Author Response

The following is the authors’ response to the original reviews.

Reviewer #1 (Recommendations For The Authors):

1) Line 99-100 The authors claimed that IQCH is a novel IQ motif-containing protein, which is essential for spermiogenesis and fertilization. However, it is not clear if the currently published paper named an ancient testis-specific IQ motif containing H gene that regulates specific transcript isoform expression during spermatogenesis.

Response: Thanks to the reviewer’s comment. Yes, IQCH is the ancient testis-specific IQ motif containing H gene. According to the reviewer’s suggestion, we have revised the statement “Here, we revealed a testis-specific IQ motif containing H gene, IQCH, which is essential for spermiogenesis and fertilization” in Introduction part of revised manuscript.

2) Line 154-159 Immunofluorescence staining for the marker of the acrosome (peanut agglutinin: PNA) as well as the mitochondrial marker (Transcription Factor A, Mitochondrial: TFAM) was performed to confirm the deficiency of the acrosomes and mitochondria in the proband's spermatozoa. It seems that the spermatozoa acrosomes and mitochondria were severely defective in the proband. The authors should indicate IQCH's role in mitochondrial and acrosome function and IQCH's role in mitochondrial and acrosome function these points by explaining how IQCH is related to mitochondrial and acrosome deficiency. In addition to staining, other functional analyses should be performed to strengthen the claim of acrosome and mitochondrial defects.

Response: We appreciate the reviewer's valuable suggestion. Indeed, in our study, the results of multiomics analysis on WT and Iqch KO testes, including LC-MS/MS analysis, proteomic analysis, and RNA-seq analysis, found a potential role of IQCH in mitochondrial and acrosome function. GO analysis of these analysis indicated a significant enrichment in mitochondrial and acrosomal functions, including acrosomal vesicle, acrosome assembly, vesicle fusion with Golgi apparatus, mitochondrion organization, mitochondrial matrix, and so on. Among the enriched molecules, in particular, HNRNPK mainly expresses at Golgi phase and Cap phase (Biggiogera et al. 1993). ANXA7 is a calcium-dependent phospholipid-binding protein that is a negative regulator of mitochondrial apoptosis (Du et al. 2015). Loss of SLC25A4 results in mitochondrial energy metabolism defects in mice (Graham et al. 1997). Furthermore, we confirmed that IQCH interacted with HNRNPK, ANXA7, and SLC25A4 through Co-IP, and exhibited downregulation in the sperm of the Iqch KO mice by immunofluorescence and western blotting. Moreover, IQCH can bind to HNRPAB, which could influence the mRNAs level of Catsper-family, such as Catsper1, Catsper2, and Catsper3, which are crucial for acrosome development (Jin ZR et al). In addition, we also detected HNRPAB binding to Dnhd1, which affects mitochondria development (Tan C et al). Therefore, in addition to staining, the other functional analyses also have provided the evidence of acrosome and mitochondrial defects caused by IQCH absence.

3) Line 180-182 IQCH knockout mice were generated. It is not clear why Mut-IQCH mice were not generated to be consistent with the human sequencing data.

Response: Thanks for reviewer’s comments. To understand IQCH's impact on fecundity in mice, we employed CRISPR-Cas9 to generate mice encoding the orthologous variant of IQCH387+1_387+10del detected in humans. Regrettably, due to sequence complexity, the designed sgRNA's specificity and efficiency were low, hindering successful Iqch knock-in mouse construction. Considering IQCH387+1_387+10del results in absent expression, we pursued Iqch knockout mice to explore IQCH's role in spermatogenesis.

4) Line 241.Figure 5A Gene Ontology (GO) analysis of the IQCH-bound proteins revealed a particular enrichment in fertilization, sperm axoneme assembly, mitochondrial organization, calcium channel, and RNA processing. But these GO functions are not shown in Figure 5A. The entire Figure 5 should be revised to enhance readability.

Response: We sincerely apologize for the oversight. These GO functions were indeed identified during the analysis of IQCH-bound proteins. Regrettably, we unintentionally omitted these GO functions when creating the plots. We have revised the plots in Figure 5 in revised manuscript to enhance readability.

5) Line 242 "33 ribosomal proteins were identified (Fig. 5B), indicating that IQCH might be involved in protein synthesis". The authors should perform an analysis to support the claim of protein synthesis defects.

Response: Thanks to reviewer’s suggestions. Initially, we have supplemented Co-IP experiments to confirm the interaction between IQCH and three ribosomal proteins (RPL4, RPS3, and RPS7), chosen from a pool of 33 ribosomal proteins based on different protein scores (Figure R1). In addition, the proteomic analysis revealed 807 upregulated proteins and 1,186 downregulated proteins in KO mice compared to WT mice. We confirmed the key downregulated proteins by western blotting and immunofluorescence staining in the previous manuscript. These results indicated that IQCH might interact with ribosomal proteins to regulate protein expression. Naturally, the regulation of protein synthesis by IQCH requires further experiments for confirmation in future studies.

Author response image 1.

The interaction between IQCH and ribosomal proteins. Co-IP assays confirmed that IQCH interacted with RPL4, RPS3, and RPS7 in WT mouse sperm.

6) Line 244 The authors mentioned too many GO functions without focus.

Response: Following reviewer’s suggestions, we have simplified IQCH-associated GO functions in the revised manuscript.

7) Figure 6, there are no negative controls in all co-IP experiments. Band sizes are not marked. Thus, all data can't be evaluated. This also raises concern about whether the LC-MS/MS experiment to identify IQCH interacting protein was well-controlled? All co-IP experiments were poorly designed to draw any conclusion.

Response: Thanks to reviewer’s comments. We have supplemented negative controls in all Co-IP experiments and provided band sizes in Figure 6 in revised manuscript.

8) The authors mentioned that IQCH can bind to CaM. But they didn't detect CaM protein in Figure 5. Did the LC-MS/MS experiment really work?

Response: Thanks to reviewer’s comments. We detected the interaction of CaM protein with IQCH in the LC-MS/MS experiment analysis, which has been submitted as new Data S1 in the revised manuscript. We also confirmed their binding in mouse sperm by Co-IP experiment and immunofluorescence staining, which results were shown in Figure 6 and Figure S10 in the previous study.

9) Figure 6D. Because IQCH is lost in Iqch KO sperm, what is the point of showing in the Co-IP assay that CaM does not bind to IQCH in Iqch KO sperm?

Response: Following reviewer’s suggestions, we have deleted the results of Co-IP assay that CaM could not bind to IQCH in Iqch KO sperm.

10) Figure 6E. The Co-IP assay does not support the authors' claim that the decreased expression of HNRPAB was due to the reduced binding of IQCH and CaM by the knockout of IQCH or CaM.

Response: Thanks to reviewer’s expert comments. Indeed, the results of Figure 6E confirmed the interaction of IQCH and CaM in K562 cells, and also showed that the expression of HNRPAB was reduced when IQCH or CaM was knocked down, suggesting that IQCH or CaM might regulate HNRPAB expression. While in Figure 6F, the downregulation of HNRPAB caused by knocking down IQCH (or CaM) cannot be rescued when overexpressed CaM (or IQCH), indicating that CaM (or IQCH) cannot mediate HNRPAB expression alone. Therefore, the reduced expression of HNRPAB in Figure 6E might result from the weakened interaction between IQCH and CaM, but not a superficial downregulation of IQCH or CaM expression. To avoid the confusion, we have modified the relevant description in the revied manuscript.

Reviewer #2 (Recommendations For The Authors):

Major comments:

1) Lines 117 and 129: Please provide the reference number (NM_xxx.x) for the IQCH isoform that was used to interpret this variant. This is key information. Also, please provide the predicted truncation consequence caused by this splicing variant to IQCH protein.

Response: Thanks to reviewer’s suggestions. We have added reference number (NM_0010317152) of IQCH in manuscript. We employed splice site prediction tools, such as SpliceAI, RDDC, and varSEAK, to assess the expression consequences of this IQCH splicing variant. These tools couldn't anticipate the outcome of this splicing variant. However, the results of minigene splicing assay showed that the IQCH c.387+1_387+10del resulted in degradation of IQCH.

2) Figure 1A: The deleted sequence indicated by the red box does not match IQCH c.387+1_387+10del. Please show a plot of the exon-intron boundary under the Sanger sequencing results of the WT allele.

Response: Thanks to reviewer’s suggestions. We are sorry for the use of non-standard descriptions about the results of Sanger sequencing. According to the HGVS nomenclature (Figure R2), we have modified the red box to match IQCH c.387+1_387+10del and have added the exon-intron boundary in Figure 1A accordingly.

Author response image 2.

HGVS nomenclature description of the IQCH variant. The picture showed a detailed HGVS nomenclature description of IQCH c.387+1_387+10del.

Minor comments:

a) Manuscript title: It is suggested to change the title to "IQCH regulates spermatogenesis by interacting with CaM to promote the expression of RNA-binding proteins".

Response: According to reviewer’s suggestions, we have modified the title as “IQCH regulates spermatogenesis by interacting with CaM to promote the expression of RNA-binding proteins”.

b) Line 116: Please introduce the abbreviation WES. Also, please introduce the other abbreviations (such as WT, SEM, TEM, etc.) the first time they appear.

Response: Thanks to reviewer’s suggestions. We have provided the full explanations for all abbreviations upon their initial appearance.

c) Line 140, "Nonfunctional IQCH": Due to "the lack of IQCH expression" in Line 137, should "Nonfunctional IQCH" be changed into "IQCH deficiency"?

Response: Thanks for reviewer’s the detailed review. We have modified this title in Results part of the revised manuscript as followed: “IQCH deficiency leads to sperm with cracked axoneme structures accompanied by defects in the acrosome and mitochondria”

d) The information on the following references is incomplete: Sechi et al., Tian et al., Wang et al., and Xu et al. Please provide issue/page/article numbers.

Response: We are sorry for our oversight. We have provided the missing issue/page/article numbers for the references.

e) The title of Figure 1: Please emphasize that the male infertile-associated variant is "homozygous".

Response: Thanks to reviewer’s suggestions. We have revised the title of Figure 1 to emphasize the homozygous variant as follows: “Identification of a homozygous splicing mutation in IQCH in a consanguineous family with male infertility”.

f) Table 1: Please provide the reference paper for the normal values. Response: We appreciate the reviewer's detailed checks. We have provided the reference paper for the normal values in Table 1.

g) Figure 5F is distorted. Please make sure that it is a perfect circle.

Response: Thanks to reviewer’s suggestions. We have revised both the graphical representation and layout of Figure 5 in revised manuscript to make sure the readability.

Reviewer #3 (Recommendations For The Authors):

While the writing is generally clear, there are multiple examples of where the writing could be improved for clarity.

1) While some terms are defined throughout the manuscript, many abbreviations are not defined upon their first mention, such as WES, RT-PCR, TYH, HTF, KSOM, KEGG, RIPA, PMSE, SDS-PAGE, H&L, and HRP.

Response: Thanks to reviewer’s suggestions. We have provided the full explanations for all abbreviations upon their initial appearance.

2) On line 44, the claim that spermatogenesis is the "most complex biological process" is rather subjective and hard to support with concrete data.

Response: Thanks to reviewer’s suggestions. We have modified this description in the Introduction section as follow: “Spermatogenesis is one of the most complex biological process in male organisms and functions to produce mature spermatozoa from spermatogonia in three phases: (i) spermatocytogenesis (mitosis), (ii) meiosis, and (iii) spermiogenesis.”

3) On line 54, I think the authors meant "heterogeneous," not "heterologous."

Response: Thanks to reviewer’s comment. We have changed “heterologous” into “heterogeneous”.

4) On line 156, I think the authors meant "deficiency," not "deficient."

Response: Thanks to reviewer’s comment. We are sorry to make this mistake. We have made the correction in the revised version of the manuscript.

5) On line 300, K562 cells are mentioned, but neither in the Methods nor the Results are any details about the biological origin of these cells (or rationale for their use other than co-expression of IQCH and CaM) provided.

Response: Thanks to reviewer’s suggestion. K562 cell line is a human leukemia cell line and is enriched in the expression of IQCH and CaM, we thus opted to use this cell line for an easier knockdown of IQCH and CaM. We have supplemented the details about the biological origin of these cells in Method section of revised manuscript.

6) For the Results section describing Figure 6H, it would be nice to provide some explanation of the results of ICHQ overexpression alone relative to control situations and not just relative to the delta-IQ version or relative to simultaneous CaM manipulation.

Response: According to the reviewer’s suggestion, we have supplemented the co-transfection of control and CaM plasmids in HEK293T cells, and the results showed that the expression of HNRPAB in cells co-transfected with control and CaM plasmids was similar to that of co-transfected with IQCH (△IQ) /CaM plasmids, but was lower than that in the cells overexpressing the WT-IQCH and CaM plasmids, confirming the nonfunction of IQCH (△IQ) plasmids. We have shown the results in Figure 6H in the revised manuscript.

7) The sentence on lines 352-354 is confusing.

Response: We apologize for any confusion caused by the sentence in question. We have revisited the sentence and made appropriate revisions to enhance its clarity as follows: “Our findings suggest that the fertilization function is the main action of IQ motif-containing proteins, while each specific IQ motif-containing protein also has its own distinct role in spermatogenesis.”

8) The use of "employee" on line 371 is awkward and not very scientific.

Response: Thanks to reviewer’s comment. We have changed “employee” in to “downstream effector protein” on line 376

Author response:

The following is the authors’ response to the original reviews.

eLife assessment

This study provides an important cell atlas of the gill of the mussel Gigantidas platifrons using a single nucleus RNA-seq dataset, a resource for the community of scientists studying deep sea physiology and metabolism and intracellular host-symbiont relationships. The work, which offers solid insights into cellular responses to starvation stress and molecular mechanisms behind deep-sea chemosymbiosis, is of relevance to scientists interested in host-symbiont relationships across ecosystems.

Public Reviews:

Reviewer #1 (Public Review):

Wang et al have constructed a comprehensive single nucleus atlas for the gills of the deep sea Bathymodioline mussels, which possess intracellular symbionts that provide a key source of carbon and allow them to live in these extreme environments. They provide annotations of the different cell states within the gills, shedding light on how multiple cell types cooperate to give rise to the emergent functions of the composite tissues and the gills as a whole. They pay special attention to characterizing the bacteriocyte cell populations and identifying sets of genes that may play a role in their interaction with the symbiotes.

Wang et al sample mussels from 3 different environments: animals from their native methane-rich environment, animals transplanted to a methane-poor environment to induce starvation, and animals that have been starved in the methane-poor environment and then moved back to the methane-rich environment. They demonstrated that starvation had the biggest impact on bacteriocyte transcriptomes. They hypothesize that the upregulation of genes associated with lysosomal digestion leads to the digestion of the intracellular symbiont during starvation, while the non-starved and reacclimated groups more readily harvest the nutrients from symbiotes without destroying them.

Strengths:

This paper makes available a high-quality dataset that is of interest to many disciplines of biology. The unique qualities of this non-model organism and the collection of conditions sampled make it of special interest to those studying deep sea adaptation, the impact of environmental perturbation on Bathymodioline mussels populations, and intracellular symbiotes. The authors do an excellent job of making all their data and analysis available, making this not only an important dataset but a readily accessible and understandable one.

The authors also use a diverse array of tools to explore their data. For example, the quality of the data is augmented by the use of in situ hybridizations to validate cluster identity and KEGG analysis provides key insights into how the transcriptomes of bacteriocytes change.

The authors also do a great job of providing diagrams and schematics to help orient non-mussel experts, thereby widening the audience of the paper.

Thank the reviewer for the valuable feedback on our study. We are grateful that the reviewers found our work to be interesting and we appreciate their thorough evaluation of our research. Their constructive comments will be considered as we continue to develop and improve our study.

Weaknesses:

One of the main weaknesses of this paper is the lack of coherence between the images and the text, with some parts of the figures never being referenced in the body of the text. This makes it difficult for the reader to interpret how they fit in with the author's discussion and assess confidence in their analysis and interpretation of data. This is especially apparent in the cluster annotation section of the paper.

We appreciate the feedback and suggestions provided by the reviewer, and we have revised our manuscript to make it more accessible to general audiences.

Another concern is the linking of the transcriptomic shifts associated with starvation with changes in interactions with the symbiotes. Without examining and comparing the symbiote population between the different samples, it cannot be concluded that the transcriptomic shifts correlate with a shift to the 'milking' pathway and not other environmental factors. Without comparing the symbiote abundance between samples, it is difficult to disentangle changes in cell state that are due to their changing interactions with the symbiotes from other environmental factors.

We are grateful for the valuable feedback and suggestions provided by the reviewer. Our keen interest lies in understanding symbiont responses, particularly at the single-cell level. However, it's worth noting that existing commercial single-cell RNA-seq technologies rely on oligo dT priming for reverse transcription and barcoding, thus omitting bacterial gene expression information from our dataset. We hope that advancements in technology will soon enable us to perform an integrated analysis encompassing both host and symbiont gene expression.

Additionally, conclusions in this area are further complicated by using only snRNA-seq to study intracellular processes. This is limiting since cytoplasmic mRNA is excluded and only nuclear reads are sequenced after the organisms have had several days to acclimate to their environment and major transcriptomic shifts have occurred.

We appreciate the comments shared by the reviewer and agree that scRNA-seq provides more comprehensive transcriptional information by targeting the entire mRNA of the cell. However, we would like to highlight that snRNA-seq has some unique advantages over scRNA-seq. Notably, snRNA-seq allows for simple snap-freezing of collected samples, facilitating easier storage, particularly for samples obtained during field trips involving deep-sea animals and other ecologically significant non-model animal samples. Additionally, unlike scRNA-seq, snRNA-seq eliminates the need for tissue dissociation, which often involves prolonged enzymatic treatment of deep-sea animal tissue/cells under atmospheric pressure. This process can potentially lead to the loss of sensitive cells or alterations in gene expression. Moreover, snRNA-seq procedures disregard the size and shape of animal cells, rendering it a superior technology for constructing the cell atlas of animal tissues. Consequently, we assert that snRNA-seq offers flexibility and represents a suitable choice for the research objects of our current research.

Reviewer #2 (Public Review):

Wang, He et al. shed insight into the molecular mechanisms of deep-sea chemosymbiosis at the single-cell level. They do so by producing a comprehensive cell atlas of the gill of Gigantidas platifrons, a chemosymbiotic mussel that dominates the deep-sea ecosystem. They uncover novel cell types and find that the gene expression of bacteriocytes, the symbiont-hosting cells, supports two hypotheses of host-symbiont interactions: the "farming" pathway, where symbionts are directly digested, and the "milking" pathway, where nutrients released by the symbionts are used by the host. They perform an in situ transplantation experiment in the deep sea and reveal transitional changes in gene expression that support a model where starvation stress induces bacteriocytes to "farm" their symbionts, while recovery leads to the restoration of the "farming" and "milking" pathways.

A major strength of this study includes the successful application of advanced single-nucleus techniques to a non-model, deep-sea organism that remains challenging to sample. I also applaud the authors for performing an in situ transplantation experiment in a deep-sea environment. From gene expression profiles, the authors deftly provide a rich functional description of G. platifrons cell types that is well-contextualized within the unique biology of chemosymbiosis. These findings offer significant insight into the molecular mechanisms of deep-sea host-symbiont ecology, and will serve as a valuable resource for future studies into the striking biology of G. platifrons.

The authors' conclusions are generally well-supported by their results. However, I recognize that the difficulty of obtaining deep-sea specimens may have impacted experimental design. In this area, I would appreciate more in-depth discussion of these impacts when interpreting the data.

Thank the reviewer for their valuable feedback on our study. We're grateful that the reviewers found our work interesting, and we appreciate their thorough evaluation of our research. We'll consider their constructive comments as we continue to develop and improve our study.

Because cells from multiple individuals were combined before sequencing, the in situ transplantation experiment lacks clear biological replicates. This may potentially result in technical variation (ie. batch effects) confounding biological variation, directly impacting the interpretation of observed changes between the Fanmao, Reconstitution, and Starvation conditions. It is notable that Fanmao cells were much more sparsely sampled. It appears that fewer cells were sequenced, resulting in the Starvation and Reconstitution conditions having 2-3x more cells after doublet filtering. It is not clear whether this is due to a technical factor impacting sequencing or whether these numbers are the result of the unique biology of Fanmao cells. Furthermore, from Table S19 it appears that while 98% of Fanmao cells survived doublet filtering, only ~40% and ~70% survived for the Starvation and Reconstitution conditions respectively, suggesting some kind of distinction in quality or approach.

There is a pronounced divergence in the relative proportions of cells per cell type cluster in Fanmao compared to Reconstitution and Starvation (Fig. S11). This is potentially a very interesting finding, but it is difficult to know if these differences are the expected biological outcome of the experiment or the fact that Fanmao cells are much more sparsely sampled. The study also finds notable differences in gene expression between Fanmao and the other two conditions- a key finding is that bacteriocytes had the largest Fanmao-vs-starvation distance (Fig. 6B). But it is also notable that for every cell type, one or both comparisons against Fanmao produced greater distances than comparisons between Starvation and Reconstitution (Fig. 6B). Again, it is difficult to interpret whether Fanmao's distinctiveness from the other two conditions is underlain by fascinating biology or technical batch effects. Without biological replicates, it remains challenging to disentangle the two.

As highlighted by the reviewer, our experimental design involves pooling multiple biological samples within a single treatment state before sequencing. We acknowledge the concern regarding the absence of distinct biological replicates and the potential impact of batch effects on result interpretation. While we recognize the merit of conducting multiple sequencing runs for a single treatment to provide genuine biological replicates, we contend that batch effects may not exert a strong influence on the observed patterns.

In addition, we applied a bootstrap sampling algorithm to assess whether the gene expression patterns within a cluster are more similar than those between clusters. This algorithm involves selecting a portion of cells per cluster and examining whether this subset remains distinguishable from other clusters. Our assumption was that if different samples exhibited distinct expression patterns due to batch effect, the co-assignment probabilities of a cluster would be very low. This expectation was not met in our data, as illustrated in Fig. S2. The lack of significantly low co-assignment probabilities within clusters suggests that batch effects may not exert a strong influence on our results.

Indeed, we acknowledge a noticeable shift in the expression patterns of certain cell types, such as the bacteriocyte. However, this is not universally applicable across all cell types. For instance, the UMAP figure in Fig. 6A illustrates a substantial overlap among basal membrane cell 2 from Fanmao, Starvation, and Reconstitution treatments, and the centroid distances between the three treatments are subtle, as depicted in Fig. 6B. This consistent pattern is also observed in DEPC, smooth muscle cells, and the food groove ciliary cells.

The reviewer also noted variations in the number of cells per treatment. Specifically, Fanmao sequencing yielded fewer than 10 thousand cells, whereas the other two treatments produced 2-3 times more cells after quality control (QC). It is highly probable that the technician loaded different quantities of cells into the machine for single-nucleus sequencing—a not uncommon occurrence in this methodology. While loading more cells may increase the likelihood of doublets, it is crucial to emphasize that this should not significantly impact the expression patterns post-QC. It's worth noting that overloading samples has been employed as a strategic approach to capture rare cell types, as discussed in a previous study (reference: 10.1126/science.aay0267).

The reviewer highlighted the discrepancy in cell survival rates during the 'doublet filtering' process, with 98% of Fanmao cells surviving compared to approximately 40% and 70% for the Starvation and Reconstitution conditions, respectively. It's important to clarify that the reported percentages reflect the survival of cells through a multi-step QC process employing various filtering strategies.

Post-doublet removal, we filtered out cells with <100 or >2500 genes and <100 or >6000 unique molecular identifiers (UMIs). Additionally, genes with <10 UMIs in each data matrix were excluded. The observed differences in survival rates for Starvation and Reconstitution cells can be attributed to the total volume of data generated in Illumina sequencing. Specifically, we sequenced approximately 91 GB of data for Fanmao, ~196 GB for Starvation, and ~249 GB for Reconstitution. As a result, the qualified data obtained for Starvation and Reconstitution conditions was only about twice that of Fanmao due to the limited data volume.

The reviewer also observed a divergence in the relative proportions of cells per cell type cluster in Fanmao compared to Reconstitution and Starvation, as depicted in Fig. S1. This discrepancy may hold true biological significance, presenting a potentially intriguing finding. However, our discussion on this pattern was rather brief, as we acknowledge that the observed differences could be influenced by the sample preparation process for dissection and digestion. It is crucial to consider that cutting a slightly different area during dissection may result in variations in the proportion of cells obtained. While we recognize the potential impact of this factor, we do not think that the sparsity of sampling alone could significantly affect the relative proportions of cells per cell type.

In conclusion, we acknowledge the reviewer's suggestion that sequencing multiple individual samples per treatment condition would have been ideal, rather than pooling them together. However, the homogenous distribution observed in UMAP and the consistent results obtained from bootstrap sampling suggest that the impact of batch effects on our analyses is likely not substantial. Additionally, based on our understanding, the smaller number of cells in the Fanmao sample should not have any significant effect on the resulting different proportion of cells or the expression patterns per each cluster.

Reviewer #3 (Public Review):

Wang et al. explored the unique biology of the deep-sea mussel Gigantidas platifrons to understand the fundamental principles of animal-symbiont relationships. They used single-nucleus RNA sequencing and validation and visualization of many of the important cellular and molecular players that allow these organisms to survive in the deep sea. They demonstrate that a diversity of cell types that support the structure and function of the gill including bacteriocytes, specialized epithelial cells that host sulfur-oxidizing or methane-oxidizing symbionts as well as a suite of other cell types including supportive cells, ciliary, and smooth muscle cells. By performing experiments of transplanting mussels from one habitat which is rich in methane to methane-limited environments, the authors showed that starved mussels may consume endosymbionts versus in methane-rich environments upregulated genes involved in glutamate synthesis. These data add to the growing body of literature that organisms control their endosymbionts in response to environmental change.

The conclusions of the data are well supported. The authors adapted a technique that would have been technically impossible in their field environment by preserving the tissue and then performing nuclear isolation after the fact. The use of single-nucleus sequencing opens the possibility of new cellular and molecular biology that is not possible to study in the field. Additionally, the in-situ data (both WISH and FISH) are high-quality and easy to interpret. The use of cell-type-specific markers along with a symbiont-specific probe was effective. Finally, the SEM and TEM were used convincingly for specific purposes in the case of showing the cilia that may support water movement.

We appreciate the valuable feedback provided by the reviewer on our study. It is encouraging to know that our work was found to be interesting and that they conducted a thorough evaluation of our research. We will take their constructive comments into account as we strive to develop and enhance our study. Thank the reviewer for all the input.

The one particular area for clarification and improvement surrounds the concept of a proliferative progenitor population within the gill. The authors imply that three types of proliferative cells within gills have long been known, but their study may be the first to recover molecular markers for these putative populations. The markers the authors present for gill posterior end budding zone cells (PEBZCs) and dorsal end proliferation cells (DEPCs) are not intuitively associated with cell proliferation and some additional exploration of the data could be performed to strengthen the argument that these are indeed proliferative cells. The authors do utilize a trajectory analysis tool called Slingshot which they claim may suggest that PEBZCs could be the origin of all gill epithelial cells, however, one of the assumptions of this analysis is that differentiated cells are developed from the same precursor PEBZC population.

However, these conclusions do not detract from the overall significance of the work of identifying the relationship between symbionts and bacteriocytes and how these host bacteriocytes modulate their gene expression in response to environmental change. It will be interesting to see how similar or different these data are across animal phyla. For instance, the work of symbiosis in cnidarians may converge on similar principles or there may be independent ways in which organisms have been able to solve these problems.

We are grateful for the valuable comments and suggestions provided by the reviewer. All suggestions have been carefully considered, and the manuscript has been revised accordingly. We particularly value the reviewer's insights regarding the characterization of the G. platifrons gill proliferative cell populations. In a separate research endeavor, we have conducted experiments utilizing both cell division and cell proliferation markers on these proliferative cell populations. While these results are not incorporated into the current manuscript, we would be delighted to share our preliminary findings with the reviewer. Our preliminary results indicate that the proliferative cell populations exhibit positivity for cell proliferation markers and contain a significant number of mitotic cells..

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Further experiments are needed to link the changes in transcriptomes of Bathymodioline mussels in the different environmental conditions to changes in their interactions with symbiotes. For example, quantifying the abundance and comparing the morphology of symbiotes between the environmental conditions would lend much support for shifting between milking and farming strategies. Without analyzing the symbiotes and comparing them across populations, it is difficult to comment on the mechanisms of interactions between symbiotes and the hosts. Without this analysis, this data is better suited towards comments about the general effect of environmental perturbation and stress on gene expression in these mussels.

We appreciate the reviewer’s comments. We are also very curious about the symbiont responses, especially at the single-cell level. However, all the current commercial single-cell RNA-seq technologies are based on oligo dT priming for reverse transcription and barcoding. Therefore, the bacterial gene expression information is omitted from our dataset. Hopefully, with the development of technology, we could conduct an integrated analysis of both host and symbiont gene expression soon.

Additionally, clarification is needed on which types of symbiotes are being looked at. Are they MOX or SOX populations? Are they homogenous? What are the concentrations of sulfur at the sampled sites?

We thank you for your valuable comments and suggestions. Gigantidas platifrons harbors a MOX endosymbiont population characterized by a single 16S rRNA phylotype. We apologize for any confusion resulting from our previous wording. To clarify, we have revised lines 57-59 of our introduction

In the text and images, consider using standardized gene names and leaving out the genome coordinates. This would greatly help with readability. Also, be careful to properly follow gene naming and formatting conventions (ie italicizing gene names and symbols).

We appreciate the reviewer’s insightful comments. In model animals, gene nomenclature often stems from forward genetic approaches, such as the identification of loss-of-function mutants. These gene names, along with their protein products, typically correspond to unique genome coordinates. Conversely, in non-model invertebrates (e.g., Gigantidas platifrons of present study), gene prediction relies on a combination of bioinformatics methods, including de novo prediction, homolog-based prediction, and transcriptomics mapping. Subsequently, the genes are annotated by identifying their best homologs in well-characterized databases. Given that different genes may encode proteins with similar annotated functions, we chose to include both the gene ID (genome coordinates) and the gene name in our manuscript. This dual labeling approach ensures that our audience receives accurate and comprehensive information regarding gene identification and annotation.

Additionally, extending KEGG analysis to the atlas annotation section could help strengthen the confidence of annotations. For example, when identifying bacteriocyte populations, the functional categories of individual marker genes (lysosomal proteases, lysosomal traffic regulators, etc) are used to justify the annotation. Presenting KEGG support that these functional categories are upregulated in this population relative to others would help further support how you characterize this cluster by showing it's not just a few specific genes that are enriched in this cell group, but rather an overall functionality.

We appreciate the valuable suggestion provided by the reviewer. Indeed, incorporating KEGG analysis into the atlas annotation section could further enhance the confidence in our annotations. However, in our study, we encountered some limitations that impeded us from conducting a comprehensive KEGG enrichment analysis.

Firstly, the number of differentially expressed genes (DEGs) that we identified for certain cell populations was relatively small, making it challenging to meet the threshold required for meaningful KEGG enrichment analysis. For instance, among the 97 marker genes identified for the Bacteriocyte cluster, only two genes, Bpl_scaf_59648-4.5 (lysosomal alpha-glucosidase-like) and Bpl_scaf_52809-1.6 (lysosomal-trafficking regulator-like isoform X1), were identified as lysosomal genes. To generate reliable KEGG enrichments, a larger number of genes is typically required.

Secondly, single-nucleus sequencing, as employed in our study, tends to yield a relatively smaller number of genes per cell compared to bulk RNA sequencing. This limited gene yield can make it challenging to achieve sufficient gene representation for rigorous KEGG enrichment analysis.

Furthermore, many genes in the genome still lack comprehensive annotation, both in terms of KEGG and GO annotations. In our dataset, out of the 33,584 genes obtained through single-nuclei sequencing, 26,514 genes have NO KEGG annotation, and 25,087 genes have NO GO annotation. This lack of annotations further restricts the comprehensive application of KEGG analysis in our study.

The claim that VEPCs are symbiote free is not demonstrated. Additional double in situs are needed to show that markers of this cell type localize in regions free of symbiotes.

We appreciate your comments and suggestions. In Figure 5B, our results demonstrate that the bacteriocytes (green fluorescent signal) are distant from the VEPCs, which are located around the tip of the gill filaments (close to the food groove). We have revised our Figure 5B to make it clear.

Additionally, it does not seem like trajectory analysis is appropriate for these sampling conditions. Generally, to create trajectories confidently, more closely sampled time points are needed to sufficiently parse out the changes in expression. More justification is needed for the use of this type of analysis here and a discussion of the limitations should be mentioned, especially when discussing the hypotheses relating to PEBZCs, VEPCs, and DEPCs.

We greatly appreciate your thoughtful commentary. It is important to acknowledge that in the context of a developmental study, incorporating more closely spaced time points indeed holds great value. In our ongoing project investigating mouse development, for instance, we have implemented time points at 24-hour intervals. However, in the case of deep-sea adult animals, we hypothesized a slower transcriptional shift in such extreme environment, which led us to opt for a time interval of 3-7 days. Examining the differential expression profiles among the three treatments, we observed that most cell types exhibited minimal changes in their expression profiles. For the cell types strongly impacted by in situ transplantation, their expression profiles per cell type still exhibited highly overlap in the UMAP analysis (Figure 6a), thus enabling meaningful comparisons. Nevertheless, we recognize that our sampling strategy may not be flawless. Additionally, the challenging nature of conducting in situ transplantation in 1000-meter depths limited the number of sampling occasions available to us. We sincerely appreciate your input and understanding.

Finally, more detail should be added on the computational methods used in this paper. For example, the single-cell genomics analysis protocol should be expanded on so that readers unfamiliar with BD single-cell genomics handbooks could replicate the analysis. More detail is also needed on what criteria and cutoffs were used to calculate marker genes. Also, please be careful to cite the algorithms and software packages mentioned in the text.

Acknowledged, thank you for highlighting this. In essence, the workflow closely resembles that of the 10x Genomics workflow (despite the use of a different software, i.e., Cell Ranger). We better explain the workflow below, and also noting that this information may no longer be relevant for newer users of BD or individuals who are not acquainted with BD, given that the workflow underwent a complete overhaul in the summer of 2023.

References to lines

Line 32: typo "..uncovered unknown tissue heterogeny" should read "uncovering" or "and uncovered")

Overall abstract could include more detail of findings (ex: what are the "shifts in cell state" in line 36 that were observed)

We apologize for the mistakes, and have revised the manuscript accordingly.

Line 60: missing comma "...gill filament structure, but also"

We apologize for the mistakes, and have revised the manuscript accordingly.

Line 62-63: further discussion here, or in the relevant sections of the specific genes identified in the referenced bulk RNA-seq project could help strengthen confidence in annotation

We appreciate the comment, and have revised the manuscript accordingly.

Line 112: what bootstrapping strategy? Applied to what?

This is a bootstrap sampling algorithm to assess the robustness of each cell cluster developed in a recent biorxiv paper. (Singh, P. & Zhai, Y. Deciphering Hematopoiesis at single cell level through the lens of reduced dimensions. bioRxiv, 2022.2006.2007.495099 (2022). https://doi.org:10.1101/2022.06.07.495099)

Lines 127-129: What figures demonstrate the location of the inter lamina cells? Are there in situs that show this?

We apologize for any errors; the referencing of figures in the manuscript has been revised for clarity

Lines 185-190: does literature support these as markers of SMCs? Are they known smooth muscle markers in other systems?

We characterized the SMCs by the expression of LDL-associated protein, angiotensin-converting enzyme-like protein, and the "molecular spring" titin-like protein, all of which are commonly found in human vascular smooth muscle cells. Based on this analysis, we hypothesize that these cells belong to the smooth muscle cell category.

Line 201: What is meant by "regulatory roles"?

In this context, we are discussing the expression of genes encoding regulatory proteins, such as SOX transcription factors and secreted-frizzled proteins.

Line 211: which markers disappeared? What in situs show this?

We apologize for the mistakes, and have revised the manuscript accordingly.

Line 211: typo, "role" → "roll"

We apologize for the mistakes, and have revised the manuscript accordingly.

Line 214: what are these "hallmark genes"

We apologize for the mistakes, here we are referring to the genes listed in figure 4B. We have revised the manuscript accordingly.

Line 220: are there meristem-like cells in metazoans? If so, this would be preferable to a comparison with plants.

In this context, we are discussing the morphological characteristics of gill proliferative cell populations found in filibranch bivalves. These populations, namely PEPC, VEPC, and DEPC, consist of cells exhibiting morphological traits akin to those of plant cambial-zone meristem cells. These cells typically display small, round shapes with a high nucleus-to-plasma ratio. We acknowledge that while these terms are utilized in bivalve studies (citations below), they lack the robust support seen in model systems backed by molecular biology evidences. The present snRNA-seq data, however, may offer valuable cell markers for future comprehensive investigations.

Leibson, N. L. & Movchan, O. T. Cambial zones in gills of Bivalvia. Mar. Biol. 31, 175-180 (1975). https://doi.org:10.1007/BF00391629

Wentrup, C., Wendeberg, A., Schimak, M., Borowski, C. & Dubilier, N. Forever competent: deep-sea bivalves are colonized by their chemosynthetic symbionts throughout their lifetime. Environ. Microbiol. 16, 3699-3713 (2014). https://doi.org:10.1111/1462-2920.12597

Cannuel, R., Beninger, P. G., McCombie, H. & Boudry, P. Gill Development and its functional and evolutionary implications in the blue mussel Mytilus edulis (Bivalvia: Mytilidae). Biol. Bull. 217, 173-188 (2009). https://doi.org:10.1086/BBLv217n2p173

Line 335: what is slingshot trajectory analysis? Does this differ from the pseudotime analysis?

Slingshot is an algorithm that uses the principal graph of the cells to infer trajectories. It models trajectories as curves on the principal graph, capturing the progression and transitions between different cellular states.

Both Slingshot and pseudotime aim to infer cellular trajectories. Slingshot focuses on capturing branching patterns which is fully compatible with the graph generated using dimensionality reduction such as UMAP and PHATE, while pseudotime analysis aims to order cells along a continuous trajectory. It does not rely on dimensionality reduction graphs. We used both in the MS for different purposes.

Line 241: introduce FISH methodology earlier in the paper, when in situ images are first referenced

We appreciate the comment, and have revised the manuscript accordingly.

Line 246-249: can you quantify the decrease in signal or calculate the concentration of symbiotes in the cells? Was 5C imaged whole? This can impact the fluorescent intensity in tissues of different thicknesses.

We appreciate your comment. In Figure 5C, most of the typical gill filament region is visible (the ventral tip of the gill filament, and the mid part of the gill filament) except for the dorsal end. The gill filament of bathymodioline mussels exhibits a simple structure: a single layer of bacteriocytes grow on the basal membrane. Consequently, the gill slices have a fairly uniform thickness (with two layers of bacteriocytes and one layer of interlamina cells in between), minimizing any potential impact on fluorescent intensity. As of now, detailed quantification of intracellular symbionts may necessitate continuous TEM or ultra-resolution confocal sections to 3D reconstruct the bacteriocytes, which may exceed the scope of the current study. Therefore, fluorescent intensity remains the only method available to us for estimating bacterial density/distribution across the gill filament.

Line 249: What is meant by 'environmental gradient?'

Here we are refereeing the gases need for symbiont’s chemosynthesis. We have revised the manuscript to make it clear.

Lines 255-256: Were the results shown in the TEM images previously known? Not clear what novel information is conveyed in images Fig 5 C and D

In the Fig 5 C and D, we’ve delivered a high-quality SEM TEM image of a typical bacteriocyte, showcasing its morphology and subcellular machinery with clarity. These electron microscopy images offer the audience a comprehensive introduction to the cellular function of bacteriocytes. Additionally, they serve as supportive evidence for the bacteriocytes' snRNA-seq data.

Line 295-296: Can you elaborate on what types of solute carrier genes have been shown to be involved with symbioses?

We appreciate the comment, and have revised the manuscript accordingly. The putative functions of the solute carriers could be found in Figure 5I.

Line 297-301: Which genes from the bulk RNA-seq study? Adding more detail and references in cluster annotation would help readers better understand the justifications.

We appreciate the comment, and have revised the manuscript accordingly.

Line 316 -322: Can you provide the values of the distances?

We also provide values in the main text, in addition to the Fig6b. We also provide a supplementary Table (Supplementary Table S19).

Line 328: What are the gene expression patterns?

We observed genes that are up- and down-regulated in Starvation and reconstitution.

LIne 334-337: A visualization of the different expression levels of the specific genes in clusters between sites might be helpful to demonstrate the degree of difference between sites.

We have prepared a new supplementary file showing the different expression levels.

Line 337: Citation needed

We appreciate the comment. Here, we hypothesize the cellular responds based on the gene’s function and their expression patterns.

Line 402-403: Cannot determine lineages from data presented. Need lineage tracing over time to determine this

We acknowledge the necessity of conducting lineage tracing over time to validate this hypothesis. Nonetheless, in practical terms, it is difficult to obtain samples for testing this. Perhaps, it is easier to use their shallow sea relatives to test this hypothesis. However, in practice, it is very difficult.

413-414: What are the "cell-type specific responses to environmental change"? It could be interesting to present these results in the "results and discussion" section

These results are shown in Supplementary Figure S8.

Line 419-424: Sampling details might go better earlier on in the paper, when the sampling scheme is introduced.

We appreciate the comments. Here, we are discussing the limitations of our current study, not sampling details.

Line 552: What type of sequencing? Paired end? How long?

We conducted 150bp paired-end sequencing.

556-563: More detail here would be useful to readers not familiar with the BD guide. Also be careful to cite the software used in analysis!

The provided guide and handbook elucidate the intricacies of gene name preparation, data alignment to the genome, and the generation of an expression matrix. It is worth mentioning that we relied upon outdated versions of the aforementioned resources during our data analysis phase, as they were the only ones accessible to us at the time. However, we have since become aware of a newer pipeline available this year, rendering the information presented here of limited significance to other researchers utilizing BD.

Many thanks for your kind reminding. We have now included a reference for STAR. All other software was cited accordingly. There are no scholarly papers or publications to refer to for the BD pipeline that we can cite.

Line 577-578: How was the number of clusters determined? What is meant by "manually combine the clusters?" If cells were clustered by hand, more detail on the method is needed, as well as direct discussion and justification in the body of the paper.

It would be more appropriate to emphasize the determination of cell types rather than clusters. The clusters were identified using a clustering function, as mentioned in the manuscript. It's important to note that the clustering function (in our case, the FindClusters function of Seurat) provides a general overview based on diffuse gene expression. Technically speaking, there is no guarantee that one cluster corresponds to a single cell type. Therefore, it is crucial to manually inspect the clustering results to assign clusters to the appropriate cell types. In some cases, multiple clusters may be assigned to the same cell type, while in other cases, a single cluster may need to be further subdivided into two or more cell types or sub-cell types, depending on the specific circumstances.

For studies conducted on model species such as humans or mice, highly and specifically expressed genes within each cluster can be compared to known marker genes of cell types mentioned in previous publications, which generally suffices for annotation purposes. However, in the case of non-model species like Bathymodioline mussels, there is often limited information available about marker genes, making it challenging to confidently assign clusters to specific cell types. In such situations, in situ hybridisation proves to be incredibly valuable. In our study, WISH was employed to visualise the expression and morphology of marker genes within clusters. When WISH revealed the expression of marker genes from a cluster in a specific type of cell, we classified that cluster as a genuine cell type. Moreover, if WISH demonstrated uniform expression of marker genes from different clusters in the same cell, we assigned both clusters to the same cell type.

We expanded the description of the strategy in the Method section.

LIne 690-692: When slices were used, what part of the gill were they taken from?

We sectioned the gill around the mid part which could represent the mature bacteriocytes.

References to figures:

General

Please split the fluorescent images into different channels with an additional composite. It is difficult to see some of the expression patterns. It would also make it accessible to colorblind readers.

We appreciate the comments and suggestions from the reviewer. We have converted our figures to CMYK colour which will help the colorblind audiences to read our paper.

Please provide the number of replicates for each in situ and what proportion of those displayed the presented pattern.

We appreciate the reviewer’s comments. We have explained in the material and methods part of the manuscript.

Figure 2.C' is a fantastic summary and really helps the non-mussel audience understand the results. Adding schematics like this to Figures 3-5 would be helpful as well.

We value the reviewer's comments. We propose that Figures 3K, 4C, and 5A-D could offer similar schematic explanations to assist the audience.

Figure 2:

Figures 2.C-F, 2.C', 2.H-J are not referenced in the text. Adding in discussions of them would help strengthen your discussions on the cluster annotation

We appreciate the reviewer's comments. We have revise the manuscript accordingly.

In 2.B. 6 genes are highlighted in red and said to be shown in in situs, but only 5 are shown.

We apology for the mistake. We didn’t include the result 20639-0.0 WISH in present study. We have changed the label to black.

Figure 3:

FIg 2C-E not mentioned.

We appreciate the reviewer's comments. We have revise the manuscript accordingly.

In 3.B 8 genes are highlighted in red and said to be shown in in situs. Only 6 are.

The result of the WISH were provided in Supplementary Figures S4 and S5.

FIgure 3.K is not referenced in the legend.

We appreciate the comment, and have revised the manuscript accordingly.

Figure 4:

In Figure D, it might be helpful to indicate the growth direction.

We appreciate the comment, and have revised the manuscript accordingly by adding an arrow in panel D to indicate growth direction.

4F: A double in situ with the symbiote marker is needed to demonstrate the nucleolin-like positive cells are symbiote free.

We appreciate the comment. The symbiont free region could be found in Figure 5A.

Figure 5:

In 5.A, quantification of symbiote concentration would help support your conclusion that they are denser around the edges.

We appreciate the comment, as we mentioned above, detailed quantification of intracellular symbionts may necessitate continuous TEM or ultra-resolution confocal sections to 3D reconstruct the bacteriocytes, which may exceed the scope of the current study. Therefore, fluorescent intensity remains the only method available to us for estimating bacterial density/distribution across the gill filament.

In 5.D, the annotation is not clear. Adding arrows like in 5.C would be helpful.

We appreciate the comment, and have revised the manuscript accordingly.

A few genes in 5.F are not mentioned in the paper body when listing other genes. Mentioning them would help provide more support for your clustering.

We appreciate the comment, and have revised the manuscript accordingly.

Is 5.I meant to be color coded with the gene groups from 5.F? Color Coding the gene names, rather than organelles or cellular structures might portray this better and help visually strengthen the link between the diagram and your dot plot.

We appreciate the suggestions. We've experimented with color-coding the gene names, but some colors are less discernible against a white background.

Figure 6:

6.B Is there a better way to visualize this data? The color coding is confusing given the pairwise distances. Maybe heatmaps?

We attempted a heatmap, as shown in the figure below. However, all co-authors agree that a bar plot provides clearer visualization compared to the heatmap. We agree that the color scheme maya be confusing because they use the same color as for individual treatment. So we change the colors.

Author response image 1.

Figure 6.D: Why is the fanmao sample divided in the middle?

Fig6C show that single-cell trajectories include branches. The branches occur because cells execute alternative gene expression programs. Thus, in Fig 6D, we show changes for genes that are significantly branch dependent in both lineages at the same time. Specifically, in cluster 2, the genes are upregulated during starvation but downregulated during reconstitution. Conversely, genes in cluster 1 are downregulated during starvation but upregulated during reconstitution. It's of note that Fig 6D displays only a small subset of significantly branch-dependent genes.

FIgure 6.D: Can you visualize the expression in the same format as in figures 2-5?

We appreciate the comments from the reviewer. As far as we know, this heatmap are the best format to demonstrate this type of gene expression profile.

Supplementary Figure S2:

Please provide a key for the cell type abbreviations

We appreciate the comment, and have added the abbreviations of cell types accordingly.

Supplementary Figures S4 and S5:

What part of the larger images are the subsetted image taken from?

We appreciate the comment, these images were taken from the ventral tip and mid of the gill slices, respectively. We have revised the figure legends to make it clear.

Supplemental Figure S7:

If clusters 1 and 2 show genes up and downregulated during starvation, what do clusters 4 and 3 represent?

Cluster 1: Genes that are obviously upregulated during Starvation, and downregulated during reconstitution; luster4： genes are downregulated during reconstitution but not obviously upregulated during Starvation.

Cluster 2 show genes upregulated during reconstitution, and cluster 3 obviously downregulated during Starvation.

Author response table 1.

Supplemental Figure S8:

This is a really interesting figure that I think shows some of the results really well! Maybe consider moving it to the main figures of the paper?

We appreciate the comments and suggestions. We concur with the reviewer on the significance of the results presented. However, consider the length of this manuscript, we have prioritized the inclusion of the most pertinent information in the main figures. Supplementary materials containing additional figures and details on the genes involved in these pathways are provided for interested readers.

Supplemental Figure S11:

Switching the axes might make this image easier for the reader to interpret. Additionally, calculating the normalized contribution of each sample to each cluster could help quantify the extent to which bacteriocytes are reduced when starving.

Thank you for the insightful suggestion, which we have implemented as detailed below. We acknowledge the importance of understanding the changes in bacteriocyte proportions across different treatments. However, it's crucial to note that the percentage of cells per treatment is highly influenced by factors such as the location of digestion and sequencing, as previously mentioned.

Author response image 2.

Reviewer #2 (Recommendations For The Authors):

The following are minor recommendations for the text and figures that may help with clarity:

Fig. 3K: This figure describes water flow induced by different ciliary cells. It is not clear what the color of the arrows corresponds to, as they do not match the UMAP (i.e. the red arrow) and this is not indicated in the legend. Are these colours meant to indicate the different ciliary cell types? If so it would be helpful to include this in the legend.

We appreciate the reviewer's comments and suggestions. The arrows indicate the water flow that might be agitated by the certain types of cilium. We have revised our figure and figure legends to make it clear.

Line 369: The incorrect gene identifier is given for the mitochondrial trifunctional enzyme. This gene identifier is identical to the one given in line 366, which describes long-chain-fatty-acid-ligase ACSBG2-like (Bpl_scaf_28862-1.5).

We appreciate the reviewer's comments and suggestions. We have revised our manuscript accordingly.

Line 554: The Bioproject accession number (PRJNA779258) does not appear to lead to an existing page in any database.

We appreciate the reviewer's comments and suggestions. We have released this Bioproject to the public.

Line 597-598: it would be helpful to know the specific number of cells that the three sample types were downsampled to, and the number of cells remaining in each cluster, as this can affect the statistical interpretation of differential expression analyses.

The number of cells per cluster in our analysis ranged from 766 to 14633. To mitigate potential bias introduced by varying cell numbers, we implemented downsampling, restricting the number of cells per cluster to no more than 3500. This was done to ensure that the differences between clusters remained less than 5 times. We experimented with several downsampling strategies, exploring cell limits of 4500 and 2500, and consistently observed similar patterns across these variations.

Data and code availability:

The supplementary tables and supplementary data S1 appear to be the final output of the differential expression analyses. Including the raw data (e.g. reads) and/or intermediate data objects (e.g. count matrices, R objects), in addition to the code used to perform the analyses, may be very helpful for replication and downstream use of this dataset. As mentioned above, the Bioproject accession number appears to be incorrect.

We appreciate the reviewer's comments and suggestions. Regarding our sequencing data, we have deposited all relevant information with the National Center for Biotechnology Information (NCBI) under Bioproject PRJNA779258. Additionally, we have requested the release of the Bioproject. Furthermore, as part of this round of revision, we have included the count matrices for reference.

Reviewer #3 (Recommendations For The Authors):

As noted in the public review, my only major concerns are around the treatment of progenitor cell populations. I am sympathetic to the challenges of these experiments but suggest a few possible avenues to the authors.

First, there could be some demonstration that these cells in G. platifrons are indeed proliferative, using EdU incorporation labeling or a conserved epitope such as the phosphorylation of serine 10 in histone 3. It appears in Mytilus galloprovincialis that proliferating cell nuclear antigen (PCNA) and phospho-histone H3 have previously been used as good markers for proliferative cells (Maiorova and Odintsova 2016). The use of any of these markers along with the cell type markers the authors recover for PEBZCs for example would greatly strengthen the argument that these are proliferative cells.

If performing these experiments would not be currently possible, the authors could use some computation approaches to strengthen their arguments. Based on conserved cell cycle markers and the use of Cell-Cycle feature analysis in Seurat could the authors provide evidence that these progenitors occupy the G2/M phase at a greater percentage than other cells? Other than the physical position of the cells is there much that suggests that these are proliferative? While I am more convinced by markers in VEPCs the markers for PEBZCs and DEPCs are not particularly compelling.

While I do not think the major findings of the paper hinge on this, comments such as "the PBEZCs gave rise to new bacteriocytes that allowed symbiont colonization" should be taken with care. It is not clear that the PBEZCs are proliferative and there does not seem to be any direct evidence that PBEZCs (or DEPCs or VEPCS for that manner) are the progenitor cells through any sort of labeling or co-expression studies.

We appreciate the comments and suggestions from the reviewer. We have considered all the suggestions and have revised the manuscript accordingly. We especially appreciate the reviewer’s suggestions about the characterisations of the G. platifrons gill proliferative cell populations. In a separate research project, we have tested both cell division and cell proliferation markers on the proliferation cell populations. Though we are not able to include these results in the current manuscript, we are happy to share our preliminary results with the reviewer. Our results demonstrate the proliferative cell populations, particularly the VEPCs, are cell proliferation marker positive, and contains high amount of mitotic cells.

Author response image 3.

Finally, there is a body of literature that has examined cell proliferation and zones of proliferation in mussels (such as Piquet, B., Lallier, F.H., André, C. et al. Regionalized cell proliferation in the symbiont-bearing gill of the hydrothermal vent mussel Bathymodiolus azoricus. Symbiosis 2020) or other organisms (such as Bird, A. M., von Dassow, G., & Maslakova, S. A. How the pilidium larva grows. EvoDevo. 2014) that could be discussed.

We appreciate the comments and suggestions from the reviewer. We have considered all the suggestions and have revised the manuscript accordingly (line 226-229).

Minor comments also include:

Consider changing the orientation of diagrams in Figure 2C' in relationship to Figure 2C and 2D-K.

We appreciate the comments and suggestions from the reviewer. The Figure 2 has been reorganized.

For the diagram in Figure 3K, please clarify if the arrows drawn for the direction of inter lamina water flow is based on gene expression, SEM, or some previous study.

We are grateful for the reviewer's valuable feedback and suggestions. The arrows in the figure indicate the direction of water flow that could be affected by specific types of cilium. Our prediction is based on both gene expression and SEM results. To further clarify this point, we have revised the figure legend of Fig. 3.

Please include a label for the clusters in Figure 5E for consistency.

We have revised our Figure 5E to keep our figures consistent.

Please include a note in the Materials and Methods for Monocle analysis in Figure 6.

We conducted Monocle analyses using Monocle2 and Monocle 3 in R environment. We have revised our material and methods with further information of Figure 6.

In Supplement 2, the first column is labeled PEBC while the first row is labeled PEBZ versus all other rows and columns have corresponding names. I am guessing this is a typo and not different clusters?

We appreciate the great effort of the reviewer in reviewing our manuscript. We have corrected the typo in the revised version.

Author Response

The following is the authors’ response to the original reviews.

Reviewer #1:

1. The most important concern that I have refers to the FDTD simulations to characterize the ZMW, as shown in Appendix 2, Figure 4. So far, the explanations given in the caption of Figure 4 are confusing and misleading: the authors should provide more detailed explanations on how the simulations were performed and the actual definition of the parameters used. In particular:

a. lines 1330-1332: it is not clear to me how the fluorescence lifetime can be calculated from the detected signal S (z), and why they are horizontal, i.e., no z dependence? Which lifetimes are the authors referring to?

b. lines 1333-1335: Where do these values come from? And how do they relate to panels D & E? From what I can see in these panels the lifetimes are highly dependent on z and show the expected reduction of lifetime inside the nanostructures.

c. lines 1336-1337: Why the quantum yield of the dyes outside the ZMW differs from those reported in the literature? In particular the changes of quantum yield and lifetime for Alexa 488 are very large (also mentioned in the corresponding part of Materials & Methods but not explained in any detail).

We thank the Reviewer for his detailed questions on the FDTD simulations. We have now added the missing equation related to the computation of signal-averaged fluorescence lifetimes from the FDTD simulations. Specifically to the three points raised:

a) The fluorescence lifetime is indeed not calculated from the detected signal S(z), but from the radiative and non-radiative rates in the presence of the ZMW as given in eq. 9-10. However, we use the detected signal S(z) to compute the average fluorescence lifetime over the whole z-profile of the simulation box, which we relate to the experimentally measured fluorescence lifetimes as given in Appendix 7, Figure 1. We have now added the equation to compute the signal-weighted fluorescence lifetimes, which we denote as <𝜏>S , in eq. 13 in the methods. To clarify this point, we have added the symbol <𝜏>S to the plots in Appendix 2, Figure 4 D-E and Appendix 7, Figure 1 C-D.

b) The estimated lifetimes were obtained as the signal-weighted average over the lifetime profiles, (<𝜏>S) as given in the new eq. 13. All plotted quantities, i.e., the detection efficiency η, quantum yield ϕ, detected signal S(z), and fluorescence lifetime, are computed from the radiative and loss rates obtained from the FDTD simulation according to eqs. 8-11. To make this clearer, we have now added the new Appendix 2 – Figure 5 which shows the z-profiles of the quantities (radiative and loss rates) used to derive the experimental observables.

c) There are multiple reasons for the differences of the quantum yields of the two analytes used in this study compared to the literature values. For cyanine dyes such as Alexa647, it is well known that steric restriction (as e.g. caused by conjugation to a biomolecule) can lead to an increase of the quantum yield and fluorescence lifetime. We observe a minor increase of the fluorescence lifetime for Alexa647 from the literature value of 1.17 ns to a value of 1.37 ns when attached to Kap95, which is indicative of this effect. In the submitted manuscript, this was discussed in the methods in lines 936-938 (lines 938-945 in the revised manuscript). For the dye Alexa488, which is used to label the BSA protein, this effect is absent. Instead, we observe (as the Reviewer correctly notes) a quite drastic reduction of the fluorescence lifetime compared to the unconjugated dye from 4 ns to 2.3 ns. In cases where a single cysteine is labeled on a protein, such a drastic reduction of the quantum yield usually indicates the presence of a quenching moiety in proximity of the labeling site, such as tryptophane, which acts via the photo-induced electron transfer mechanism. Indeed, BSA contains two tryptophanes that could be responsible for the low quantum yield of the conjugated dyes. The situation is complicated by the fact that BSA contains 35 cysteines that can potentially be labeled (although 34 are involved in disulfide bridges). The labeled BSA was obtained commercially and the manufacturer lists the degree of labeling as ~6 dye molecules per protein, with a relative quantum yield of 0.2 compared to the standard fluorescein. This corresponds to an absolute quantum yield of ~0.16, which is low compared to the literature value for Alexa488 of ~0.8.

Based on the measured fluorescence lifetime, we estimate a quantum yield of 0.46, which is higher than the photometrically obtained value of 0.16 reported by the manufacturer. Fully quenched, nonfluorescent dyes will not contribute to the lifetime measurement but are detected in the photometric quantum yield estimates. The difference between the lifetime and photometric based quantum yield estimates thus suggest that part of the fluorophores are almost fully quenched. While it is unknown where the dyes are attached to the protein, the low quantum yield could be indicative of dye-dye interactions via pi-pi stacking, which can often lead to non-fluorescent dimers. This is supported by the fact that the manufacturer reports color differences between batches of labeled protein, which indicate spectral shifts of the absorption spectrum when dye-dye adducts are formed by π-π stacking. We have now added a short discussion of this effect in lines 938-941. We note that the conclusions drawn on the quenching effect of the metal nanostructure remain valid despite the drastic reduction of the quantum yield for Alexa488, which leads to a further quantum yield reduction of the partly quenched reference state.

2) A second important concern refers to Figure 3: Why is there so much variability on the burst intensities reported on panels C, D? They should correspond to single molecule translocation events and thus all having comparable intensity values. In particular, the data shown for BSA in panel D is highly puzzling, since it not only reflects a reduced number of bursts (which is the main finding) but also very low intensity values, suggesting a high degree of quenching of the fluorophore being proximal to the metal on the exit side of the pore. In fact, the count rates for BSA on the uncoated pore range form 50-100kcounts/s, while on the coated pores thy barely reach 30 kcounts/s, a clear indication of quenching. Importantly, and in direct relation to this, could the authors exclude the possibility that the low event rates measured on BSA are largely due to quenching of the dye by getting entangled in the Nsp mesh just underneath the pore but in close contact to the metal?

The Reviewer raises a valid concern, but further analysis shows that this is unproblematic. Notably, the burst intensities are in fact not reduced, in contrast to the visual impression obtained from the time traces shown in the figure. The time trace of the BSA intensity is visually dominated by high-intensity bursts which mask the low-intensity bursts in the plot. In contrast, in Figure 3 the reduced number of BSA events results in a sparser distribution of the intensity spikes, which allows low-intensity events to be seen. Different to the visual inspection, the spike-detection algorithm does not exhibit any bias in terms of the duration or the number of photons of the detected events between the different conditions for both BSA and Kap95, as shown in the new Appendix 7 – Figure 1. Using FCS analysis it can be tested whether the event duration varies between the different conditions shown in Figure 3 C-D. This did not show a significant difference in the estimated diffusion time for BSA (Appendix 7 – Figure 1 C,D). Contrary to the suggestion of the Reviewer, we also do not observe any indication of quenching by the metal between uncoated and Nsp1-coated pores for BSA. Such quenching should result in differences of the fluorescence lifetimes, which however is not evident in our experimental data (Appendix 7 – Figure 1 F).

3) Line 91: I suggest the authors remove the word "multiplexed" detection since it is misleading. Essentially the authors report on a two-color excitation/detection scheme which is far from being really multiplexing.

We have changed the word to “simultaneous” now and hope this avoids further confusion.

4) Line 121: why are the ZMW fabricated with palladium? Aluminum is the gold-standard to reduce light transmissivity. An explanation for the choice of this material would be appreciated by the community.

In a previous study (Klughammer and Dekker, Nanotechnology, 2021), we established that palladium can have distinct advantages compared to other ZMW metals such as aluminum and gold, most prominently, an increased chemical stability and reduced photoluminescence. For this study, we chose palladium over aluminum as it allowed the use of simple thiol chemistry for surface modification. In the beginning of the project, we experimented with aluminum pores as well. We consistently found that the pores got closed after measuring their ionic conductance in chlorine-containing solutions such as KCl or PBS. This problem was avoided by choosing palladium.

5) Lines 281-282: This statement is somewhat misleading, since it reads such that the molecules stay longer inside the pore. However, if I understand correctly, these results suggest that Kap95 stays closer to the metal on the exit side. This is because measurements are being performed on the exit side of the pore as the excitation field inside the pore is quite negligible.

We thank the Reviewer for this comment and have clarified the text in lines 290-292 as suggested to: “(…) this indicates that, on the exit side, Kap95 diffuses closer to the pore walls compared to BSA due to interactions with the Nsp1 mesh”

6) Lines 319-320: Although the MD simulations agree with the statement being written here, the variability could be also due to the fact that the proteins could interact in a rather heterogenous manner with the Nsp mesh on the exit side of the pore, transiently trapping molecules that then would stay longer and/or closer to the metal altering the emission rate of the fluorophores. Could the authors comment on this?

The variation mentioned in the text refers to a pore-to-pore variation and thus needs to be due to a structural difference between individual pores. This effect would also need to be stable for the full course of an experiment, typically hours. We did not find any structural changes in the fluorescence lifetimes measured on individual pores such as suggested by the Reviewer. We think that the suggested mechanism would show up as distinct clusters in Appendix 7 – Figure 1 E,F where we found no trace of such a change to happen. If we understand correctly, the Reviewer suggests a mechanism, not based on changes in the Nup layer density, that would lead to a varying amount of trapping of proteins close to the surface. Such a behavior should show up in the diffusion time of each pore ( Appendix 7 – figure 1 C,D), where we however find no trace of such an effect.

7) Lines 493-498: These claims are actually not supported by the experimental data shown in this contribution: a) No direct comparison in terms of signal-to-noise ratio between fluorescence-based and conductance-based readouts has been provided in the ms. b) I would change the word multiplexed by simultaneous since it is highly misleading. c) The results shown are performed sequentially and thus low throughput. d) Finally, the use of unlabeled components is dubious since the detection schemes relies on fluorescence and thus requiring labeling.

We thank the Reviewer for pointing this out.

a) We have now added a section in appendix 3 that discusses the signal-to-noise ratios. In brief, there are three observations that led us to conclude that ZMWs provide beneficial capabilities to resolve individual events from the background:

1. The signal-to-background ratio was determined to be 67±53 for our ZMW data of Kap95 which is an order of magnitude higher compared to the ~5.6 value for a conductance-based readout.

2. The detection efficiency for ZMWs is independent of the Kap95 occupancy within the pore. This is different from conductance based approaches that have reduced capability to resolve individual Kap95 translocations at high concentrations.

3. The fraction of detected translocations is much higher for ZMWs than for conductance-based data (where lots of translocations occur undetected) and matches closer to the theoretical predictions.

b) We have changed the wording accordingly.

c) We agree with the Reviewer that our method is still low throughput. However, the throughput is markedly increased compared to previous conductance-based nanopore measurements. This is because we can test many (here up to 8, but potentially many more) pores per chip in one experiment, whereas conductance-based readouts are limited to a single pore. We have now changed the wording to “increased throughput” in line 507 to avoid confusion.

d) We agree that only labeled components can be studied directly with our methods. However, the effect of unlabeled analytes can be assessed indirectly without any perturbation of the detection scheme due to the specificity of the fluorescent labeling. This is distinct from previous nanopore approaches using a conductance-based readout that lack specificity. In our study, we have for example used this advantage of our approach to access event rates at high concentrations (1000nM Kap95, 500nM BSA) and large pore diameters by reducing the fraction of labeled analyte in the sample. Finally, the dependence of the BSA leakage rate as a function of the concentration of Kap95 (Figure 6) relies on a specific readout of BSA events in the presence of large amounts of Kap95, which would be impossible in conductance-based experiments.

8) Line 769: specify the NA of the objective. Using a very long working distance would also affect the detection efficiency. Have the authors considered the NA of the objective on the simulations of the detection efficiency? This information should be included and it is important as the authors are detecting single molecule events.

We used an NA of 1.1 for the simulation of the Gaussian excitation field in the FDTD simulations, corresponding to the NA of the objective lens used in the experiments and as specified in the methods. The Reviewer is correct that the NA also affects the absolute detection efficiency of the fluorescence signal due to the finite opening angle of the collection cone of ~56˚. In our evaluation of the simulations, we have neglected this effect for simplicity, because the finite collection efficiency of the objective lens represents only an additional constant factor that does not depend on the parameters of the simulated system, such as the pore diameter. Instead, we focused solely the effect of the ZMW and defined the detection efficiency purely based on the fraction of the signal that is emitted towards the detection side and can potentially be detected in the experiment, which also provides the benefit that the discussed numbers are independent of the experimental setup used.

To clarify this, we have now made this clearer in the method text on lines 917-920.

9) Line 831: I guess that 1160ps is a mistake, right?

This is not a mistake. We performed a tail fit of the fluorescence decay curves, meaning that the initial rise of the decay was excluded from the fit. The initial part of the fluorescence decay is dominated by the instrument response function (IRF) of the system, with an approximate width of ~500 ps. To minimize the influence of the IRF on the tail fit, we excluded the first ~1 ns of the fluorescence decay.

10) Lines 913-917: Why are the quantum yield of Alexa 488 and lifetime so much reduced as compared to the published values in literature?

See answer to point 1. We have added a short discussion at lines 938-941 where we speculate that the reduced quantum yield is most likely caused by dye-dye interactions due to the high degree of labeling of ~6 dyes per protein.

11) Lines 1503-1509: The predicted lifetimes with the Nsp-1 coating have not been shown in Appendix 2 - Figure 4. How have they been estimated?

We have not performed predictions of fluorescence lifetimes in the presence of an Nsp1 coating. Predictions of the fluorescence lifetime in the absence of the Nsp1 coating were obtained by assuming a uniform occupancy of the molecules over the simulation box. A prediction of the fluorescence lifetimes in the presence of the Nsp1 coating would require a precise knowledge of the spatial distribution of analytes, which depends, among other factors, on the extension of the Nsp1 brushes and the interaction strengths with the FG repeats. While simulations provide some insights on this, we consider a quantitative comparison of predicted and measured fluorescence lifetimes in the presence of the Nsp1 coating beyond the scope of the present study.

12) Lines 1534-1539: I disagree with this comment, since the measurements reported here have been performed outside the nano-holes, and thus the argument of Kap95 translocating along the edges of the pore and being responsible for the reduced lifetime does not make sense to me.

In accordance with our answer to point 5 above, we have now changed the interpretation to the proximity of Kap95 to the metal surface on the exit side, rather than speculating on the path that the protein takes through the pore (lines 1662-1664), as follows:

“This indicates that, in the presence of Nsp1, Kap95 molecules diffuse closer to or spend more time in proximity of the metal nanoaperture on the exit side.”

Reviewer #2:

(Numbers indicate the line number.)

48: should cite more recent work: Timney et al. 2016 Popken et al 2015

59: should cite Zilman et al 2007, Zilman et al 2010

62: should cite Zilman et al 2010

We thank the Reviewer for the suggestions and have added them to the manuscript now.

65: one should be careful in making statements that the "slow" phase is immobile, as it likely rapidly exchanging NTRs with the "fast" phase.

We have removed this description and replaced it by “This 'slow phase' exhibits a reduced mobility due to the high affinity of NTRs to the FG-Nup mesh.” to avoid misunderstanding.

67: Schleicher 2014 does not provide evidence of dedicated channels

We agree with the Reviewer and therefore moved the reference to an earlier position in the sentence.

74-75: must cite work by Lusk & Lin et al on origami nanochannels

We thank the Reviewer for this suggestion. We have now added a reference to the nanotraps of Shen et al. 2021, JACS, in line 75. In addition, we now also refer to Shen et al. 2023, NSMB, in the discussion where viral transport is discussed.

77: Probably Jovanovic- Talisman (2009)?

We thank the Reviewer for pointing out this typo.

93; should cite Auger&Montel et al, PRL 2014

We thank the Reviewer for pointing out this reference. To give proper credit to previous ZMW, we have now incorporated a sentence in lines 100-102 citing this reference.

111-112: there appears to be some internal inconsistency between this interpretation and the BSA transport mostly taking place through the "central hole" (as seems to be implied by Equation (3). Probably it should be specified explicitly that the "central hole" in large channels is a "void".

We thank the Reviewer for this suggestion and have added a clarifying sentence.

115-177: This competition was studied in Jovanovic-Talisman 2009 and theoretically analysed in Zilman et al Plos Comp Biol 2010. The differences in the results and the interpretation should be discussed.

We agree, therefore it is discussed in the discussion section (around line 594) and now added the reference to Zilman et al.

Figure 2 Caption: "A constant flow..." - is it clear that is flow does not generate hydrodynamic flow through the pore?

The Reviewer raises an important point. Indeed, the pressure difference over the membrane generates a hydrodynamic flow through the pore that leads to a reduction of the event rate compared to when no pressure is applied. However, as all experiments were performed under identical pressures, one can expect a proportional reduction of the absolute event rates due to the hydrodynamic flow against the concentration gradient. In other words, this will not affect the conclusions drawn on the selectivity, as it is defined as a ratio of event rates.

We have now added additional data on the influence of the hydrodynamic flow on the translocation rate in Appendix 3 – Figure 2, where we have measured the signal of free fluorophores at high concentration on the exit side of the pore as a function of the applied pressure. The data show a linear dependence of the signal reduction on the applied pressure. At the pressure values used for the experiments of 50 mbar, we see a ~5% reduction compared to the absence of pressure, implying that the reported absolute event rates are underestimated only by ~5%. Additionally we have added such data for Kap95 translocations that shows a similar effect (however less consistent). Measuring the event rate at zero flow is difficult, since this leads to an accumulation of fluorophores on the detection side.

Figure 3: it would help to add how long is each translocation, and what is the lower detection limit. A short explanation of why the method detects actual translocations would be good

With our method, unfortunately, we can not assess the duration of a translocation event since we only see the particle as it exists the pore. Instead, the measured event duration is determined by the time it takes for the particle to diffuse out of the laser focus. This is confirmed by FCS analysis of translocation events that show the same order of magnitude of diffusion times as for free diffusion (Appendix 7 – Figure 1 C,D) in contrast to a massively reduced diffusion time within a nanopore. In Figure 2D we show the detection efficiency at different locations around the ZMW as obtained from FDTD simulations and discuss the light blocking. This clearly shows that the big majority of the fluorescence signal comes from the laser illuminated side and therefore only particles that translocated through the ZMW are detected as presented between lines 170-190. In Yang et al. 2023, bioRxiv (https://doi.org/10.1101/2023.06.26.546504) a more detailed discussion about the optical properties of Pd nanopores is given.

This point also explains why we see actual translocations: since the light is blocked by the ZMW, fluorophores can only be detected after they have translocated. On parts of the membrane without pores and upstream the amount of spikes found in a timetrace was found to be negligibly small. Additionally, if a significant part of the signal would be contributed by leaking fluorescence from the dark top side, there should no difference in BSA event rate found between small open and Nsp1 pores which we did not observe.

With respect to the lower detection limit for events: In the burst search algorithm we require a false positive level rate of lower than 1 event in 100. Additionally, as described in Klughammer and Dekker, Nanotechnology (2021), we apply an empirical filtering to remove low signal to noise ratio events that contain less than 5 detected photons per event or a too low event rate. From the event detection algorithm there is no lower limit set on the duration of an event. Such a limit is then set by the instrument and the maximum frequency it which it can detect photons. This time is below 1μs. Practically we don’t find events shorter than 10μs as can be seen in the distribution of events where also the detection limits can be estimated (Appendix 7 – figure 1 A and B.)

Equation (1): this is true only for passive diffusion without interactions (see eg Hoogenboom et al Physics Reports 2021 for review). Using it for pores with interactions would predict, for instance, that the inhibition of the BSA translocation comes from the decrease in D which is not correct.

We agree with the Reviewer that this equation would not reproduce the measured data in a numerically correct way. We included it to justify why we subsequently fit a quadratic function to the data. As we write in line 260 we only used the quadratic equation “as a guide to the eye and for numerical comparison” and specifically don’t claim that this fully describes the translocation process. In this quadratic function, we introduced a scaling factor α that can be fitted to the data and thus incorporates deviations from the model. In appendix 5 we added a more elaborate way to fit the data including a confinement-based reduction of the diffusion coefficient (although not incorporating interactions). Given the variations of the measured translocation rates, the data is equally well described by both the simple and the more complex model function.

Equation (1): This is not entirely exact, because the concentration at the entrance to the pore is lower than the bulk concentration, which might introduce corrections

We agree with the Reviewer and have added that the concentration difference Δc is measured at the pore entrance and exit, and this may be lower than the bulk concentration. As described in our reaction to the Reviewer’s previous comment, equation (1) only serves as a justification to use the quadratic dependence and any deviations in Δc are absorbed into the prefactor α in equation (2).

Equation (3): I don't understand how this is consistent with the further discussion of BSA translocation. Clearly BSA can translocate through the pore even if the crossection is covered by the FG nups (through the "voids" presumably?).

The Reviewer raises an important point here. Equation 3 can only be used for a pore radius r > rprot + b. b was determined to be 11.5 nm and rprot is 3.4 nm for BSA, thus it needs to be that r > 15 nm. We would like to stress, however, that b does not directly give a height of a rigid Nsp1 ring but is related to the configuration of the Nsp1 inside the pore. Equation (3) (and equation (2)) were chosen because even these simple equations could fit the experimentally measured translocation rates well, and not because they would accurately model the setup in the pore. As we found from the simulations, the BSA translocations at low pore diameters presumably happen through transient openings of the mesh. The dynamics leading to the stochastic opening of voids on average leads to the observed translocation rate.

296-297: is it also consistent with the simulations?

We compare the experimentally and simulated b values in lines 387-388 and obtained b=9.9 ± 0.1 nm from the simulations (as obtained from fitting the translocation rates and not from measuring the extension of the Nsp1 molecules) and 11.5 ± 0.4 nm from the experiments – which we find in good agreement.

331: has it been established that the FG nups equilibrate on the microsecond scale?

As an example, we have analyzed the simulation trajectory of the most dense nanopore (diameter = 40 nm, grafting = 1/200 nm2). In Author response image 1 we show for each of the Nsp1-proteins how the radius of gyration (Rg) changes in time over the full trajectory (2 μs + 5 μs). As expected, the Rg values reached the average equilibrium values very well within 2 μs simulation time, showing that the FG-Nups indeed equilibrate on the (sub)microsecond scale.

Author response image 1.

334-347: the details of the method should be explained explicitly in the supplementary (how exactly voids distributions are estimated and the PMF are calculated etc)

The void analysis was performed with the software obtained from the paper of Winogradoff et al. In our Methods we provide an overview of how this software calculates the void probability maps and how these are converted into PMFs. For a more detailed description of how exactly the analysis algorithm is implemented in the software, we refer the reader to the original work. The analysis codes with the input files that were used in this manuscript have been made public ( https://doi.org/10.4121/22059227.v1 ) along with the manuscript.

Equation (4) is only an approximation (which works fine for high barriers but not the low ones). Please provide citations/derivation.

To our knowledge, the Arrhenius relation is a valid approximation for our nanopore simulations. We are unaware of the fact that it should not work for low barriers and cannot find mention of this in the literature. It would be helpful if the Reviewer can point us to relevant literature.

Figure 4: how was transport rate for Kaps calculated?

As mentioned in lines 388-391, we assumed that the Kap95 translocation rate through Nsp1-coated pores is equal to that for open pores, as we did not observe any significant hindrance of Kap95 translocation by the Nsp1 mesh in the experiment (Figure 4 A,C).

378: It's a bit strange to present the selectivity ratio as prediction of the model when only BSA translocation rate was simulated (indirectly).

We agree with the Reviewer that ideally we should also simulate the Kap95 translocation rate to obtain an accurate selectivity measure of the simulated nanopores. However, as the experiments showed very similar Kap95 translocation rates for open pores and Nsp1-coated pores, we believe it is reasonable to take the Kap95 rates for open and Nsp1-pores to be equal.

Figure 5C and lines 397: I am a bit confused how is this consistent with Figure 4D?

Figure 5C and figure 4D both display the same experimental data, where 4D only focuses on a low diameter regime. In relation to line 397 (now 407), the Nsp1 mesh within the 60-nm pore dynamically switches between closed configurations and configurations with an open channel. When taking the temporal average of these configurations, we find that the translocation rate is higher than for a closed pore but lower than for a fully open pore. The stochastic opening and closing of the Nup mesh results in the continuous increase of the translocation rates with increasing diameter, which is in contrast to a step-wise increase that would be expected from an instantaneous collapse of the Nsp1 mesh at a certain pore diameter.

428-439: Please discuss the differences from Jovanovic-Talisman 2009.

How our results for a Kap95 induced change of the BSA translocation rate are related to previous literature is discussed extensively in the lines 598-620.

440: How many Kaps are in the pore at different concentrations?

This is a very interesting question that we were, unfortunately, not able to answer within the scope of this project. With our fluorescent based methods we could not determine this number because the excitation light does not reach well into the nanopore.

In our previous work on Nsp1-coated SiN nanopores using conductance measurements, we quantified the drop in conductance at increasing concentrations of Kap95 (Fragasso et al., 2023, NanoResearch, http://dx.doi.org/10.1007/s12274-022-4647-1). From this, we estimated that on average ~20 Kap95 molecules are present in a pore with a diameter of 55 nm at a bulk concentration of 2 µM. In these experiments, however, the height of the pore was only ~20 nm, which is much lower compared to 100 nm long channel used here, and the grafting density of 1 per 21 nm2 was high compared to the grafting density here of 1 per 300 nm2. Assuming that the Kap95 occupancy scales linearly with the number of binding sites (FG repeats) in the vicinity of the pore, and hence the amount of Nsp1 molecules bound to the pore, we would expect approximately ~7 Kap95 molecules in a pore of similar diameter under saturating (> 1 µM) concentrations.

On the other hand, the simulations showed that the density of Nsp1 within the pore is equal to the density within the 20-nm thick SiN pores (line 380). For the longer channel and lower grafting density used here, Nsp1 was also more constrained to the pore compared to thinner pores used in previous studies (Fragasso et al., 2023, NanoResearch), where the grafted protein spilled out from the nanopores. Thus assuming that the Kap95 occupancy depends on the protein density in the pore volume rather than the total protein amount grafted to the pore walls, we would estimate a number of 100 Kap95 molecules per pore.

These varying numbers already show that we cannot accurately provide an estimate of the Kap95 occupancy within the pore from our data due to limitations of the ZMW approach.

445: how is this related to the BSA translocation increase?

For the calculation of the selectivity ratio, we assumed the normalized Kap95 translocation rate to be independent of the Kap95 concentration. Hence, the observed trends of the selectivity ratios at different concentrations of Kap95, as shown in Figure 6 D, are solely due to a change in the BSA translocation rate at different concentrations of Kap95, as given in Figure 6 B,C.

462-481: it's a bit confusing how this interfaces with the "void" analysis ( see my previous comments)

We agree that the phenomenological descriptions in terms of transient openings (small, dynamic voids) that for larger pores become a constantly opened channel (a single large, static void) might cause some confusion to the reader. In the last part of the results, we aimed to relate the loss of the BSA rate to a change of the Nsp1 mesh. We acknowledge that the model of a rim of Nsp1 and an open center described in Figure 5F is highly simplifying . We now explain this in the revised paper at lines 483-486 by referring to an effective layer thickness which holds true under the simplifying assumption of a central transport channel.

Figure 6D: I think the illustration of the effect of kaps on the brush is somewhat misleading: at low pore diameters, it is possible that the opposite happens: the kaps concentrate the polymers towards the center of the pore. It should be also made clear that there are no kaps in simulations (if I understand correctly?)

Indeed, at small pore diameters we think it would be possible to observe what the Reviewer describes. The illustration should only indicate what we think is happening for large pore diameters where we observed the opening of a central channel. To avoid confusion, we now shifted the sketches to panel G where the effective layer thickness is discussed.

Indeed, as stated in lines 331-340 no Kap95 or BSA molecules were present in the simulations. We have now clarified this point in lines 872-876.

518: Please provide more explanation on the role of hydrodynamics pressure.

We have now performed additional experiments and quantified the effect of the pressure to be a ~5% reduction of the event rates, as described in the answer to a previous question above.  

Reviewer #3 (Recommendations For The Authors):

No experiments have been performed with the Ran-Mix regeneration system. It would be beneficial to add Ran-Mix to the trans compartment and see how this would affect Kap95 translocation events frequency and passive cargo diffusion. As the authors note in their outlook, this setup offers an advantage in using Ran-Mix and thus could also be considered here or in a future follow-up study.

We thank the Reviewer for this suggestion. We think, however, that it is beyond the scope of this paper and an interesting subject for a follow-up study.

Author Response

The following is the authors’ response to the original reviews.

eLife assessment

This study and associated data is compelling, novel, important, and well-carried out. The study demonstrates a novel finding that different chemotherapeutic agents can induce nucleolar stress, which manifests with varying cellular and molecular characteristics. The study also proposes a mechanism for how a novel type of nucleolar stress driven by CDK inhibitors may be regulated. The study sheds light on the importance of nucleolar stress in defining the on-target and offtarget effects of chemotherapy in normal and cancer cells.

We are thankful to the reviewers and the editor for their feedback and thorough assessment of our work. Our responses to the comments and suggestions are below.

Reviewer #1 (Public Review):

The study titled "Distinct states of nucleolar stress induced by anti-cancer drugs" by Potapova and colleagues demonstrates that different chemotherapeutic agents can induce nucleolar stress, which manifests with varying cellular and molecular characteristics. The study also proposes a mechanism for how a novel type of nucleolar stress driven by CDK inhibitors may be regulated. As a reviewer, I appreciate the unbiased screening approach and I am enthusiastic about the novel insights into cell biology and the implications for cancer research and treatment. The study has several significant strengths: i) it highlights the understudied role of nucleolar stress in the on- and off-target effects of chemotherapy; ii) it defines novel molecular and cellular characteristics of the different types of nucleolar stress phenotypes; iii) it proposes novel modes of action for well-known drugs. However, there are several important points that should be addressed:

• The rationale behind choosing RPE cells for the screen is unclear. It might be more informative to use cancer cells to study the effects of chemotherapeutic agents. Alternatively, were RPE cells selected to evaluate the side effects of these agents on normal cells? Clarifying these points in the introduction and discussion would guide the reader.

RPE1, a non-cancer-derived cell line, was chosen for this study to evaluate the effects of anticancer drugs on normal nucleolar function, with the underlying premise that nucleolar stress in normal cells can contribute to non-specific toxicity. This clarification is added to the introduction. Another factor that played in selecting a normal cell line for the drug screen and subsequent experiments was the spectrum of known and unknown genetic and metabolic alterations present in various cancer cell lines. These variables are often unique to a particular cancer cell line and may or may not impact nucleolar proteome and function. Therefore, the nucleolar stress response can be influenced by the spectrum of alterations inherent to each cancer. Our primary focus was to determine the impact of these drugs under normal conditions.

That said, the selected hits of main drug classes were validated in a panel of cell lines that included two other hTERT lines (BJ5TA and CHON-002) and two cancer lines (DLD1 and HCT116). In cancer cells starting nucleolar normality scores were lower than in hTERT cells, suggesting that genetic and metabolic changes in these cells may indeed affect nucleolar morphology. Nonetheless, all drugs from a panel of selected hits from different target classes validated in both cancer cell lines (Fig. 2F).

• Figure 2F indicates that DLD1 and HCT116 cells are less sensitive to nucleolar changes induced by several inhibitors, including CDK inhibitors. It would be crucial to correlate these differences with cell viability. Are these differences due to cell-type sensitivity or variations in intracellular drug levels? Assessing cell viability and intracellular drug concentration for the same drugs and cells would provide valuable insights.

One of the reasons for the reduced magnitude of the effects of selected drugs in DLD1 and HCT116 cells is their lower baseline normality scores compared to hTERT cells (now shown in Sup. Fig. 1B-C). Other potential factors include proteomic and metabolic shifts and alterations in signaling pathways that control ribosome production. The less-likely possibility of variations in intracellular drug levels cannot be excluded, but measuring this for every compound in every cell line was not feasible in this study. These limitations are now noted in the results section.

Regarding the point about viability - our initial screen output, in addition to normality scores, included cell count (cumulative count of cells in all imaged fields), which serves as a proxy for viability. By this measure, all hit compounds in our screen were cytostatic or cytotoxic in RPE1 cells (Fig. 2C). The impact of these drugs on the viability of cancer cells that can have various degrees of addiction to ribosome biogenesis merits a separate study of a large cancer cell line panel.

• Have the authors interpreted nucleolar stress as the primary cause of cell death induced by these drugs? When cells treated with CDK inhibitors exhibit the dissociated nucleoli phenotype, is this effect reversible? Is this phenotype indicative of cell death commitment? Conducting a washout experiment to measure the recovery of nucleolar function and cell viability would address these questions.

Whether nucleolar toxicity is the primary cause of cytotoxicity for a given chemotherapy drug is an incisive and thought-provoking question. Our screen did not discern whether the cytotoxic effects of our hits were due to inhibition of their intended targets, their impact on the nucleolus, or a combined effect. This point is now mentioned in the results section. Regarding the reversibility of the nucleolar disassembly phenotype seen in CDK inhibitors –in the case of flavopiridol, which is a reversible CDK inhibitor, we demonstrated that nucleoli re-assembled within 4-6 hours after the drug was washed out. An example of this is shown in Sup. Figure 3 and in Video 5. For these experiments, cells were pretreated with the drug for 5 hours, not long enough to cause cell death.

• The correlation between the loss of Treacle phosphorylation and nucleolar stress upon CDK inhibition is intriguing. However, it remains unclear how these two events are related. Would Treacle knockdown yield the same nucleolar phenotype as CDK inhibition? Moreover, would point mutations that abolish Treacle phosphorylation prevent its interaction with Pol-I? Experiments addressing these questions would enhance our understanding of the correlation/causation between Treacle phosphorylation and the effects of CDK inhibition on nucleolar stress.

We agree that the Treacle finding is interesting and warrants further investigation. In our attempts to knock down Treacle with siRNA, its protein levels were reduced by no more than 50%, which was not sufficient to cause a strong nucleolar stress response. Therefore, these data were not incorporated into the manuscript. However, in our view, Treacle is unlikely to be the only nucleolar CDK substrate whose dephosphorylation is causing the “bare scaffold” phenotype caused by the transcriptional CDK inhibitors. Our phospho-proteomics studies identified multiple nucleolar CDK substrates with established roles in the formation of the nucleolus. For instance, the granular component protein Ki-67 was also dephosphorylated on multiple sites and dispersed throughout the nucleus (shown in Sup. Fig 4). Given that CDKs typically phosphorylate many substrates that can have multiple phosphorylation sites, identifying a sole protein or phosphorylation site responsible for nucleolar disassembly may be an unattainable target.

Overall, this study is significant and novel as it sheds light on the importance of nucleolar stress in defining the on-target and off-target effects of chemotherapy in normal and cancer cells.

Thank you, we appreciate the positive and constructive assessment of our study.

Reviewer #2 (Public Review):

This is an interesting study with high-quality imaging and quantitative data. The authors devise a robust quantitative parameter that is easily applicable to any experimental system. The drug screen data can potentially be helpful to the wider community studying nucleolar architecture and the effects of chemotherapy drugs. Additionally, the authors find Treacle phosphorylation as a potential link between CDK9 inhibition, rDNA transcription, and nucleolar stress. Therefore I think this would be of broad interest to researchers studying transcription, CDKs, nucleolus, and chemotherapy drug mechanisms. However, the study has several weaknesses in its current form as outlined below.

1) Overall the study seems to suffer from a lack of focus. At first, it feels like a descriptive study aimed at characterizing the effect of chemotherapy drugs on the nucleolar state. But then the authors dive into the mechanism of CDK inhibition and then suddenly switch to studying biophysical properties of nucleolus using NPM1. Figure 6 does not enhance the story in any way; on the contrary, the findings from Fig. 6 are inconclusive and therefore could lead to some confusion.

This study was specifically designed to examine a broad range of chemotherapy drugs. The newly created nucleolar normality score enabled us to measure nucleolar stress precisely and in high throughput. Our primary objective was to find drugs that disrupt the normal nucleolar morphology and then study in-depth the most interesting and novel hits. We have made revisions to emphasize that these are the primary focal points of the manuscript.

As context, we were motivated to explore the biophysical properties of the nucleolus because they are thought to underlie its formation and function, which also suggested a potential predictive value for modeling nucleolar responses to drug treatments. For this, we edited the RPE1 cell line by endogenously tagging NPM1, a granular component protein that behaves in line with the phase-separation paradigm in vitro and when over-expressed. We fully expected to confirm that its behavior in vivo would be consistent with LLPS, but instead found that even in an untreated scenario, the dynamics of endogenous NPM1 could not be fully explained by the phase separation theory (Fig. 6 A-C). Our message is that accurately predicting drug responses using the nucleolar normality score as a readout, based on our current understanding of the biophysical forces governing nucleolar assembly, is unworkable. For instance, normality scores decrease and NPM1 dynamics increase radically when CDKs are inhibited, without changes in NPM1 concentration or concentrations of other protein components (Fig.6 E-H). These observations are important because they highlight our gaps in understanding the relative contribution of phase separation versus active assembly in nucleolar formation. We believe that these observations are worth sharing with the scientific community.

2) The justification for pursuing CDK inhibitors is not clear. Some of the top hits in the screen were mTOR, PI3K, HSP90, Topoisomerases, but the authors fail to properly justify why they chose CDKi over other inhibitors.

We decided to focus on CDK inhibitors for several reasons. First, their effects were completely new and unexpected, suggesting the existence of an unknown mechanism regulating nucleolar structure and function. In addition, CDK inhibitors caused a very strong and distinct nucleolar stress phenotype with the lowest normality scores that merited its own term, the “bare scaffold” phenotype. One more reason for pursuing CDK-inhibiting drugs was their high rate of failure in clinics because of the intense and hard-to-explain toxicity. We suspect that this toxicity may be due at least in part to their profound effect on nucleolar organization and ribosome production throughout the body. We stated this rationale more explicitly in the manuscript.

3) In addition to poor justification, it seems like a very superficial attempt at deciphering the mechanism of CDK9imediated nucleolar stress. I think the most interesting part of the study is the link between CDK9, Pol I transcription, and nucleolar stress. But the data presented is not entirely convincing. There are several important controls missing as detailed below.

We agree with the reviewer that follow-up studies of CDK9, Pol I, and nucleolar stress connection are important long-term goals. However, the primary objective of this study was to ascertain the scope of anticancer agents that can cause nucleolar stress and the establishment of nucleolar stress categories. This is an important advance and could serve as the foundation for a standalone in-depth study or multiple studies. We have included the complete screen, proteomics, and phospho-proteomics results (Sup. Tables 1, 2, and 3), which will enable other investigators to mine the screen information based on their specific interests. Furthermore, we have made multiple text revisions to clarify rationale and interpretation, and incorporated additional data that strengthen the manuscript.

4) The authors did not test if inhibition of CDK7 and/or CDK12 also induces nucleolar stress. CDK7 and CDK12 are also major kinases of RNAPII CTD, just like CDK9. Importantly, there are well-established inhibitors against both these kinases. It is not clear from the text whether these inhibitors were included in the screen library.

Our anticancer compound library contained CDK7 inhibitor THZ1⦁2HCL, and it was a hit at both 1 and 10 uM concentrations (Sup. Table 1). However, its nucleolar stress phenotype was morphologically distinct from CDK9 inhibitors, resembling the stress caps phenotype instead of the bare scaffold phenotype. We did not pursue CDK7 because of its two hard-to-separate functions: in addition to its role as an RNAPII CTD kinase, it also acts as a CDK-activating kinase (CAK) by promoting the associations of multiple CDKs with their cyclin partners. This dual role of CDK7 makes the interpretation of THZ1-induced nucleolar stress phenotype difficult because it could be attributed to either or both of these functions. Moreover, it was reported to cause DNA damage, which may explain why it causes stress caps. An image depicting nucleolar stress phenotype caused by THZ1⦁2HCL is provided in Author response image 1.

Author response image 1.

Control and THZ1 - treated RPE1 cells, images from screen plates.

We are not aware of specific inhibitors of CDK12, as they also reportedly inhibit CDK13. None of the CDK12/CDK13 inhibitors were present in our library, therefore we can neither confirm nor exclude the possible involvement of these kinases in regulating nucleolar structure. Many other existing CDK inhibitors were absent from our library. Our work highlights the importance of assessing their potential to induce nucleolar stress and offers an approach for this assessment.

5) In Figure 4E, the authors show that Pol I is reduced in nucleolus/on rDNA. The authors should include an orthogonal method like chromatin fractionation and/or ChIP

We acknowledge the reviewer’s request for additional validation of reduced occupancy of rDNA by Pol I.<br /> Nucleolar chromatin fractionation in cells treated with CDK inhibitors is unlikely to work due to nearly complete nucleolar disassembly. Chromatin immunoprecipitation would require finding and validating a suitable ChIP-grade antibody. Moreover, the evaluation of repetitive regions by ChIP is non-trivial and error-prone. To help address this request and further confirm the POLR1A immunofluorescence results in 4E, we included additional immunofluorescence data obtained with a different POLR1A antibody (Sup. Fig. 3D), and the results were similar.

6) In Fig. 5D, in vitro kinase lacks important controls. The authors should include S to A mutants of Treacle S1299A/S1301A to demonstrate that CDK9 phosphorylates these two residues specifically.

7) To support their model, the authors should test if overexpression of Treacle mutants S1299A/S1301A can partially phenocopy the nucleolar stress seen upon CDK9 inhibition. This would considerably strengthen the author's claim that reduced Treacle phosphorylation leads to Pol I disassociation from rDNA and consequently leads to nucleolar stress.

8) Additionally, it would be interesting if S1299D/S1301D mutants could partially rescue CDK9 inhibition.

Points (6-8):

We reiterate that transcriptional CDKs target multiple nucleolar proteins, and the observed phenotype might be due to the combined effects of de-phosphorylation of multiple substrates. We concur that deconstructing the role of Treacle phosphorylation sites is very interesting and warrants further in-depth studies. The phospho-proteomics enrichment method, while an effective first-pass strategy, might not capture 100% of the phosphorylated sites. Treacle is a phospho-protein with an abundance of serine and threonine residues. It could potentially have been selectively dephosphorylated on more sites than were detected by this method. Therefore, the suggested mutations may not be the exclusive contributors responsible for the functional phenotype. Additionally, overexpressing Treacle impairs the viability of RPE1 cells, complicating the interpretation of experiments involving overexpression of both wild-type and mutant proteins. A conceivable strategy would involve generating phosphomimetic and non-phosphorylatable mutants by gene editing, studying their interactions by biochemical approaches, and determining their impact on nucleolar function, but this may take years of additional work. We hope that our work will inspire further studies that explore Treacle phosphorylation and other functions of transcriptional CDKs in nucleolar formation.

Thank you for the thoughtful review and suggestions.

Reviewer #2 (Recommendations For The Authors):

1) The manuscript could be re-organized to focus on 'CDK9-Treacle-Pol I-nucleolar stress' as the central part of the story.

While we acknowledge this suggestion, it's important to emphasize that the primary focus of this manuscript is on the identification of anticancer drugs that induce nucleolar stress and the establishment of nucleolar stress categories.

2) Include a "no ATP" control in the in vitro kinase assay and indicate molecular sizes.

We provided an additional kinase assay (Sup. Fig. 4B) that includes no ATP control lanes and a fragment of a Coomassie blue stained gel showing molecular weight markers. No ATP control assays (lanes 4 and 5) were blank as expected. Molecular weight markers were added to all other kinase assays based on the known sizes of isolated Pol II holoenzyme subunits Rbp1 (191 kDa) and Rbp2 (138 kDa).

3) For in vitro phosphorylation, please provide an explanation for using CDK9/cyclin K instead of Cyclin T1 which is the predominant cyclin for CDK9

Recombinant CDK9/cyclin K complex was used for in vitro kinase assays for a technical reason: CDK9/cyclin T obtained from the same vendor appeared to be low quality, as it showed only minimal activity toward our positive control, the isolated Pol II complex. The kinase assays using recombinant CDK9/cyclin T in parallel with CDK9/cyclin K are now presented it Sup. Fig. 4B. The first two assays in this experiment contained Pol II as a substrate, and it is evident that Pol II was phosphorylated much stronger by CDK9/cyclin K than CDK9/cyclin T (comparing lane 1 vs lane 2). Therefore, the lack of detectable Treacle phosphorylation by CDK9/Cyclin T (lane 7), in contrast to strong phosphorylation by CDK9/cyclin K (lane 6), was likely attributable to poor reagent quality rather than physiological differences. We can conclude that CDK9/cyclin K reliably phosphorylates Treacle in vitro, but CDK9/cyclin T kinase assays were inconclusive.

Author response:

The following is the authors’ response to the original reviews.

eLife assessment

This important study presents a novel pipeline for the large-scale genomic prediction of members of the non-ribosomal peptide group of pyoverdines based on a dataset from nearly 2000 Pseudomonas genomes. The advance presented in this study is largely based on solid evidence, although some main claims are only incompletely supported. This study on bacterial siderophores has broad theoretical and practical implications beyond a singular subfield.

Thank you for the supportive and encouraging words. We appreciate the editor’s and reviewers’ careful and professional assessment of this manuscript. The reviewers’ scrutiny has helped us to improve the presentation and discussion of our work. We have now carefully revised the manuscript following their instructive suggestions and comments. Please find below our detailed responses (marked in blue) to each of the comments.

Public Reviews:

Reviewer #1 (Public Review):

The manuscript introduces a bioinformatic pipeline designed to enhance the structure prediction of pyoverdines, revealing an extensive and previously overlooked diversity in siderophores and receptors. Utilizing a combination of feature sequence and phylogenetic approaches, the method aims to address the challenging task of predicting structures based on dispersed gene clusters, particularly relevant for pyoverdines.

Predicting structures based on gene clusters is still challenging, especially pyoverdines as the gene clusters are often spread to different locations in the genome. An improved method would indeed be highly useful, and the diversity of pyoverdine gene clusters and receptors identified is impressive.

However, so far the method basically aligns the structural genes and domains involved in pyoverdine biosynthesis and then predicts A domain specificity to predict the encoded compounds. Both methods are not particularly new as they are included in other tools such as PRISM (10.1093/nar/gkx320) or Sandpuma (https://doi.org/10.1093/bioinformatics/btx400) among others. The study claims superiority in A domain prediction compared to existing tools, yet the support is currently limited, relying on a comparison solely with AntiSMASH. A more extensive and systematic comparison with other tools is needed.  

Thanks for pointing this out. In the revised manuscript, we have included a comprehensive comparative analysis, in which we compared our pipeline to six different commonly used methods, including NP.searcher, PRISM4, AdenPredictor, SeMPI2, SANDPUMA, antiSMASH5 (see Supplementary_table 6 for details, and lines 281-286). These approaches either consist of a single specific algorithm or integrate several methods. Our approach performs best (see table below), demonstrating a clear improvement over previous tool. The improvements are due to several methodological differences inherent to our approach. Additionally, while exploring existing prediction tools, we found that some had not been maintained for years. For instance, we were unable to access NRPSsp (www.nrpssp.com) and NRPSpredictor2 (http://nrps.informatik.uni-tuebingen.de/). Below, we briefly explain these differences, particularly in relation to PRISM and SANDPUMA, as highlighted by the reviewer. 

Author response table 1.

PRISM annotates biosynthetic gene clusters (BGC) and reconstructs the linear structures of NRPS synthetases, with this function depending on proper annotations of open reading frames. This pipeline can have difficulties in assembling the linear structure into a final product. In our approach, we found that the annotations of NRPS gene are frequently truncated because of sequencing errors and annotation issues. Our method fixes this problem through rescanning all possible reading frames of the BGC to rebuild complete pyoverdine synthetase genes. 

Sandpum and our approach are based on similar ideas (using the prediCAT algorithm) to predict A domain substrates, namely by using the closest reference A domain annotated. However, our method uses a self-adaptive feature extraction step to reduce the co-founding influence of phylogeny. This small adjustment significantly improves the performance of our approach and even works well for small training sets (101 experimentally validated A domains with our approach as opposed to 494 A domains used by Sandpuma from MIBiG).

Additionally, in contradiction to the authors' claims, the method's applicability seems constrained to well-known and widely distributed gene clusters. The absence of predictions for new amino acids raises concerns about its generalizability to NRPS beyond the studied cases.

We thank the reviewers for this comment. We acknowledge that our method cannot directly predict new amino acids. Nevertheless, for several reasons we believe that our approach is not constrained and can be widely applied in the future.

First, our method can identify A domains that select new unknown amino acid substrates. In fact, three of the four unresolved cases in our experimental verification analysis (Fig. 3d) represent new amino acids. Obviously, experimental verification is required to characterize the unknown substrate. Once verified, the new A domains and their substrates can expand the reference dataset, allowing targeted improvement of our phylogeny-focused prediction technique. We now discuss this aspect in lines 634-645.

Second, despite that the overall substrate diversity in NRPS is high across the microbial kingdom, our analysis suggests that the number of amino acids used for a specific group of secondary metabolites quickly reaches a saturation point. The discovery rate of new amino acids was 1.7% for our experimental Pseudomonas data set (Fig. 3d). The discovery rate of new amino acids was even 0.0 % for the Burkholderiales data set. This suggests that as the database expands, the discovery rate of novel amino acid substrates is expected to drop rapidly.

Third, we acknowledge that the inability to predict the substrates of unknown domains is a common limitation among all knowledge-guided learning algorithms, including ours. However, we have made significant improvements in prediction accuracy. As the database grows, we expect the rate of unknown substrates to decrease, and the prediction accuracy to increase.

The manuscript lacks clarity on how the alignment of structural genes operates when dealing with multiple NRPS gene clusters on different genome contigs. How would the alignment of each BGC work?

We thank the reviewers for this comment. The pyoverdine molecules consist of a conserved fluorescent chromophore (Flu) and a peptide chain (Pep), both synthesized by NRPS enzymes. In most instances (over 90%), Flu and Pep are produced by two separate biosynthetic gene clusters (BGCs). In these cases, we merge the two BGCs by positioning Flu at the head and Pep at the tail. For the remaining less than 10%, there are two scenarios: 1. Flu and Pep are located on the same BGC, which eliminates any issues with BGC alignment. 2. In very rare cases, Flu and Pep are synthesized by three BGCs. Here, Flu is still synthesized by one BGC at the head, while Pep is produced by two BGCs. We put the BGC containing the Thioesterase (TE) domain as the tail and the BGC not containing the TE domain in the middle.

(see lines 165-169).

Another critical concern is that a main challenge in NRPS structure prediction is not the backbone prediction but rather the prediction of tailoring reactions, which is not addressed in the manuscript at all, and this limitation extensively restricts the applicability of the method.

While we thank the reviewer for this comment, we only partly agree with it. Peptide backbone predictions are still a significant challenge. This challenge is clearly visible in our new analysis comparing prediction accuracies of different pipelines, such as antiSMASH5, PRISM4, AdenPredictor, SeMPI2, NP.searcher, Sandpuma. Unresolved and wrong substrate predictions are still common, highlighting the importance of our contribution in developing a new approach with improved high accuracy. 

However, we agree with the reviewer that our current algorithm does not predict tailoring reactions (now discussed on lines 680-685). Although tailoring reactions are important for predicting the final NRPS product structure, none of the other existing pipelines address this issue either, and it remains a challenge for future work. For our study, it is important to note that the specificity of pyoverdines is primarily determined by the backbone composition, whereas tailoring reactions seem to play a minor role.

The manuscript presents a potentially highly useful bioinformatic pipeline for pyoverdine structure prediction, showcasing a commendable exploration of siderophore diversity. However, some of the claims made remain unsubstantiated. Overall, while the study holds promise, further validation and refinement are required to fulfill its potential impact on the field of bioinformatic structure prediction.

Thank you for the supportive and encouraging words. We deeply appreciate your constructive comments and suggestions. 

Reviewer #2 (Public Review):

Pyoverdines, siderophores produced by many Pseudomonads, are one of the most diverse groups of specialized metabolites and are frequently used as model systems. Thousands of Pseudomonas genomes are available, but large-scale analyses of pyoverdines are hampered by the biosynthetic gene clusters (BGCs) being spread across multiple genomic loci and existing tools' inability to accurately predict amino acid substrates of the biosynthetic adenylation (A) domains. The authors present a bioinformatics pipeline that identifies pyoverdine BGCs and predicts the A domain substrates with high accuracy. They tackled a second challenging problem by developing an algorithm to differentiate between outer membrane receptor selectivity for pyoverdines versus other siderophores and substrates. The authors applied their dataset to thousands of Pseudomonas strains, producing the first comprehensive overview of pyoverdines and their receptors and predicting many new structural variants.

The A domain substrate prediction is impressive, including the correction of entries in the MIBiG database. Their high accuracy came from a relatively small training dataset of A domains from 13 pyoverdine BGCs. The authors acknowledge that this small dataset does not include all substrates, and correctly point out that new sequence/structure pairs can be added to the training set to refine the prediction algorithm. 

The authors could have been more comprehensive in finding their training set data. For instance, the authors claim that histidine "had not been previously documented in pyoverdines", but the sequenced strain P. entomophila L48, incorporates His (10.1007/s10534-009-9247-y). 

Thank you for highlighting this issue. We agree that stating histidine has not been reported before in pyoverdine was incorrect. We have reviewed the full text and made the necessary corrections.

The primary reason for excluding the sequenced strains P. syringae 1448a (10.1186/14712180-11-218) and P. entomophila L48 (10.1007/s10534-009-9247-y) from the training set is that the pyoverdine structures of these strains were not determined solely through experimental methods. In these works, the pyoverdine structures were predicted based on the synthetic gene sequence using bioinformatical analysis, followed by structural analysis experiments based on this predicted structure. We found that pre-prediction probably has introduced biases into downstream analyses. Specifically, in the case of Pseudomonas entomophila L48, we discovered inaccuracies in the annotation of certain domains (see figures below). For example, the third A domain of the peptide chain in P. entomophila L48 pyoverdine was initially annotated with Dab specificity. However, upon closer examination, it appears to differ significantly from other Dab references (top) or Dab from our experimentally validated (right) domains (left panel in the figure below). By analyzing the interface (I) domain (10.1073/pnas.1903161116) in its predicted site, we suggested that it should actually recognize OHHis. The OHAsp domain of P. entomophila L48 reported in the paper is actually close in sequence similarity to the OHAsp domain (left panel in the figure below), while the Ala domain reported is more similar to the Ser domain (right panel in the figure below). For these reasons, we did not include this supervised pyoverdine structure analysis strain in the training set data.

Author response image 1.

The workflow cannot differentiate between different variants of Asp and OHOrn, and it's not clear if this is a limitation of the workflow, the training data, or both. 

Thanks for pointing this out. It is generally challenging to differentiate between variants of the same amino acid (for all the algorithms existing to date). In this sense, it is a limitation of our but also of all other workflows. Nonetheless, we wish to stress that we observed feature sequence divergence (using the A motif4-5 region), which helped us to separate some (but not all) of the Asp and Orn variants. For example, separations between Asp-variants are distinct (left panel in the figure below). To be on the conservative side, we only differentiated between OHAsp and Asp for our predictions, but also differentiation between DOHAsp and OHAsp would be possible. In the case of Orn-variants, there was a clear separation between Orn and the OHOrn variants (right panel). In contrast, it was difficult to differentiate between the subgroups of OHOrn variants. We believe that no A domain prediction tool will be able to solve this issue. Instead, it would be important to include information on substrate-modifying enzymes in future approaches.

Author response image 2.

The prediction workflow holds up well in Burkholderiales A domains, however, they fail to mention in the main text that they achieved these numbers by adding more A domains to their training set.

We thank the reviewers for this comment. We apologize for not having mentioned the training data set in the main text, while we described it in detail in the methods section (lines 714-732). We now provided more details on the analysis procedure in the main text (lines 307313). Important to note is that we did not add more A domains to the training data set but built up a new independent data set for Burkholderiales. The aim was to mirror the analysis we performed for pyoverdines with a completely new data set, featuring 124 A domains for training and 178 A domains as test set.

To validate their predictions, they elucidated structures of several new pyoverdines, and their predictions performed well. However, the authors did not include their MS/MS data, making it impossible to validate their structures. In general, the biggest limitation of the submitted manuscript is the near-empty methods section, which does not include any experimental details for the 20 strains or details of the annotation pipeline (such as "Phydist" and "Syndist"). The source code also does not contain the requisite information to replicate the results or re-use the pipeline, such as the antiSMASH version and required flags. That said, skimming through the source code and data (kindly provided upon request) suggests that the workflow itself is sound and a clear improvement over existing tools for pyoverdine BGC annotation.

Thank you for highlighting these issues. We agree that the methods section is short. This is because the entire paper is a step-by-step methodological introduction to our pipeline. We have now carefully revised the main text to add the information requested by the reviewer. Moreover, we have included a supplementary file with the MS/MS data of the experimentally analyzed pyoverdine structures. Finally, we further include a link to a one-click online notebook that can be used to replicate the annotation and substrate prediction results See: https://drive.google.com/drive/folders/1JsfyPUGDTFo8BDDZk8JLSvKry8emzMhr?usp=drive_ link , following a more detail explanation on code.

Predicting outer membrane receptor specificity is likewise a challenging problem and the authors have made a promising achievement by finding specific gene regions that differentiate the pyoverdine receptor FpvA from FpvB and other receptor families. Their predictions were not tested experimentally, but the finding that only predicted FpvA receptors were proximate to the biosynthesis genes lends credence to the predictive power of the workflow. The authors find predicted pyoverdine receptors across an impressive 468 genera, an exciting finding for expanding the role of pyoverdines as public goods beyond Pseudomonas. However, whether or not these receptors can recognize pyoverdines (and if so, which structures!) remains to be investigated.

Thank you for the supportive and encouraging words. The bioinformatic analysis and experimental testing of pyoverdine-receptor matching is complicated and it is not part of this paper. We treated it in a separate manuscript in which we developed an experimentally verified co-evolution algorithm that matches pyoverdines to receptors. With this algorithm, we can identify self-receptors (i.e. receptors used to take up the self-produced pyoverdine), and therefore establish pyoverdine sharing and interaction networks across strains in communities.

Please see DOI:10.1101/2023.11.05.565711 for details.

In all, the authors have assembled a rich dataset that will enable large-scale comparative genomic analyses. This dataset could be used by a variety of researchers, including those studying natural product evolution, public good eco/evo dynamics, and NRPS engineering.

Thank you for the supportive and encouraging words. We are grateful for the reviewers’ instructive suggestions and comments.

Reviewer #3 (Public Review):

Summary:

Secondary metabolites are produced by numerous microorganisms and have important ecological functions. A major problem is that neither the function of a secondary metabolite enzyme nor the resulting metabolite can be precisely predicted from gene sequence data.

In the current paper, the authors addressed this highly relevant question.

The authors developed a bioinformatic pipeline to reconstruct the complete secondary metabolism pathway of pyoverdines, a class of iron-scavenging siderophores produced by Pseudomonas spp. These secondary metabolites are biosynthesized by a series of nonribosomal peptide synthetases and require a specific receptor (FpvA) for uptake. The authors combined knowledge-guided learning with phylogeny-based methods to predict with high accuracy encoding NRPSs, substrate specificity of A domains, pyoverdine derivatives, and receptors. After validation, the authors tested their pipeline with sequence data from 1664 phylogenetically distinct Pseudomonas strains and were able to determine 18,292 enzymatic A domains involved in pyoverdine synthesis, reliably predicted 97.8% of their substrates, identified 188 different pyoverdine molecule structures and 4547 FpvA receptor variants belonging to 94 distinct groups. All the results and predictions were clearly superior to predictions that are based on antiSMASH. Novel pyoverdine structures were elucidated experimentally by UHPLC-HR-MS/MS.

To assess the extendibility of the pipeline, the authors chose Burkholderiales as a test case which led to the results that the pipeline consistently maintains high prediction accuracy within Burkholderiales of 83% which was higher than for antiSMASH (67%).

Together, the authors concluded that supervised learning based on a few known compounds produced by species from the same genus probably outperforms generalized prediction algorithms trained on many products from a diverse set of microbes for NRPS substrate predictions. As a result, they also show that both pyoverdine and receptor diversity have been vastly underestimated.

Strengths:

The authors developed a very useful bioinformatic pipeline with high accuracy for secondary metabolites, at least for pyoverdines. The pipelines have several advantages compared to existing pipelines like the extensively used antiSMASH program, e.g. it can be applied to draft genomes, shows reduced erroneous gene predictions, etc. The accuracy was impressively demonstrated by the discovery of novel pyoverdines whose structures were experimentally substantiated by UHPLC-HR-MS/MS.

The manuscript is very well written, and the data and the description of the generation of pipelines are easy to follow.

Weaknesses:

The only major comment I have is the uncertainty of whether the pipeline can be applied to more complex non-ribosomal peptides. In the current study, the authors only applied their pipeline to a very narrow field, i.e., pyoverdines of Pseudomonas and Burkholderia strains.

Thanks for your positive and encouraging comment. Regarding your only major comment, we think that the design concept of our pipeline has the potential to be applied to more complex non-ribosomal peptides. Currently, our method is tailored to accurately predict the structural composition of the Pseudomonas siderophore pyoverdine (see also response 3). A key point emphasized in our article is the importance of considering phylogeny in developing substrate prediction algorithms for A domains. Currently, the main challenge in advancing these algorithms is the limited availability of data on A domains and their corresponding substrates. However, with the future accumulation of more reference data, we are confident that the design principles of our method will enable precise predictions of the structural compositions of all products synthesized by non-ribosomal peptide synthetases (see our discussions in lines 634-

645). 

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

I believe that the manuscript would benefit from focusing solely on the task of improving pyoverdine predictions. This aspect alone is significant, and robustly supporting this claim would strengthen the manuscript. The diversity analysis provided is valuable and would undoubtedly benefit the scientific community. However, additional systematic comparisons with other methods are necessary. Furthermore, clarification of certain terms, such as 'featurebased' (e.g., whether it refers to NRPS domains or CDS), would enhance clarity.

Thank you for the supportive and encouraging words. We followed the reviewer’s suggestion and now provide the requested method comparison, see also response 2 for details. Furthermore, we have carefully checked the main text to clarify terms whenever needed. Specifically, we now define the terms “feature sequence” and “feature sequence distance” in lines 227-229.  

Additionally, several minor points could be improved upon:

In line 85, clarification is needed on how pyoverdine genes were identified.

Thank you for your thorough review. In the introduction section, we provided a brief overview of our work, while the detailed methodology is outlined in the results section on lines 160-174.

In line 382, it would be helpful to know the source of the sequences.

We agree and have now carefully revised the manuscript following your suggestions (lines 403-405).

Line 392 could be explained more clearly. Does it mean that the authors used an hmm search to search pHMMs against each reference sequence?

Thanks for your comment. Yes, we used an hmm search to search pHMMs against each reference sequence. We have now revised the manuscript to improve explanations (lines 413-418).

Reviewer #2 (Recommendations For The Authors):

The authors state they "elucidated the chemical structure of the 20 pyoverdines using culturebased methods combined with UHPLC-HR-MS/MS", so I was alarmed to see that KR and LB already published several of those structures in the cited paper. I hope that this "double dipping" will be fixed in a revision process.

Thank you for pointing this out. We agree that we have not explained clearly enough what steps were conducted in this study and which data were used from a previous paper (https://doi.org/10.1007/s00216-022-03907-w). The genomes of the 20 strains used for the verification analysis (Fig. 3d) were sequenced as part of this study (access code now provided). 14 out of the 20 pyoverdine structures were elucidated with UHPLC-HR-MS/MS in this study. For 6 out of the 20 pyoverdines, we had structural information already at hand from the previous paper. We have now clarified these details in our manuscript (lines 276-280). 

Thank you for providing the source code and data, and I hope that the final non-redundant dataset will be uploaded to Zenodo or another repository. Please deposit the 20 newlysequenced genomes to GenBank or another public repository. Please also show the UHPLC-

HR-MS/MS data, preferably in the form of raw data uploaded to GNPS.

We have followed the reviewer’s advice and deposited our data:

- The sequences of the 20 newly sequenced strains are available on ENA accession PRJEB76792.

- The MS/MS plots of the 14 newly analyzed pyoverdines are shown in the Supplementary Materials.

- We provide a one-click online notebook to allow readers to replicate the pyoverdine cluster annotation and substrate prediction of the 20 experimentally analyzed strains.

I suggest adding "at least" or a similar qualifier when the 73 variants are mentioned unless the literature search was truly exhaustive. What were the criteria for inclusion of the 13 strains in Table S2? For instance, sequenced strains P. syringae 1448a (10.1186/1471-2180-11-218) and P. entomophila L48 (10.1007/s10534-009-9247-y) were not included.

Thank you for your comment. We have now carefully revised the manuscript following your suggestions (lines 291-295). Regarding the criteria for including the 13 strains in Table S2, we aimed to select strains with the high credibility for inclusion in the training set data. The primary reason for excluding the two strains from the training set is that their siderophore structures were analyzed through supervised experiments. We wanted to avoid any form of biases that bioinformatic pre-predictions could introduce to downstream analyses (see Response 13 for details).

OHAsp in pyoverdines has been reported to arise from hydroxylation of Asp after it's already been activated by the A domain (10.1073/pnas.1903161116). Was there a clear difference between A domains that lead to Asp and OHAsp? Conversely, acetylation and formylation of OHOrn occur before adenylation. Can your workflow be used to differentiate cOHOrn, fOHOrn, and AcOHOrn, which are currently difficult to predict through genome mining?

Thank you for these considerations. We treated these aspects in our response 8.  

Throughout, define non-proteinogenic AA substrate abbreviations (ex: Rsc, Dab).

Revised as per suggestion (lines 329-333).

Additional line comments:

189: Mention PhyloPhlAn in the main text.

Revised as per suggestion (lines 189).

191: Define these filtering/selection criteria.

Thanks for your comment, we have added the criteria in the main text (line 196 and line 198). 

309, 620: An A domain presumably loading histidine is present in sequenced strain P. entomophila L48 (10.1007/s10534-009-9247-y). Please also clarify that Val has previously been seen in a pyoverdine (it is in Table S1) albeit not sequenced.

We have clarified these aspects as per suggestion (lines 314-315 and line 630).

310: The pipeline can "highlight" new substrates, but not identify them.

Revised as per suggestion (line 295).

354: Please clarify "13 amino acid substrates form the core of all the 188 pyoverdine structures", considering that 279 A domain substrates couldn't be predicted.

Thanks for your comments. We have now clarified “our analysis found that 13 amino acids form the main structural substrates of all the 188 pyoverdine structures.” (lines

360-363)

630: "discovered" implies that there is experimental evidence. I suggest something like "here we predicted 151 putatively new variants".

Revised as per suggestion (line 648).

Reviewer #3 (Recommendations For The Authors):

Weakness:

The only major comment I have is the uncertainty of whether the pipeline can be applied to more complex non-ribosomal peptides. In the current study, the authors only applied their pipeline to a very narrow field, i.e., pyoverdines of Pseudomonas and Burkholderia strains

Thanks for your comment. Please see our Responses 3+13 above, where we treat this concern in detail. Moreover, we discussed the possibility of extension to other groups of secondary metabolites in our discussion. We believe that we deliver a balanced view on the applicability of our approach and the next steps to be taken.  

Please comment on this aspect.

Minor:

(1)  When you speak about "synthesis" it is rather biosynthesis. Synthesis is chemical synthesis.

Please replace all instances of the word synthesis with biosynthesis.

Revised as per suggestion.

(2)  Line 188: synthetase is rather synthetases

Revised as per suggestion (line 191).

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1:

Point 1: While the manuscript is methodologically sound, the following aspects of image acquisition and data analysis need to be clarified to ensure replicability and reproducibility. The authors state that the sample is a "population-derived adult lifespan sample", the lack of demographic information makes it impossible to know if the sample is truly representative. Though this may seem inconsequential, education may impact both cognitive performance and functional activation patterns. Moreover, the authors do not report race/ethnicity in the manuscript. This information is essential to ensure representativeness in the sample. It is imperative that barriers to study participation within minoritized groups are addressed to ensure rigor and reproducibility of findings.

First, the section Methods-Participants has been updated to refer readers to a prior article where the sample’s demographics are broken down into nine decile age groups (see Wu et al. 2023 Table 1), including information about their education levels. Secondly, we have updated the Data Availability section text to indicate that all Cam-CAN IDs are included in the available OSF datasets, allowing anyone to verify additional participant demographics described in the Cam-CAN protocol article (Shafto et al., 2014). Third, we have updated the Participants section text to refer to another prior study that reported on the representativeness of the Cam-CAN sample indicating that at least some elements of the sample have been independently deemed as representative (e.g., Sex).

Page-24

“A healthy population-derived adult lifespan human sample (N = 223; ages approximately uniformly distributed from 19 - 87 years; females = 112; 50.2%) was collected as part of the Cam-CAN study (Stage 3 cohort; Shafto et al., 2014). Participants were fluent English speakers in good physical and mental health, based on the Cam-CAN cohort’s exclusion criteria which includes poor mini mental state examination, ineligibility for MRI and medical, psychiatric, hearing or visual problems. Throughout analyses, age is defined at the Home Interview (Stage 1; Shafto et al., 2014). The study was approved by the Cambridgeshire 2 (now East of England–Cambridge Central) Research Ethics Committee and participants provided informed written consent. Further demographic information of the sample is reported in Wu et al. (2023) and is openly available (see section Data Availability) with a recent report indicating the representativeness of the sample across sexes (Green et al., 2018).”

Page-30

“Raw and minimally pre-processed MRI (i.e., from automatic analysis; Taylor et al., 2017) and behavioural data are available by submitting a data request to Cam-CAN (https://camcan-archive.mrc-cbu.cam.ac.uk/dataaccess/). The univariate and multivariate ROI data, and behavioural data, can be downloaded from the Open Science Framework, which includes Cam-CAN participant identifiers allowing the retrieval of any additional demographic data (https://osf.io/v7kmh), while the analysis code is available on GitHub.”

Point 2: For the whole-brain analysis in which the ROIs were derived, the authors used a threshold-free cluster enhancement (TFCE; Smith & Nichols 2009). The methodological paper cited suggests that individuals' TCFE image should still be corrected for multiple comparisons using the following: "to correct for multiple comparisons, one [...] has to build up the null distribution (across permutations of the input data) of the maximum (across voxels) TFCE score, and then test the actual TFCE image against that. Once the 95th percentile in the null distribution is found then the TFCE image is simply thresholded at this level to give inference at the p < 0.05 (corrected) level." (Smith & Nichols, 2009). Although the authors mention that clusters were estimated using 2000 permutations, there is no mention of the TFCE image itself being thresholded. While this would impact the overall size of the ROIs used in the study, the remaining analyses are methodologically sound.

We have updated the text to detail the t=1.97 (i.e., p = .05) threshold we applied before interpretation of the resultant TFCE images to the section: Experimental Design & Statistical Analysis. This threshold value can also be verified in the analytics code that is referenced on GitHub from the section Data Availability within the requisite toolbox functions: https://github.com/kamentsvetanov/CommonalityAnalysis/blob/main/code/ca_vba_tfce_threshold.m#L24 and https://github.com/kamentsvetanov/CommonalityAnalysis/blob/main/code/external/ca_matlab_tfce_transform.m

Page-30

“For whole-brain voxelwise analyses, clusters were estimated using threshold-free cluster enhancement (TFCE; Smith & Nichols 2009) with 2000 permutations and the resulting images were thresholded at a t-statistic of 1.97 before interpretation.”

Point 3: The authors should consider moving the ROI section to results. The way the manuscript currently reads, the ROIs seem to be derived a priori as opposed to being derived from activation maps in the current study.

After consideration of this point, we have decided to leave the methodological details regarding the definition of ROIs in the methods, to maintain the focus of the Results section. However, we have improved signposting in the results section to highlight that the ROIs were derived from the overlapped activation maps.

Page-8

“Crucially, two areas of the brain showed spatially-overlapping positive effects of age and performance, which is suggestive of an age-related compensatory response (Figure 2A yellow intersection). These were in bilateral cuneal cortex (Figure 2B magenta) and bilateral frontal cortex (Figure 2B brown), the latter incorporating parts of the middle frontal gyri and anterior cingulate. Therefore, based on traditional univariate analyses, these are two candidate regions for age-related functional compensation (Cabeza et al. 2013; 2018). Accordingly, we defined regions of interest within these two regions using the overlap activation maps (see section: ROIs) to be used for subsequent univariate and multivariate analysis.”

Point 4: The manuscript can be strengthened by explaining why the authors chose a greedy search algorithm over a dynamic Bayesian model.

The text is updated to refer to appropriateness of the computationally efficient greedy search implementation, due to the size of the fMRI cohort dataset.

Page-28

“The pattern weights specifying the mapping of data features to the target variable are optimized with a greedy search algorithm using a standard variational scheme (Friston et al., 2007) which was particularly appropriate given the large dataset.”

Reviewer #2:

Point 1: However, it might have been nice to see an analysis of a more crystallised intelligence task included too, as a contrast since this is an area that does not demonstrate such a decline (and perhaps continues to improve over aging).

We (Samu et al., 2017) have previously investigated, but failed to find, univariate evidence for functional compensation in this cohort’s performance on a sentence comprehension task that is more closely aligned to a measure of crystallised intelligence. Based on the additional previous studies where we have applied these types of univariate and multivariate criteria of functional compensation (Morcom & Henson, 2018; Knights et al., 2021), we have consistently observed that the uni-/multivariate effects are in the same direction. Therefore, we would not initially expect a different conclusion here, where the univariate and multivariate effects suggest different outcomes. Notably, the univariate analysis approach in Samu et al. (2017) did differ from focusing on the age x behaviour interaction term here, so it could still be worth future investigation, but it does seem less likely that evidence of compensation would be observed than for fluid intelligence. However, as the Reviewer suggests, such a task may make another good contrast to show evidence against the existence of functional compensation (as in Morcom & Henson, 2018; Knights et al., 2021).

Point 2: Figure 1B: Consider adding coefficients describing relationships to plots.

Annotations of the coefficients have been added to Figure 1B:

Point 3: Figure 2C. The scale of the axis for RSFA-Scales cuneal cortex ROI activations should be the same as the other 3 plots.

Figure axes are updated such that ROIs are on matching scales, according to whether data were RSFA-scaled or not.

Point 4: Figure 2C. Adding in the age ranges for each of the three groups following the tertile split may be informative to the reader.

The age group tertile definition used for Figure 2C visualisations is now added to the Figure description.

Page-10

“Figure 2. Univariate analysis. (A) Whole-brain effects of age and performance. Age (green) and performance (red) positively predicted unique aspects of increased task activation, with their spatial overlap (yellow) being overlaid on a template MNI brain, using p < 0.05 TFCE. (B) Intersection ROIs. A bilateral cuneal (magenta) and frontal cortex (brown) ROI were defined from voxels that showed a positive and unique effect of both age and performance (yellow map in Figure 2A). (C) ROI Activation. Activation (raw = left; RSFA-scaled = right) is plotted against behavioural performance based on a tertile split between three age groups (19-44, 45-63 & 64-87 years).”

Reviewer #3:

Point 1: [Public Review] 1) I don't quite follow the argumentation that compensatory recruitment would need to show via non-redundant information carried by any given non-MDN region (cf. p14). Wouldn't the fact that a non-MDN region carries task-related information be sufficient to infer that it is involved in the task and, if activated increasingly with increasing age, that its stronger recruitment reflects compensation, rather than inefficiency or dedifferentiation? Put differently, wouldn't "more of the same" in an additional region suffice to qualify as compensation, as compared to the "additional information in an additional region" requirement set by the authors? As a consequence, in my honest opinion, showing that decoding task difficulty from non-MDN ROIs works better with higher age would already count as evidence for compensation, rather than asking for age-related increases in decoding boosts obtained from adding such ROIs. It would be interesting to see whether the arguably redundant frontal ROI would satisfy this less demanding criterion. At any rate, it seems useful to show whether the difference in log evidence for the real vs. shuffled models is also related to age.

We agree with the logic for conducting a weaker assessment of functional compensation whereby a brain region does not necessarily have to provide a unique contribution beyond that of the ordinarily activated task-relevant network. However, although non-unique recruitment is predicted by a compensation theory, it can also be explained by a nonspecific mechanism that recruits multiple regions in tandem. In contrast, unique additional recruitment is compatible with compensation but not with nonspecific recruitment. In this article, and those prior (Morcom & Henson, 2018; Knights et al. 2021), we have also deliberately avoided using the specific kind of analysis proposed (i.e., testing for an effect of age on differential log evidence) because these would involve applying statistical tests directly to the log evidence, a variable that is already a statistical test output.

Nevertheless, temporarily putting these caveats aside, we did run the suggested test. Results from multiple regression showed that using log evidence from frontal cortex models still did not meet this less demanding criterion for functional compensation as there was an effect of age in the opposite direction to that expected by functional compensation: there was a significant negative effect of age (t(218) = -7.95, p = < .001) indicating that as age increased, the difference in log evidence decreased. This effect is visualised below for transparency, but we preferred not to add this information to the article because we do not wish to encourage using this kind of analysis for the reason mentioned above. Thus, although our main multivariate test of interest is stringent, the additional step of mapping log evidence back to the boost-likelihood categories (e.g., boost vs. no difference to model performance) lends itself to the more appropriate logistic regression statistical approach.

Author response image 1.

Negative effect of age on MVB log evidence model outcomes for frontal cortex.

A different approach that could be taken to assess a more lenient definition of functional compensation would be to analyse the effects of age on the spread of multivariate responses predicting task difficulty (i.e., standard deviation of fitted MVB voxel weights; also see Morcom & Henson, 2018; Knights et al., 2021) specifically from models that only include the candidate ‘compensation’ ROIs.

Accordingly, these analyses and their discussion have been added to the article. To summarise, these analyses showed that (1) the frontal cortex still did not show evidence of functional compensation (i.e., a negative effect of age like in Morcom & Henson, 2018) and (2) no effect of age on the cuneal ROI, implying that the original model comparison approach (i.e., Figure 2C in the manuscript now) can provide more sensitivity for detecting evidence of functional compensation (perhaps because of the importance of including task-relevant network responses when building decoding models).

Page-15

“As a final analysis, we also tested a more lenient definition of functional compensation, whereby the multivariate contribution from the “compensation ROI” does not necessarily need to be above and beyond that of the task-relevant network (Morcom & Henson, 2018; Knights et al., 2021). To do this, we again assessed whether age was associated with an increase in the spread (standard deviation) of the weights over voxels, for smaller models containing only the cuneal or frontal ROI. This tested whether increased age led to more voxels carrying substantial information about task difficulty, a pattern predicted by functional compensation (but also consistent with non-specific additional recruitment). In this case, the results of this test did not support functional compensation, as there was no effect detected for the cuneal cortex and even a negative effect of age for the frontal cortex where the spread of the information across voxels was lower for older age (Figure 3C; Table 2).”

Page-21

“The age- and performance-related activation in our frontal region satisfied the traditional univariate criteria for functional compensation, but our multivariate (MVB) model comparison analysis showed that additional multivariate information beyond that in the MDN was absent in this region, which is inconsistent with the strongest definition of compensation. In fact, the results from the spread analysis showed that as age increased, this frontal area processed less, rather than more, multivariate information about the cognitive outcome (Figure 3C) as previously observed in two (memory) tasks for a comparable ROI within the same Cam-CAN cohort (Morcom & Henson, 2018).”

Page-24

“This said, univariate criteria for functional compensation will continue to play a role in hypothesis testing. For instance, the over-additive interaction observed in the cuneal cortex - where the increase in activity with better performance is more pronounced in older adults - offers stronger evidence of compensation compared to the simple additive effect of age and performance observed in the frontal cortex (Figure 2C). So far, the two studies that have combined these rigorous univariate, behavioral and multivariate approaches to assess functional compensation (i.e., Knights et al., 2021; the present study) have generally found converging evidence regardless of the method used. However, it is important to note that the MVB approach uniquely shifts the focus from individual differences to the specific task-related information that compensatory neural activations are assumed to carry and provides a specific test of region- (or network-) unique information. With further studies, it may also be that multivariate approaches prove more sensitive for detecting compensation effects than when using mean responses over voxels (e.g., Friston et al., 1995) particularly since over-additive effects are challenging to observe because compensatory effects are typically ‘partial’ and do not fully restore function (for review see Scheller et al., 2014; Morcom & Johnson, 2015). Within the multivariate analysis options themselves, it is also interesting to highlight that the stringent MVB boost likelihood analysis could detect functional compensation unlike the more lenient analysis focusing on the spread of MVB voxel weights. This suggests the importance of including task-relevant network responses when building decoding models to assess compensation.”

Page-32

“Alongside the MVB boost analysis, we also included an additional measure using the spread (standard deviation) of voxel classification weights (Morcom & Henson, 2018). This measure indexes the absolute amplitude of voxel contributions to the task, reflecting the degree to which multiple voxels carry substantial task-related information. When related to age this can serve as a multivariate index of information distribution, unlike univariate analyses. However, it is worth highlighting that even if an ROI shows an effect of age on this spread measure, such an effect could instead be explained by a non-specific mechanism that represents the same information in tandem across multiple regions (rather than reflecting compensation) as seen previously (Knights et al., 2021; also see Morcom & Johnson, 2015). Thus, it is the MVB boost analysis that is the most compelling assessment of functional compensation because it can directly detect novel information representation.”

Point 2: [Public Review] 2) Relatedly, does the observed boost in decoding by adding the cuneal ROI (in older adults) really reflect "additional, non-redundant" information carried by this ROI? Or could it be that this boost is just a statistical phenomenon that is obtained because the cuneus just happens to show a more clear-cut, less noisy difference in hard vs. easy task activation patterns than does the MDN (which itself may suffer from increased neural inefficiency in older age), and thus the cuneaus improves decoding performance without containing additional (novel) pieces of information (but just more reliable ones)? If so, the compensation account could still be maintained by reference to the less demanding rationale for what constitutes compensation laid out above.

We agree that this is a possibility and have added this as an additional explanation to the Discussion. We have also discussed why we think it is a less likely possibility, but do concede that it cannot be ruled out currently.

Page-20

“Another possibility is that the age-related increases in fMRI activations (for hard versus easy) in one or both of our ROIs do not reflect greater fMRI signal for hard problems in older than younger people, but rather lower fMRI signal for easy problems in the older. Without a third baseline condition, we cannot distinguish these two possibilities in our data. However, a reduced “baseline” level of fMRI signal (e.g., for easy problems) in older people is consistent with other studies showing an age-related decline in baseline perfusion levels, coupled with preserved capacity of cerebrovascular reactivity to meet metabolic demands of neuronal activity at higher cognitive load  (Calautti et al., 2001; Jennings et al., 2005). Though age-related decline in baseline perfusion occurs in the cuneal cortex (Tsvetanov et al., 2021), the brain regions showing modulation of behaviourally-relevant Cattell fMRI activity by perfusion levels did not include the cuneal cortex (Wu et al., 2023). This suggests that the compensatory effects in the cuneus are unlikely to be explained by age-related hypo-perfusion, consistent with the minimal effect here of adjusting for RSFA (Figure 2C).

One final possibility is whether the observed boost in decoding from adding the cuneal ROI simply reflects less noisy task-related information (i.e., a better signal-to-noise ratio (SNR)) than the MDN and, consequently, the boosted decoding is the result of more resilient patterns of information (rather than the representation of additional information) based on a steeper age-related decline of SNR in the MDN. Overall then, as none of the explanations above agree with all aspects of the results, to functionally explain the role of the cuneal cortex in this task would require further investigation.”

Point 3: [Public Review] 3) On page 21, the authors state that "...traditional univariate criteria alone are not sufficient for identifying functional compensation." To me, this conclusion is quite bold as I'd think that this depends on the unvariate criterion used. For instance, it could be argued that compensation should be more clearly indicated by an over additive interaction as observed for the relationship of cuneal activity with age and performance (i.e., the activity increase with better performance becomes stronger with age), rather than by an additive effect of age and performance as observed for the prefrontal ROI (see Fig. 2C). In any case, I'd appreciate it if the authors discussed this issue and the relationship between univariate and multivariate results in more detail (e.g. how many differences in sensitivity between the two approaches have contributed), in particular since the sophisticated multivariate approach used here is not widely established in the field yet.

We have now considered this point further in a section of the Discussion (which is merged with points 1 & 2 above) about the relevance and distinction of univariate / multivariate criteria for functional compensation. As described in text below, whilst we agree that univariate / behavioural approaches have a role in testing functional compensation, we still view the MVB boost analysis to be a particularly compelling approach for assessing this theory.

Page-22

“This said, univariate criteria for functional compensation will continue to play a role in hypothesis testing. For instance, the over-additive interaction observed in the cuneal cortex - where the increase in activity with better performance is more pronounced in older adults - offers evidence of compensation compared to the simple additive effect of age and performance observed in the frontal cortex (Figure 2C). However, the conclusions that can be drawn from age-related differences in cross-sectional associations of brain and behaviour are limited, mainly because individual performance differences are largely lifespan-stable (see Lindenberger et al., 2011; Morcom & Johnson, 2015). So far, the two studies that have combined these univariate-behavioral and multivariate approaches to assess functional compensation (i.e., Knights et al., 2021; the present study) have generally found converging evidence regardless of the method used. However, it is important to note that the MVB approach uniquely shifts the focus from individual differences to the specific task-related information that compensatory neural activations are assumed to carry. With further studies, it may also be that multivariate approaches prove more sensitive for detecting compensation effects than when using mean responses over voxels (e.g., Friston et al., 1995) particularly since over-additive effects are challenging to observe because compensatory effects are typically ‘partial’ and do not fully restore function. Within the multivariate analysis options themselves, it is also interesting to highlight that the stringent MVB boost likelihood analysis could detect functional compensation unlike the more lenient analysis focusing on the spread of MVB voxel weights. This suggests the importance of including task-relevant network responses when building decoding models to asses compensation.”

Point 4: [Public Review] 4) As to the exclusion of poorly performing participants (see p24): If only based on the absolute number of errors, wouldn't you miss those who worked (overly) slowly but made few errors (possibly because of adjusting their speed-accuracy tradeoff)? Wouldn't it be reasonable to define a criterion based on the same performance measure (correct - incorrect) as used in the main behavioural analyses?

This is a good point, though if we were to exclude participants using a chance level exclusion rate based on the formulae used for measuring behavioural performance, this removes identical subjects to those originally excluded. Based on this, the text has been updated to reflect this more parsimonious approach for defining exclusion criteria.

Page-25

“In a block design, participants completed eight 30-second blocks which contained a series of puzzles from one of two difficulty levels (i.e., four hard and four easy blocks completed in an alternating block order; Figure 1A). The fixed block time allowed participants to attempt as many trials as possible. Therefore, to balance speed and accuracy, behavioural performance was measured by subtracting the number of incorrect from correct trials and averaging over the hard and easy blocks independently (i.e., ((hard correct - hard incorrect) + (easy correct - easy incorrect))/2; Samu et al., 2017). For assessing reliability and validity, behavioural performance (total number of puzzles correct) was also collected from the same participants during a full version of the Cattell task (Scale 2 Form A) administered outside the scanner at Stage 2 of the Cam-CAN study (Shafto et al., 2014). Both the in- and out-of-scanner measures were z-scored. We excluded participants (N = 28; 17 females) who performed at chance level ((correct + incorrect) / incorrect < 0.5) on the fMRI task, leading to the same subset as reported in Samu et al. (2017).”

Point 5: [Public Review] 5) Did the authors consider testing for negative relationships between performance and brain activity, given that there is some literature arguing that neural efficiency (i.e. less activation) is the hallmark of high intelligence (i.e. high performance levels in the Cattell task)? If that were true, at least for some regions, the set of ROIs putatively carrying task-related information could be expanded beyond that examined here. If no such regions were found, it would provide some evidence bearing on the neural efficiency hypothesis.

No, we did not test for negative relationships between performance and brain activity in this study. However, In Wu et al. (2023) we did specifically test for this and neither of the relevant results reported in section 3.3.1 (i.e., unique relationship between activity and performance) nor section 3.3.2 (i.e., age-related relationship between activity and performance) showed the queried direction of effects. Note that the negative effect in section 3.3.2 (Age U Performance) is a more unique suppression effect representing a positive relationship between performance and activity where this becomes stronger as age is added to the model.

Point 6: [Recommendations for the authors] 1) Page 26: It is not quite clear how the authors made sure their age and performance covariates functioned as independent regressors in the univariate group-level GLM, given the correlation between age and performance (i.e. shared variance).

We included age and performance as covariates (of the age x performance effect of interest) by simply including these as independent regressors in the group-level GLM design matrix in addition to the interaction term (i.e., activity ~ age*performance + covariates equivalent to activity ~ age:performance + age + performance + covariates; Wilkinson & Roger 1973 notation), allowing us to examine the unique variance explained by each predictor (Table 1 and Table 2) and to control for their shared variance.

We should note that while the GLM approach we used accounts for unique and shared effects, it does not explicitly report shared effects in its standard output. To directly examine shared variance, one would need to employ commonality analysis. For reference, results from a commonality analysis on this task have been previously reported in Wu et al. (2023).

Prompted by this point, we have made some further minor improvements to help ensure our methodological steps are reproducible, as highlighted below.

Page-30

“Continuous age and behavioural performance variables were standardised and treated as linear predictors in multiple regression throughout the behavioural (Figure 1B), wholebrain voxelwise (Figure 1C/2A), univariate (Table 1; Figure 1B/2B) and MVB (Table 2; Figure 3) analyses. Throughout, sex was included as a covariate. The models, including interaction terms, can be described, according to Wilkinson & Roger’s (1973) notation, as activity ~ age * performance + covariates (which is equivalent to activity ~ age:performance + age + performance + covariates), allowing us to examine the unique variance explained by each predictor (Table 1) and to control for their shared variance. For whole-brain voxelwise analyses, clusters were estimated using threshold-free cluster enhancement (TFCE; Smith & Nichols 2009) with 2000 permutations and the resulting images were thresholded at a t-statistic of 1.97 before interpretation. Bonferroni correction was applied to a standard alpha = 0.05 based on the two ROIs (cuneal and frontal) that were examined. For Bayes Factors, interpretation criteria norms were drawn from Jarosz & Wiley (2014).”

Point 7: [Recommendations for the authors] 2) Figure 3: I suggest changing the subheading in panel B to "Joint vs. MDN-only Model," in line with the wording in the main text.

The subheading of Figure 3B is updated as suggested to `Joint vs. MDN-only Model`.

Point 8: [Recommendations for the authors] 3) In Figures 1C and 2A, MNI z coordinates should be added to the section views. The appreciation of Figure 2B could be enhanced by adding some rendering with a saggital (medial and/or lateral) view.

The slice mosaics in Figure 1C and 2A are now updated with each slice’s MNI Z coordinates and mentioned in the figure descriptions.

Point 9: [Recommendations for the authors] 4) Page 7 (l. 135): What exactly is meant by "lateral occipital temporal cortex"?

The text is updated to specify the anatomical landmarks that were used for guidance when referring to activation within the lateral occipital temporal cortex, based on ROI criteria definitions used in Knights, Mansfield et al. (2021):

Page-7 Line-135:

“Additional activation was observed bilaterally in the inferior/ventral and lateral occipital temporal cortex (i.e., a cluster around the lateral occipital sulcus that extended anteriorly beyond the anterior occipital sulcus), likely due to the visual nature of the task.”

Point 10: [Recommendations for the authors] 5) On p18ff. (ll. 259-318) the authors discuss in quite some detail how the age-related decoding boost seen with the cuneus ROI can be functionally explained, but it seems like none of the explanations agrees with all aspects of the results. While this is not a major problem for the paper, it may be advisable if this part of the discussion ends with a clearer statement that this issue is not fully solved yet and provides material for future research.

A more direct sentence has been added to make it clear that future investigation will be needed to explain the role of the cuneal cortex here.

Page-20 Line-322:

“Another possibility is that the age-related increases in fMRI activations (for hard versus easy) in one or both of our ROIs do not reflect greater fMRI signal for hard problems in older than younger people, but rather lower fMRI signal for easy problems in the older. Without a third baseline condition, we cannot distinguish these two possibilities in our data. However, a reduced “baseline” level of fMRI signal (e.g., for easy problems) in older people is consistent with other studies showing an age-related decline in baseline perfusion levels, coupled with preserved capacity of cerebrovascular reactivity to meet metabolic demands of neuronal activity at higher cognitive load  (Calautti et al., 2001; Jennings et al., 2005). Though age-related decline in baseline perfusion occurs in the cuneal cortex (Tsvetanov et al., 2021), the brain regions showing modulation of behaviourally-relevant Cattell fMRI activity by perfusion levels did not include the cuneal cortex (Wu et al., 2021). This suggests that the compensatory effects in the cuneus are unlikely to be explained by age-related hypo-perfusion, consistent with the minimal effect here of adjusting for RSFA (Figure 2C). Overall then, as none of the explanations above agree with all aspects of the results, to functionally explain the role of the cuneal cortex in this task will require further investigation.”

Point 11: [Recommendations for the authors] 6) The threshold choice for Bayesian log evidence (> 3) should be motivated in some more detail, rather than just pointing to a book reference, as there is no established convention in the field, the choice may depend on the type of data and/or analysis, and a sizeable part of the readership may not be deeply familiar with the particular Bayesian approach used here.

Text is updated to further clarify our motivation for using the log evidence BF>3 criterion:

Page-29

“The outcome measure was the log evidence for each model (Morcom & Henson, 2018; Knights et al., 2021). To test whether activity from an ROI is compensatory, we used an ordinal boost measure (Morcom & Henson, 2018; Knights et al., 2021) to assess the contribution of that ROI for the decoding of task-relevant information (Figure 3B). Specifically, Bayesian model comparison assessed whether a model that contains activity patterns from a compensatory ROI and the MDN (i.e., a joint model) boosted the prediction of task-relevant information relative to a model containing the MDN only. The compensatory hypothesis predicts that the likelihood of a boost to model decoding will increase with older age. The dependent measure, for each participant, was a categorical recoding of the relative model evidence to indicate the outcome of the model comparison. The three possible outcomes were: a boost to model evidence for the joint vs. MDN-only model (difference in log evidence > 3), ambiguous evidence for the two models (difference in log evidence between -3 to 3), or a reduction in evidence for the joint vs. MDN-only model (difference in log evidence < -3).These values were selected because a log difference of three corresponds to a Bayes Factor of 20, which is generally considered strong evidence (Lee & Wagenmakers, 2014). Further, with uniform priors, this chosen criterion (Bayes Factor > 3) corresponds to a p-value of p<~.05 (since the natural logarithm of 20 equals three, as evidence for the alternative hypothesis).”

Point 12: [Recommendations for the authors] 7) Adding page numbers would be helpful.

Page numbers have been added to the manuscript file – apologies for this oversight.

References

Green, E., Bennett, H., Brayne, C., & Matthews, F. E. (2018). Exploring patterns of response across the lifespan: The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study. BMC Public Health, 18, 1-7.

Knights, E., Mansfield, C., Tonin, D., Saada, J., Smith, F. W., & Rossit, S. (2021). Hand-selective visual regions represent how to grasp 3D tools: brain decoding during real actions. Journal of Neuroscience, 41(24), 5263-5273.

Samu, D., Campbell, K. L., Tsvetanov, K. A., Shafto, M. A., & Tyler, L. K. (2017). Preserved cognitive functions with age are determined by domain-dependent shifts in network responsivity. Nature communications, 8(1), 14743.

Shafto, M. A., Tyler, L. K., Dixon, M., Taylor, J. R., Rowe, J. B., Cusack, R., ... & Cam-CAN. (2014). The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing. BMC neurology, 14, 1-25.

Wu, S., Tyler, L. K., Henson, R. N., Rowe, J. B., & Tsvetanov, K. A. (2023). Cerebral blood flow predicts multiple demand network activity and fluid intelligence across the adult lifespan. Neurobiology of aging, 121, 1-14.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Reviewer):

It is not clear from the analysis presented in the paper how persistent those environmentally induced changes, do they remain with the bats till the end of their lives.

Currently, the long-term effects of enrichment on the bats remain uncertain. Preliminary results suggest that these differences may persist throughout the bats’ lifetimes; however, further data analysis is ongoing to determine the extent of these effects. We also addressed now at the manuscript discussion

Reviewer #2 (Public Reviewer):

(1) Assessing personality metrics and the indoor paradigm: While I applaud this effort and think the metrics used are justified, I see a few issues in the results as they are currently presented:

(a) [Major] I am somewhat concerned that here, the foraging box paradigm is being used for two somewhat conflicting purposes: (1) assessing innate personality and (2) measuring changes in personality as a result of experience. If the indoor foraging task is indeed meant to measure and reflect both at the same time, then perhaps this can be made more explicit throughout the manuscript. In this circumstance, I think the authors could place more emphasis on the fact that the task, at later trials/measurements, begins to take on the character of a "composite" measure of personality and experience.

Personality traits should generally be stable over time, but personality can also somewhat change with experience. We used the foraging box to assess individual personality, but we also examined the assumption that what we are measuring is a proxy of personality and hence is stable over time. We now clarify this in the manuscript. 

(b) [Major] Although you only refer to results obtained in trials 1 and 2 when trying to estimate "innate personality" effects, I am a little worried that the paradigm used to measure personality, i.e. the stable components of behavior, is itself affected by other factors such as age (in the case of activity, Fig. 1C3, S1C1-2), the environment (see data re trial 3), and experience outdoors (see data re trials 4/5).

We found that boldness was the most consistent trait, showing persistence between trials 1 to 5, i.e., 144 days apart on average. We thus also used Boldness as the primary parameter for assessing the effects of personality on the outdoors behavior. While we evaluated other traits for completeness, boldness was the only one that consistently met the criteria for personality, which is why we focused on it in our analyses. The other traits which were not stable over time could be used to assess the effects of experience on behavior

Ideally, a study that aims to disentangle the role of predisposition from early-life experience would have a metric for predisposition that is relatively unchanging for individuals, which can stand as a baseline against a separate metric that reflects behavioral differences accumulated as a result of experience.

I would find it more convincing that the foraging box paradigm can be used to measure personality if it could be shown that young bats' behavior was consistent across retests in the box paradigm prior to any environmental exposure across many baseline trials (i.e. more than 2), and that these "initial settings" were constant for individuals. I think it would be important to show that personality is consistent across baseline trials 1 and 2. This could be done, for example, by reproducing the plots in Fig. 1C1-3 while plotting trial 1 against trial 2. (I would note here that if a significant, positive correlation were to be found (as I would expect) between the measures across trial 1 and 2, it is likely that we would see the "habituation effect" the authors refer to expressed as a steep positive slope on the correlation line (indicating that bold individuals on trial 1 are much bolder on trial 2).)

We agree and thus used boldness which was found to be stable over five trials (three of which were without external experience). We note that if Boldness as we measured it increased over time, the differences between individuals remained similar and this is what is expected from personality traits measured in the same paradigm several times (after the animal acquires experience).  

(c) Related to the previous point, it was not clear to me why the data from trial 2 (the second baseline trial) was not presented in the main body of the paper, and only data from trial 1 was used as a baseline.

We added a main figure, showing the correlation between the two baseline trials

In the supplementary figure and table, you show that the bats tended to exhibit more boldness and exploratory behavior, but fewer actions, in trial 2 as compared with trial 1. You explain that this may be due to habituation to the experimental setup, however, the precise motivation for excluding data from trial 2 from the primary analyses is not stated. I would strongly encourage the authors to include a comparison of the data between the baseline trials in their primary analysis (see above), combine the information from these trials to form a composite baseline against which further analyses are performed, or further justify the exclusion of data as a baseline.

We had no intention of excluding data from baseline 2. As we have shown several times before (e.g., Harten, 2021) bats’ boldness as we measure it in the box experiment increases over sessions performed nearby in time. This means that trial 2’s boldness was higher than that of trial 1 and trial 3 which made the data less suitable for a Linear model. Moreover, our measurement of boldness is capped (with a maximum of 1) again making it less suitable for a Linear model. However, following the reviewer’s question we now ran all analyses with trial 2’s data included and not only that the results remained the same, some of the models fit better (based on the AIC criterion). We added this information to the revised manuscript.  

(2) Comparison of indoor behavioral measures and outdoor behavioral measures Regarding the final point in the results, correlation between indoor personality on Trial 4 and outdoor foraging behavior: It is not entirely clear to me what is being tested (neither the details of the tests nor the data or a figure are plotted). Given some of the strong trends in the data - namely, (1) how strongly early environment seems to affect outdoor behavior, (2) how strongly outdoor experience affects boldness, measured on indoor behavior (Fig. 1D) - I am not convinced that there is no relationship, as is stated here, between indoor and outdoor behavior. If this conclusion is made purely on the basis of a p-value, I would suggest revisiting this analysis.

We agree that the relationship between indoor personality measures and outdoor foraging behavior is of great interest and had expected to find some correspondence between the two. To test this, we conducted multiple GLM analyses using the different indoor behavioral traits as predictors of outdoor behaviors. These analyses did not reveal any significant correlations. We also performed a separate analysis using PC1 (derived from the indoor behavioral variables) as a predictor, and again found no significant associations with outdoor behavior.

We were indeed surprised by this outcome. It is possible that the behavioral traits we assessed indoors (boldness, exploration, and activity) do not fully capture the dimensions of behavior that are most relevant to foraging in the wild. For example, traits such as neophobia or decisionmaking under risk, which we did not assess directly, may have had stronger predictive value for outdoor behavior. We now highlight this point more clearly in the Discussion and acknowledge the possibility that alternative or additional personality traits might have revealed meaningful relationships.

(3) Use of statistics/points regarding the generalized linear models While I think the implementation of the GLMM models is correct, I am not certain that the interpretation of the GLMM results is entirely correct for cases where multivariate regression has been performed (Tables 4s and S1, and possibly Table 3). (You do not present the exact equation they used for each model (this would be a helpful addition to the methods), therefore it is somewhat difficult to evaluate if the following critique properly applies, however...)

The "estimate" for a fixed effect in a regression table gives the difference in the outcome variable for a 1 unit increase in the predictor variable (in the case of numeric predictors) or for each successive "level" or treatment (in the case of categorical variables), compared to the baseline, the intercept, which reflects the value of the outcome variable given by the combination of the first value/level of all predictors. Therefore, for example, in Table 4a - Time spend outside: the estimate for Bat sex: male indicates (I believe) the difference in time spent outside for an enriched male vs. an enriched female, not, as the authors seem to aim to explain, the effect of sex overall. Note that the interpretation of the first entry, Environmental condition: impoverished, is correct. I refer the authors to the section "Multiple treatments and interactions" on p. 11 of this guide to evaluating contrasts in G/LMMS: https://bbolker.github.io/mixedmodelsmisc/notes/contrasts.pdf

We are not certain we fully understand the comment; however, if our understanding is correct, we respectfully disagree. A GLM analysis without interaction terms—as conducted in our study—functions as a multiple linear regression, wherein each factor's estimate reflects its individual effect on the dependent variable. For example in the case of sex, it examines he effect of sex on the tie spent out independently of enrichment. An interaction term would be needed to test sex*enrichment. We have added the models’ formula, and we hope this clarifies our approach

Reviewer #1 (Recommendations for the authors):

I would recommend the following:

(1) As video tracking and behavioral analysis softwares are wide spread, it would be great to see this applied to the bat behavior indoor to answer questions like how does the bat velocity or heading or acceleration correlate with the behavioral measures boldness , activity or exploration? In the same gist, can one infer boldness, activity or exploration from measured bat velocity or other parameters? I think this will further make the indoor behavior more quantitative.

In a tent of the size used in our study, bats’ flight behavior tends to be highly stereotypical: they typically perch on the wall, take off, circle the tent—sometimes multiple times—and then either land or not, and enter or not. Flight velocity is largely determined by individual maneuverability and the physical constraints of the space; thus, precise tracking is unlikely to provide further insight into boldness. In contrast, decision-making behaviors—such as whether to land or enter—more accurately reflect personality traits, as we have shown previously (Harten et al., 2018). Moreover, accurate 3D tracking in such an environment is possible but definitely not easy due to the many blind-spots resulting from the cameras being inside the 3D volume.  Nonetheless, we quantified flight activity and assessed its correlation with the other behavioral axes. As it was highly correlated with general activity, we did not include it as an independent parameter in the main analysis. However, in response to the reviewer’s suggestion, we now present this analysis in the Supplementary Materials.

(2) It is not clear whether the bats come from the same genetic background. they might be but it is not mentioned in the methods under the experimental subjects.

We have shown in the past that there is no familial relations in a randomly caught sample of bats in the colony where we usually work (Harten et al., 2018). The bats were caught in three, not related wild colonies. The text referring to the table was clarified in the revised manuscript

(3) It will be great to include the author's thoughts about mechanisms underlying those environmentally induced changes in behavior in the discussion section along with how this will affect the bats' social foraging abilities. Another question that comes to mind is whether growing up with a large number of bats constitute an enriched environment in itself.

We agree that this could count as an enrichment, and we thus ensured similar group sizes in both groups for this reason. We clarify this in the revised manuscript. 

We have elaborated on the underlying mechanisms in the discussion, focusing on how they contribute to behavioral changes.

Reviewer #2 (Recommendations for the authors):

(1) Outdoor foraging behavior

If I understand correctly, the data you display in Fig. 3A is only from the 2nd to 3rd weeks of exploration, i.e. just before the first post-exploration trial.

What does the data look like for the second outdoor exploration data, i.e. before the final trial?

Is there a specific reason why these measures were only computed on the GPS data from the 3rd week outside? If so, can this sampling of the data be motivated or briefly addressed (in the methods and wherever else necessary)?

In order to allow a comparison between individuals, we had to restrict ourself to a period we had data from many individuals (some dissapeared later on).

Following the reviewer suggestion – we added a supplemenry figure including days 21-26

I would find it important and of great interest to see movement maps for more animals, as these give very rich information that is not entirely captured by the three proxies of outdoor activity.

Are these four exemplary animals sampled from both seasons?

Did you check to see if there were any overall differences in outdoor foraging behavior as a function of the season in which the bats were captured?

Yes, the samples represent individuals from both tested years. This was clarified, and additional examples were included in a supplementary figure.

Variable of time spent outdoors: You mention that you did not include the nights that the bat spent in the colony in these calculations. Did you also look to see if 'the number of nights when the bats left the colony' predicted the bat's earlier enrichment treatment? This could also be interesting to consider.

In response to the reviewer’s comment, we conducted an additional analysis to test whether the proportion of nights each bat spent foraging outside the roost was predicted by its earlier environmental condition (enriched vs. impoverished). We also examined whether sex or age influenced this variable. This analysis showed no significant effect of environmental condition, sex, or age on the proportion of nights spent foraging outside the roost

[Following on point 3 in public review...]

When wishing to discuss the effect/significance of predictors overall, it is common to present the modelling results as an analysis of variance table. See, for example, the two-way anova section (p. 182) in the book Practical Regression and ANOVA using R: https://cran.r-project.org/doc/contrib/Faraway-PRA.pdf

I think the output of passing the model object to an "anova" yields the table that you may be looking for, where the variance accounted for by a predictor is given overall, and not just relative to the first level of all predictors. Naturally, this information can be used in combination with the information provided by the raw model output presented in the paper.

I assume you have done this analysis in R, but am not sure, as the statistical software used is not mentioned. There are several packages in R that allow users to quickly plot the graphical interaction of the parameters they use in models, which aids in interpreting results. It would be good to check results of model fitting in this manner.

Relatedly, I was unable to locate the data and code for this paper using the DOI provided. Neither searching the internet using the doi nor entering the doi on the Mendeley Data website returned the right results. I tried searching Mendeley Data using the senior author's last name, but the most recent entry does not appear to be from this paper. https://data.mendeley.com/datasets/fr48bmnhxj/1

We thank the reviewer for the helpful comment. The analysis was indeed conducted in MATLAB, and this has now been clarified in the manuscript. We have also revised the result tables to improve clarity and included the exact formulas used for each model. Regarding the data availability, the reviewer is correct — the dataset had not yet been published at the time of submission. It is now available at the provided DOI link.

### Suggestions and questions for the present paper, grouped thematically:

[Major] Expansion and development of results: I thought there were many interesting and suggestive points in this data that could be expanded upon. I mention some of these here. While the authors of course do not need to implement all of these suggestions, I think the paper would benefit from a more substantial presentation of this rich data set:

(a) Individual differences as such are not emphasized in the paper so much, as the analyses, particularly those expressed as boxplots, are grouped. The scatter plots in Figure 1 give the richest insight into how individual behavior changes throughout the course of the experiment. I would advocate for the authors to show additional comparisons using such scatter plots (perhaps in the supplementary, if needed).

We thank the reviewer and added scatter plots to figure 2

(b) In the second paragraph of the results, the authors introduce the concept of a pareto front and that of personality archetypes (lines 101-107). I found this very interesting, but these concepts were never reiterated upon later in the results or in the discussion. In fact, at many points, I found myself curious as to how the three indoor measures of personality might be combined to form a composite measure of personality (and likewise for outdoor measures). Have you tried to combine measures into a composite and tried to measure whether this composite metric provides any additional insight into these phenomena? For example, what if you mapped the starting position of each bat as a point in a three-dimensional space, given by the three personality measures, and then evaluated their trajectory through this space with measurements taken at later trials. Could innate personality be interpreted as the starting vector in this space (measured across the two baseline trials)? 

Following the reviewer’s (justified) curiosity we ran a PCA analysis on the behavioral data from trials 1 and 5 and found that there is a significant correlation between the individual scores on PC1. This can be thought of as a measurement that takes both boldness and exploration into account (the weight of activity was very low). We added this information to the revised manuscript and also use this new behavioral parameter as a predisposition in the models (instead of exploration and activity). 

Could environmental exposure be quantified as a warping of the trajectory through this space? Finally, could outdoor experience also be incorporated to evaluate how an individual arrives at its final measurement of personality combined with experience (trial 5)?

The paper currently tries to explain outdoors behavior given personality and not vice versa. While this is a very interesting suggestion, we feel that adding this analysis would make the premise of the paper less clear and since the paper is already somewhat complex, we prefer to leave this analysis for a future study. 

Examining the 3D trajectories of the individuals through the personality space did not reveal any immediate clear pattern (triangles mark the first trial and colours depict the environmental treatment) – 

Author response image 1.

Related to this point: I think the strongest part of the paper is the result showing that bats exposed to enriched environments explore farther, more often, and over larger distances than bats that were raised in an impoverished environment.

We completely agree and tried to further emphasize this  

(c) While these results of the outdoor GPS tracking are very clear, I wish that more information were extracted from the tracking data, which is incredibly rich and certainly can be used to derive many interest parameters beyond those that the authors have shown here. Examples might include: distance travelled (as opposed to estimated km2 or farthest point), a metric of navigational ability (how much "dead reckoning" the animal engages in). I even wonder if the areas or landmarks visited by the enriched bats might be found to be more complex, challenging, or richer by some measure.

This study was a first step, aiming to establish a connection between early exposure and outdoors foraging

We agree that there are many more analyses that can be done and indeed that ones related to navigation capabilities are missing. We are still collecting data on these bats and hope to present a more advanced analysis with a time span of years. 

(d) Related to the above point: I find it very interesting that in 3 of the 4 bats for which you show exemplary movement data (Fig. 3, panels B and C), they appear to travel to the farthest distances and cover the most ground early on, and become more "conservative" in their flight paths on later evenings. This point is not explored in the discussion, nor related to earlier measurements.

During the first months of exploration, bats will occasionally perform long exploratory flights in between bouts of shorter flights where they return to nearby familiar trees. This behavior can be seen in more detail in Harten et al Science 2020. We are currently quantifying this more carefully for another study. 

(e) Finally, my points about the possible strength of a composite measure of the three personality metrics is related to my concern about one of the conclusions, which is that innate personality does not have an effect on outdoor foraging behavior. I think the manner in which this was tested statistically is likely to bias the results against finding such a result given that personality metrics are used to predict outdoor behaviors in an individual manner (6 models in total, each examining a single comparison of predisposition to outdoor behavior), while both indoor personality metrics (Fig 1B) and outdoor behaviors appear to be correlated with each other (Table 5).

Are there other analyses you have performed that are not presented in the paper and that have led you to conclude that there is no relationship here?

We agree with the reviewer, that our findings do not exclude an effect of innate personality on foraging but only suggest no such affect for the parameter we measured. That said, we did expect to find an effect of boldness because this parameter has been shown to differentiate much between groups (Harten et al., 2018), and to correlate with other parameters of behavior. We were therefore surprised to find no significant effects, as we had anticipated observing some differences.

Following the reviewer’s previous comment we now also tested another predisposition parameter – the PC1 score and also found that it did not explain foraging. 

(f) Personality measured before and after early environmental exposure (related to point (a) above): I find it interesting that the positive correlation in boldness between baseline and post-enrichment or baseline and post-release suggests that the individuals that were the most bold remained bold (and likewise for less adventurous individuals). The correlation for activity, too, still suggests that more active individuals early in life are likely to remain very active after enrichment, even accounting for the fact that activity is confounded with age.

Perhaps you could place some emphasis on the fact that the initial variation between individuals also appears to be relatively stable over repeated trials. You might also consider measuring this directly (population variance over successive trials; relationship of population variance on indoor measures vs. outdoor measures...)

Yes – this is a main point of interest. We further emphasize that in the revised manuscript 

(g) Effect of indoor behavior following early experience on outdoor behavior: You evaluate the effect of predisposition (measured on baseline trial 1) and environmental condition on measures of outdoor activity (Table 4). I wonder if you also tried using indoor behavioral measures measured on the post-enrichment trial 3 to predict outdoor foraging behavior.

Assuming that these measures are in fact reflecting a combination of predisposition and accumulated experience, then measurements at this closer time point may tell you how the combination of innate traits and early acquired experience affect behavior in the wild.

We appreciate the reviewer’s insightful suggestion to test whether indoor behavior from post-enrichment Trial 3, reflecting both innate traits and experience, predicts outdoor foraging behavior. We conducted this analysis, but found that the boldness in Trial 3 did not significantly predict any of the outdoor activity measures.

(2) [Minor] Age/development: While the authors discuss the effect of their manipulations on behavioral measures, they do not much discuss the effect of age.

I think it would be important to include at some point a mention of the developmental stages of Rousettus, giving labels to certain age ranges, e.g. pup, juvenile, adult, and to provide more context about the stages at which bats were tested in the discussion. Presently, age is only really mentioned as an explanation for declining activity levels, but I wonder if it might also have an influence on boldness.

It would also be very elegant for figures where age is given in days, to additional label then with these stages.

All bats were juveniles during the trials (approximately 4 to 8 months old), so they could not be divided into distinct age groups. To assess the effect of age, it was included as a predictor (in days) in the GLM analysis.

(3) [Major] Effect of early experience and outdoor experience on the indoor task: In the paragraph on lines 278-285, you argue that the effect of seeing earlyenriched bats exhibit more boldness in trial 5 was likely due to post-sampling bias...

I tend to disagree with this conclusion. I actually find this result both interesting and intuitive - that bats that were exposed to an enriched environment and have had experience in the wild, show much bolder activity on a familiar indoor foraging test (i.e. outside experience has made the animals bolder than before) (Fig 1, lines 159-161, Fig. S1). I did not notice this possibility mentioned in the discussion of the results.

I also do not fully understand this argument. Could you please explain further?

We accept the reviewer's comment and updated the manuscript (lines 336346) explaining the two hypotheses more clearly and arguing that it is difficult to tell them apart with the current data.

[Minor] You also say that "this difference... can be seen in Figure 2 when examining only the bats that had remained until the last trial (Figure 2A2)." Do you mean supplementary Figure S1 A2? In fact, I am entirely unclear on what data is plotted in the supplementary Figure S1 and what differentiates the two columns of figures and the two models presented in the supplementary table. Did you plot data similar to that in Figure 2, with only bats that were present for all trials, but not show this data?

There was a mistake: what was previously referred to as 2A2 is actually S2 A2.

On the right side—only among the individuals with GPS data—the change is already evident at Baseline 2, where only the bolder individuals remain. If you have suggestions for a better analysis approach, we would be happy to hear them.

### Minor points

General points regarding figures:

For Figures 2 and 3A1-3 (as well as Fig. S1): Authors must show the raw data points over the box plots. It is very difficult to interpret the data and conclusions without being able to see the true distribution.

Done

For all figures showing grouped individual data, please annotate all panels or sets of boxplots with the number of bats whose data entered into each, as it is a little difficult to keep track of the changing sample sizes across experimental stages.

To enhance transparency, we have added individual data points to all boxplots, allowing visual estimation of sample sizes across experimental stages. While numerical annotations are not included on the figures, the exact number of bats contributing to each group is provided in the Methods section (Table 8), ensuring this information is readily accessible to readers.In response to the reviewer’s request, we have updated all relevant figures to display individual data points within each boxplot. This addition makes it easier to track changes in sample size across different experimental stages.

Unless I've missed the reason behind differences in axis labelling across the figures, it seems that trials are not always referred to consistently. E.g. Fig. 1 labels say "Trial 1 (baseline)" and fig. 2 labels say "Baseline 1 0 days." I'm not entirely sure if these correspond to exactly the same data. If so, perhaps the labels can be made uniform. I think the descriptive ones (Baseline 1, Postenrichment...) may be more helpful to the reader than providing the trial number (Trial 1, etc....).

Done

Figure 1:

Very good Fig. 1A and 1B.

For panels C1-3 & D, I think it would make it easier for the reader if the personality measure labels were placed at the top of each panel, e.g. "Boldness (entrance proportion)". The double axis labels are not only harder to read, they are also redundant, as the personality measure label repeats on both axes.

Done

Panel C1: For the first panel in this sequence, I think it would be elegant to include an annotation in the figure that indicates what the datapoints lying on either side of the dashed line means, i.e. "bolder after enrichment treatment" in the upper left corner, and "bolder before enrichment treatment" in the bottom right corner.

Panel C2: It appears as though many of the data points in this panel overlap, and it appears to me that the blue data points in particular are overlaid by the orange ones. I am guessing this happens because proportion values based on entrances to only 6 boxes end up giving a more "discrete" looking distribution. I wonder if you can find a way to allow all the data to be visible by, e.g., jittering the data slightly; if there is rounding being done to the proportions, perhaps don't round them so that minute differences will allow them to escape the overlap; or possibly split the panel by enrichment treatment.

Caption for C1-3: it may be helpful to mention the correlation line color scheme: "enriched (blue lines), the impoverished (orange lines)". The caption also says positive correlations were found for "both environments together," but this correlation line is not shown. Perhaps mention "(not shown)" or show line. Please rephrase the sentence "Dashed line represents the Y=X line." for more transparency and clarity. I understand you mean an "equality" or "unity" line, but perhaps you can explicitly state the information that this line provides, something like e.g. "Dashed line indicates equal values measured on both trials."

We added the line for a reference, the caption was corrected

Figure 3:

Panels B1-C2: I would suggest giving these panels supertitles that indicate that B panels are enriched, C panels are impoverished, and that each panel is data from a different individual.

The legend was corrected to be more clear about the figure

General points regarding tables:

Please revisit tables for formatting and typos, particularly in Table 4. Please also revise table captions for clarity. E.g. "first exploration as predisposition" to "Exploration (Baseline 1)" or similar

Done

Supplementary Tables and Figure: these are missing captions and explanations.

The missing parts were adddad and corrected

Points of clarification/style:

It would seem to me more logical to present the results shown in Table 3 before those in Table 2, given that the primary in-lab manipulation is discussed with relation to Table 3, and the analysis in Table 2 is discussed rather as a limitation (though I believe this result can be expanded upon further, see above).

For the activity metric, I would suggest showing this data as actions/hour instead of actions/minute. I think it is much more intuitive to consider, for example, that a bat makes 2 actions every hour, than that it makes 0.002 actions per minute.

Done

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

The authors present a model for multisensory correlation detection that is based on the neurobiologically plausible Hassenstein Reichardt detector. It modifies their previously reported model (Parise & Ernst, 2016) in two ways: a bandpass (rather than lowpass) filter is initially applied and the filtered signals are then squared. The study shows that this model can account for synchrony judgement, temporal order judgement, etc in two new data sets (acquired in this study) and a range of previous data sets.

Strengths:

(1) The model goes beyond descriptive models such as cumulative Gaussians for TOJ and differences in cumulative Gaussians for SJ tasks by providing a mechanism that builds on the neurobiologically plausible Hassenstein-Reichardt detector.

(2) This modified model can account for results from two new experiments that focus on the detection of correlated transients and frequency doubling. The model also accounts for several behavioural results from experiments including stochastic sequences of A/V events and sine wave modulations.

Additional thoughts:

(1) The model introduces two changes: bandpass filtering and squaring of the inputs. The authors emphasize that these changes allow the model to focus selectively on transient rather than sustained channels. But shouldn't the two changes be introduced separately? Transients may also be detected for signed signals.

We updated the original model because our new psychophysical evidence demonstrates the fundamental role of unsigned transient for multisensory perception. While the original model received input from sustained unimodal channels (low-pass filters), the new version receives input from unsigned unimodal transient channels. Transient channels are normally modelled through bandpass filters (to remove the DC and high-frequency signal components) and squaring (to remove the sign). While these may appear as two separate changes in the model, they are, in fact, a single one: the substitution of sustained with unsigned transient channels (for a similar approach, see Stigliani et al. 2017, PNAS). Either change alone would not be sufficient to implement a transient channel that accounts for the present results.

That said, we were also concerned with introducing too many changes in the model at once. Indeed, we simply modelled the unimodal transient channels as a single band-pass filter followed by squaring. This is already a stripped-down version of the unsigned transient detectors proposed by Adelson and Bergen in their classic Motion Energy model. The original model consisted of two biphasic temporal filters 90 degrees out of phase (i.e., quadrature filters), whose output is later combined. While a simpler implementation of the transient channels was sufficient in the present study, the full model may be necessary for other classes of stimuli (including speech, Parise, 2024, BiorXiv). Therefore, for completeness, we now include in the Supplementary Information a formal description of the full model, and validate it by simulating our two novel psychophysical studies. See Supplementary Information “The quadrature MCD model” section and Supplementary Figure S8.

(2) Because the model is applied only to rather simple artificial signals, it remains unclear to what extent it can account for AV correlation detection for naturalistic signals. In particular, speech appears to rely on correlation detection of signed signals. Can this modified model account for SJ or TOJ judgments for naturalistic signals?

It can. In a recent series of studies we have demonstrated that a population of spatially-tuned MCD units can account for audiovisual correlation detection for naturalistic stimuli, including speech (e.g. the McGurk Illusion). Once again, unsigned transients were sufficient to replicate a variety of previous findings. We have now extended the discussion to cover this recent research: Parise, C. V. (2024). Spatiotemporal models for multisensory integration. bioRxiv, 2023-12.

Even Nidiffer et al. (2018) which is explicitly modelled by the authors report a significant difference in performance for correlated and anti-correlated signals. This seems to disagree with the results of study 1 reported in the current paper and the model's predictions. How can these contradicting results be explained? If the brain detects correlation on signed and unsigned signals, is a more complex mechanism needed to arbitrate between those two?

We believe the reviewer here refers to our Experiment 2 (where, like Nidiffer at al. (2018) we used periodic stimuli, not Experiment 1, which consists of step stimuli). We were also puzzled by the difference between our Experiment 2 and Nidiffer et al. (2018): we induced frequency doubling, Nidiffer did not. Based on quantitative simulations, we concluded that this difference could be attributed to the fact that while Nidiffer included on each trial an intensity ramp in their periodic audiovisual stimuli, we did not. As a result, when considering the ramp (unlike in Nidiffer’s analyses), all audiovisual signals used by Nidiffer were positively correlated (irrespective of frequency and phase offset), while our signals in Experiment 2 were sometimes correlated and other times not (depending on the phase offset). This important simulation is included in Supplementary Figure S7; we also have now updated the text to better highlight the role of the pedestal in determining the direction of the correlation.

(3) The number of parameters seems quite comparable for the authors' model and descriptive models (e.g. PSF models). This is because time constants require refitting (at least for some experimental data sets) and the correlation values need to be passed through a response mode (i.e. probit function) to account for behavioural data. It remains unclear how the brain adjusts the time constants to different sensory signals.

This is a deep question. For simplicity, here the temporal constants were fitted to the empirical psychometric functions. To avoid overfitting, whenever possible we fitted such parameters over some training datasets, while trying to predict others. However, in some cases, it was necessary to fit the temporal constants to specific datasets. This may suggest that the temporal tuning of those units is not crystalised to some pre-defined values, but is adjusted based on recent perceptual history (e.g., the sequence of trials and stimuli participants are exposed to during the various experiments).

For transparency, here we show how varying the tuning of the temporal constants of the filters affects the goodness of fit of our new psychophysical experiments (Supplementary Figure S8). As it can be readily appreciated, the relative temporal tuning of the unimodal transient detector was critical, though their absolute values could vary over a range of about 15 to over 100ms. The tuning of the low-pass filters of the correlation detector (not shown here) displayed much lower temporal sensitivity over a range between 0.1s to over 1s.

This simulation shows the impact of temporal tuning in our simulations, however, the question remains as to how such a tuning gets selected in the first place. An appealing explanation relies on natural scene statistics: units are temporally tuned to the most common audiovisual stimuli. Although our current empirical evidence does not allow us to quantitatively address this question, in previous simulations (see Parise & Ernst, 2016, Supplementary Figure 8), by analogy with visual motion adaptation, we show how the temporal constants of our model can dynamically adjust and adapt to recent perceptual history. We hope these new and previous simulations address the question about the nature of the temporal tuning of the MCD units.

(4) Fujisaki and Nishida (2005, 2006) proposed mechanisms for AV correlation detection based on the Hassenstein-Reichardt motion detector (though not formalized as a computational model).

This is correct, Fujisaki and Nishida (2005, 2007) also hypothesized that AV synchrony could be detected using a mechanism analogous to motion detection. Interestingly, however, they ruled out such a hypothesis, as their “data do not support the existence of specialized low-level audio-visual synchrony detectors”. Yet, along with our previous work (Parise & Ernst, 2016, where we explicitly modelled the experiments of Fujisaki and Nishida), the present simulations quantitatively demonstrate that a low-level AV synchrony detector is instead sufficient to account for audiovisual synchrony perception and correlation detection. We now credit Fujusaki and Nishida in the modelling section for proposing that AV synchrony can be detected by a cross-correlator.

Finally, we believe the reviewer is referring to the 2005 and 2007 studies of Fujisaki and Nishida (not 2006); here are the full references of the two articles we are referring to:

Fujisaki, W., & Nishida, S. Y. (2005). Temporal frequency characteristics of synchrony–asynchrony discrimination of audio-visual signals. Experimental Brain Research, 166, 455-464.

Fujisaki, W., & Nishida, S. Y. (2007). Feature-based processing of audio-visual synchrony perception revealed by random pulse trains. Vision Research, 47(8), 1075-1093.

Reviewer #2 (Public Review):

Summary:

This is an interesting and well-written manuscript that seeks to detail the performance of two human psychophysical experiments designed to look at the relative contributions of transient and sustained components of a multisensory (i.e., audiovisual) stimulus to their integration. The work is framed within the context of a model previously developed by the authors and is now somewhat revised to better incorporate the experimental findings. The major takeaway from the paper is that transient signals carry the vast majority of the information related to the integration of auditory and visual cues, and that the Multisensory Correlation Detector (MCD) model not only captures the results of the current study but is also highly effective in capturing the results of prior studies focused on temporal and causal judgments.

Strengths:

Overall the experimental design is sound and the analyses are well performed. The extension of the MCD model to better capture transients makes a great deal of sense in the current context, and it is very nice to see the model applied to a variety of previous studies.

Weaknesses:

My one major issue with the paper revolves around its significance. In the context of a temporal task(s), is it in any way surprising that the important information is carried by stimulus transients? Stated a bit differently, isn't all of the important information needed to solve the task embedded in the temporal dimension? I think the authors need to better address this issue to punch up the significance of their work.

In hindsight, it may appear unsurprising that transient signals carry most information for audiovisual integration. Yet, so somewhat unexpectedly, this has never been investigated using perhaps the most diagnostic psychophysical tools for perceived crossmodal timing; namely temporal order and simultaneity judgments–along with carefully designed experiments with quantitative predictions for the effect of either channel. The fact that the results conform to intuitive expectations further supports the value of the present work: grounding empirically with what is intuitively expected. This offers solid psychophysical evidence that one can build on for future advancements. Importantly, developing a model that builds on our new results and uses the same parameters to predict a variety of classic experiments in the field, further supports the current approach.

If “significance” is intended as shaking previous intuitions or theories, then no: this is not a significant contribution. If instead, by significance we intend to build a solid empirical and theoretical ground for future work, then we believe this study is not significant, it is foundational. We hope that this work's significance is better captured in our discussion.

On a side note, there is an intriguing factor around transient vs. sustained channels: what matters is the amount of change, not the absolute stimulus intensity. Previous studies, for example, have suggested a positive cross modal mapping between auditory loudness and visual lightness or brightness [Odegaard et al., 2004]. This study, conversely, challenges this view and demonstrates that what matters for multisensory integration in time is not the intensity of a stimulus, but changes thereof.

In a more minor comment, I think there also needs to be a bit more effort into articulating the biological plausibility/potential instantiations of this sustained versus transient dichotomy. As written, the paper suggests that these are different "channels" in sensory systems, when in reality many neurons (and neural circuits) carry both on the same lines.

The reviewer is right, in our original manuscript we glossed over this aspect. We have now expanded the introduction to discuss their anatomical basis. However, we are not assuming any strict dichotomy between transient and sustained channels; rather, our results and simulations demonstrate that transient information is sufficient to account for audiovisual temporal integration.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) Related to point 2 of the public review, can the authors provide additional results showing that the model can also account for naturalistic signals and more complex stochastic signals?

While working on this manuscript, we were also working in parallel on a project related to audiovisual integration of naturalistic signals. A pre-print is available online [Parise, 2024, BiorXiv], and the related study is now discussed in the conclusions.

(2) As noted in the public review, Fujisaki and Nishida (2005, 2006) already proposed mechanisms for AV correlation detection based on the Hassenstein-Reichardt motion detector. Their work should be referenced and discussed.

We have now acknowledged the contribution of Fujisaki and Nishida in the modelling section, when we first introduce the link between our model and the Hassenstein-Reichardt detectors.

(3) Experimental parameters: Was the phase shift manipulated in blocks? If yes, what about temporal recalibration?

To minimise the effect of temporal recalibration, the order of trials in our experiments was randomised. Nonetheless, we can directly assess potential short-term recalibration effects by plotting our psychophysical responses against both the current SOA, and that of the previous trials. The resulting (raw) psychometric surfaces below are averaged across observers (and conditions for Experiment 1). In all our experiments, responses are obviously dependent on the current SOA (x-axis). However, the SOA of the previous trials (y-axis) does not seem to meaningfully affect simultaneity and temporal order judgments. The psychometric curves above the heatmaps represent the average psychometric functions (marginalized over the SOA of the previous trial).

All in all, the present analyses demonstrate negligible temporal recalibration across trials, likely induced by a random sequence of lags or phase shifts. Therefore, when estimating the temporal constants of the model, it seems reasonable to ignore the potential effects of temporal recalibration. To avoid increasing the complexity of the present manuscript, we would prefer not to include the present analyses in the revised version.

Author response image 1.

Effect of previous trial. Psychometric surfaces for Experiments 1 and 2 plotted against the lag in the current vs. the previous trial. While psychophysical responses are strongly modulated by the lag in the last trial (horizontal axis), they are relatively unaffected by the lag in the previous trial (vertical axis).

(4) The model predicts no differences for experiment 1 and this is what is empirically observed. Can the authors support these null results with Bayes factors?

This is a good suggestion: we have now included a Bayesian repeated measures ANOVA to the analyses of Experiment 1. As expected, these analyses provide further, though mild evidence in support for the null hypothesis (See Table S2). For completeness, the new Bayesian analyses are presented alongside the previous frequentist ones in the revised manuscript.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors aim to consider the effects of phonotactics on the effectiveness of memory reactivation during sleep. They have created artificial words that are either typical or atypical and showed that reactivation improves memory for the latter but not the former.

Comment 1:

Strengths:

This is an interesting design and a creative way of manipulating memory strength and typicality. In addition, the spectral analysis on both the wakefulness data and the sleep data is well done. The article is clearly written and provides a relevant and comprehensive of the literature and of how the results contribute to it.

We thank the reviewer for his/her positive evaluation of our manuscript. 

Comment 2:

Weaknesses:

(1) Unlike most research involving artificial language or language in general, the task engaged in this manuscript did not require (or test) learning of meaning or translation. Instead, the artificial words were arbitrarily categorised and memory was tested for that categorisation. This somewhat limits the interpretation of the results as they pertain to language science, and qualifies comparisons with other language-related sleep studies that the manuscript builds on.

We thank the reviewer for this comment. We agree that we did not test for meaning or translation but used a categorization task in which we trained subjects to discriminate artificial words according to their reward associations (rewarded vs. non-rewarded). Previous language studies (Batterink et al., 2014; Batterink and Paller, 2017; Reber, 1967) used artificial words to investigate implicit learning of hidden grammar rules. Here, the language researchers studied generalization of the previously learned grammar knowledge by testing subject’s ability to categorize correctly a novel set of artificial words into rule-congruent versus rule-incongruent words. These differences to our study design might limit the comparability between the results of previous language studies of artificial grammar learning and our findings. We discussed now this aspect as a limitation of our novel paradigm. 

We added the following sentences to the discussion on p.14, ll. 481-488:

Based on our paradigm, we investigated categorization learning of artificial words according to their reward associations (rewarded vs. unrewarded) and did not studied aspects of generalization learning of artificial grammar rules (Batterink et al., 2014; Batterink and Paller, 2017; Reber, 1967). This difference might limit the comparability between these previous language-related studies and our findings. However, the usage of artificial words with distinct phonotactical properties provided a successful way to manipulate learning difficulty and to investigate word properties on TMR, whereas our reward categorization learning paradigm had the advantage to increase the relevance of the word learnings due to incentives.    

Comment 3:

(2) The details of the behavioural task are hard to understand as described in the manuscript. Specifically, I wasn't able to understand when words were to be responded to with the left or right button. What were the instructions? Were half of the words randomly paired with left and half with right and then half of each rewarded and half unrewarded? Or was the task to know if a word was rewarded or not and right/left responses reflected the participants' guesses as to the reward (yes/no)? Please explain this fully in the methods, but also briefly in the caption to Figure 1 (e.g., panel C) and in the Results section.

We thank the reviewer for this comment and added additional sentences into the document to provide additional explanations. We instructed the participants to respond to each word by left- and right-hand button presses, whereas one button means the word is rewarded and the other button means the word is unrewarded. The assignment of left- and right-hand button presses to their meanings (rewarded versus unrewarded) differed across subjects. In the beginning, they had to guess. Then over trial repetitions with feedback at the end of each trial, they learned to respond correctly according to the rewarded/unrewarded associations of the words.        

We added the following sentences to the results section on p.5, ll. 161-168: 

As a two alternative forced-choice task, we assigned left- and right-hand button presses to the rewarded and the unrewarded word category, counterbalanced across subjects. We instructed the participants to respond to each word by left- or right-hand button presses, whereas one button means the word is rewarded (gain of money points) and the other button means the word is unrewarded (avoid the loss of money points). In the beginning, they had to guess. By three presentations of each word in randomized order and by feedback at the end of each trial, they learned to respond correctly according to the rewarded/unrewarded associations of the words (Fig. 1c). 

We added the following sentences to the caption of Figure 1 on p.6, ll. 188-194:

As a two alternative forced-choice task, responses of left- and right-hand button presses were assigned to the rewarded and the unrewarded word category, respectively. The participants were instructed to respond to each word by left- or right-hand button presses, whereas one button means the word is rewarded (gain of money points) and the other button means the word is unrewarded (avoid the loss of money points). d) Feedback matrix with the four answer types (hits: rewarded and correct; CR, correct rejections: unrewarded and correct; misses: rewarded and incorrect; FA, false alarms: unrewarded and incorrect) regarding to response and reward assignment of the word.

We added the following sentences to the methods on p.19, ll. 687-692:  

As a two alternative forced-choice task, we assigned left- and right-hand button presses to the rewarded and the unrewarded word category, counterbalanced across subjects. We instructed the participants to respond to each word by left- or right-hand button presses, whereas one button means the word is rewarded (gain of money points) and the other button means the word is unrewarded (avoid the loss of money points).

Comment 4:  

(3) Relatedly, it is unclear how reward or lack thereof would translate cleanly into a categorisation of hits/misses/correct rejections/false alarms, as explained in the text and shown in Figure 1D. If the item was of the non-rewarded class and the participant got it correct, they avoided loss. Why would that be considered a correct rejection, as the text suggests? It is no less of a hit than the rewarded-correct, it's just the trial was set up in a way that limits gains. This seems to mix together signal detection nomenclature (in which reward is uniform and there are two options, one of which is correct and one isn't) and loss-aversion types of studies (in which reward is different for two types of stimuli, but for each type you can have H/M/CR/FA separably). Again, it might all stem from me not understanding the task, but at the very least this required extended explanations. Once the authors address this, they should also update Fig 1D. This complexity makes the results relatively hard to interpret and the merit of the manuscript hard to access. Unless there are strong hypotheses about reward's impact on memory (which, as far as I can see, are not at the core of the paper), there should be no difference in the manner in which the currently labelled "hits" and "CR" are deemed - both are correct memories. Treating them differently may have implications on the d', which is the main memory measure in the paper, and possibly on measures of decision bias that are used as well.

We thank the reviewer for this comment giving us the opportunity to clarify. As explained in the previous comment, for our two alternative forced-choice task, we instructed the participants to press one button when they were thinking the presented word is rewarded and the other button, when they were thinking the word is unrewarded. Based on this instruction, we applied the signal detection theory (SDT), because the subjects had the task to detect when reward was present or to reject when reward was absent. Therefore, we considered correct responses of words of the rewarded category as hits and words of the unrewarded category as correct rejections (see Table below). However, the reviewer is correct because in addition to false alarms, we punished here the incorrect responses by subtraction of money points to control for alternative task strategies of the participants instead of reward association learning of words. We agree that further explanation/argumentation to introduce our nomenclature is necessary.  

Author response table 1.

We adjusted the results section on p.5, ll. 169-177:

To obtain a measurement of discrimination memory with respect to the potential influence of the response bias, we applied the signal detection theory (Green and Swets, 1966). Because, we instructed the participants to respond to each word by left- or right-hand button presses and that one button means reward is present whereas the other button means reward is absent, we considered correct responses of words of the rewarded category as hits and words of the unrewarded category as correct rejections. Accordingly, we assigned the responses with regard to the reward associations of the words to the following four response types: hits (rewarded, correct); correct rejections (unrewarded, correct); misses (rewarded, incorrect); and false alarms (unrewarded, incorrect). Dependent on responses, subjects received money points (Fig. 1d). 

Comment 5:

(4) The study starts off with a sample size of N=39 but excludes 17 participants for some crucial analyses. This is a high number, and it's not entirely clear from the text whether exclusion criteria were pre-registered or decided upon before looking at the data. Having said that, some criteria seem very reasonable (e.g., excluding participants who were not fully exposed to words during sleep). It would still be helpful to see that the trend remains when including all participants who had sufficient exposure during sleep. Also, please carefully mention for each analysis what the N was.

Our study was not pre-registered. Including all the subjects independent of low prememory performance, but with respect to a decent number of reactivations (> 160 reactivations, every word at least 2 times), resulted in a new dataset with 15 and 13 participants of the high- and low-PP cueing condition, respectively. Here, statistical analyses revealed no significant overnight change anymore in memory performance in the high-PP cueing condition (Δ memory (d'): t(14) = 1.67, p = 0.12), whereas the increase of the bias in decision making towards risk avoidance still remained significant (Δ bias (c-criterion): t(14) = 3.36, p = 0.005).

We modified and added the following sentences to the discussion on p.13, ll. 456-458:

Our study has limitations due to a small sample size and between-subject comparisons. The criteria of data analyses were not pre-registered and the p-values of our behavior analyses were not corrected for multiple comparisons.

Comment 6:             

(5) Relatedly, the final N is low for a between-subjects study (N=11 per group). This is adequately mentioned as a limitation, but since it does qualify the results, it seemed important to mention it in the public review.

We agree with the reviewer that the small sample size and the between subject comparisons represent major limitations of our study. Accordingly, we now discussed these limitations in more detail by adding alternative explanations and further suggestions for future research to overcome these limitations.        

We added the following sentences to the discussion about the limitations on p.14, ll. 465-488: 

To control for potential confounders despite the influence of difficulty in word learning on TMR, we compared parameters of sleep, the pre-sleep memory performance and the vigilance shortly before the post-sleep memory test, revealing no significant group differences (see Table S1 and S2). Nevertheless, we cannot rule out that other individual trait factors differed between the groups, such as the individual susceptibility to TMR. To rule out these alternative explanations based on individual factors, we suggest for future research to replicate our study by conducting a within-subject design with cueing of subsets of previously learned low- and high-PP words providing all conditions within the same individuals as shown in other TMR studies (Cairney et al., 2018; Schreiner and Rasch, 2015).

Comment 7:

(6) The linguistic statistics used for establishing the artificial words are all based on American English, and are therefore in misalignment with the spoken language of the participants (which was German). The authors should address this limitation and discuss possible differences between the languages. Also, if the authors checked whether participants were fluent in English they should report these results and possibly consider them in their analyses. In all fairness, the behavioural effects presented in Figure 2A are convincing, providing a valuable manipulation test.

We thank the reviewer pointing to the misalignment between the German-speaking participants and the used artificial words based on American English. Further, we did not assessed the English language capability of the participants to control it as a potential confounder, whereas comparative control analyses revealed no significant differences between the both cueing groups in pre-sleep memory performance (see Table S1). 

We now discussed these comments as limitations on p.14, ll. 473-481: 

Further, we used artificial words based on American English in combination with German speaking participants, whereas language differences of pronunciation and phoneme structures might affect word perception and memory processing (Bohn and Best, 2012). On the other hand, both languages are considered to have the same language family (Eberhard et al., 2019) and the phonological distance between English and German is quite short compared for example to Korean (Luef and Resnik, 2023). Thus, major common phonological characteristics across both languages are still preserved. In addition, our behavior analyses revealed robust word discrimination learning and distinct memory performance according to different levels of phonotactic probabilities providing evidence of successful experimental manipulation. 

Comment 8:

(7) With regard to the higher probability of nested spindles for the high- vs low-PP cueing conditions, the authors should try and explore whether what the results show is a general increase for spindles altogether (as has been reported in the past to be correlated with TMR benefit and sleep more generally) or a specific increase in nested spindles (with no significant change in the absolute numbers of post-cue spindles). In both cases, the results would be interesting, but differentiating the two is necessary in order to make the claim that nesting is what increased rather than spindle density altogether, regardless of the SW phase.

We conducted additional analyses based on detected sleep spindles to provide additional data according to this question. 

We added the following section to the supplementary data on pp. 31-32, ll. 1007-1045:  

After conducting a sleep spindle detection (frequency range of 12-16Hz, see methods for details), we compared the sleep spindle density between the TMR conditions of high- and lowPP showing no significant difference (see Fig. S8a and Table S9). Next, we subdivided the detected sleep spindles into coupled and uncoupled sleep spindles with the previously detected slow waves (SW; analyses of Fig. 4). Sleep spindles were defined as coupled when their amplitude peak occurred during the SW up-state phase (0.3 to 0.8s time-locked to the SW troughs). A two-way mixed design ANOVA on the amplitude size of the sleep spindles with the cueing group as a between-subject factor (high-PP-cued vs. low-PP-cued) and SW-coupling as a within-subject factor (coupled vs. uncoupled) showed a significant interaction effect (cueing group × SW-coupling: F(1,20) = 4.51, p = 0.046, η2 = 0.18), a significant main effect of SW-coupling (F(1,20) = 85.02, p < 0.001, η2 = 0.81), and a trend of significance of the main effect of the cueing group (F(1,20) = 3.54, p = 0.08). Post-hoc unpaired t-tests revealed a significant higher amplitude size of the coupled sleep spindles of the cueing group of high- compared to low-PP (t(20) = 2.13, p = 0.046, Cohen’s d = 0.91; Fig. S8b) and no significant group difference of the uncoupled sleep spindles (t(20) = 1.62, p = 0.12). An additional comparison of the amount of coupled sleep spindles between the cueing groups revealed no significant difference (see Table S9). 

Here, we found that detected sleep spindles coupled to the SW up-state phase occurred with higher amplitude after TMR presentations of the high-PP words in comparison to the low-PP words, whereas the sleep spindle density and the amount of sleep spindles coupled to the SW up-state phase did not differed between the cueing conditions.     

We added the following sentences to the methods on pp. 22-23, ll. 822-839:  

Sleep spindle analyses 

We detected fast sleep spindles by band-pass filtering (12-16Hz) the signal of the Pz electrode during the auditory cueing trials in the time windows of -2 to 8s according to stimulus onsets. The amplitude threshold was calculated individually for each subject as 1.25 standard deviations (SDs) from the mean. The beginning and end times of the sleep spindles were then defined as the points at which the amplitude fell below 0.75 SDs before and after the detected sleep spindle. Only sleep spindles with a duration of 0.5-3 s were included in subsequent analyses. 

To compare the sleep spindle densities between the different cueing conditions of high- and low-PP, we computed the grand average sleep spindle density distribution in number per trial with a bin size of 0.5s from -0.5 to 6s time-locked to stimulus onset in each condition (see Fig. S8a and Table S9).     

Based on the detected slow waves and sleep spindles, we defined coupling events when the positive amplitude peak of a detected sleep spindle was occurring during the slow wave upstate phase in a time window of 0.3 to 0.8s according to the trough of a slow wave. 

We computed the averaged amplitude size of each detected sleep spindle by calculating the mean of the absolute amplitude values of all negative and positive peaks within a detected sleep spindle (see Fig. S8b).

We added the following sentences to the results on p.10, ll. 338-343:  

By conducting an additional analyses based on detection of fast sleep spindles (12-16Hz; see methods), we confirmed that fast sleep spindles during the SW up-states (from 0.3 to 0.8s after the SW trough) occurred with significantly higher amplitude after the cueing presentation of high- compared to low-PP words, whereas parameters of sleep spindle density and the amount sleep spindles coupled to the SW up-state did not differed between the cueing conditions (see Fig. S8 and Table S9).       

Reviewer #2 (Public Review):

Summary:

The work by Klaassen & Rasch investigates the influence of word learning difficulty on sleepassociated consolidation and reactivation. They elicited reactivation during sleep by applying targeted memory reactivation (TMR) and manipulated word learning difficulty by creating words more similar (easy) or more dissimilar (difficult) to our language. In one group of participants, they applied TMR of easy words and in another group of participants, they applied TMR of difficult words (between-subjects design). They showed that TMR leads to higher memory benefits in the easy compared to the difficult word group. On a neural level, they showed an increase in spindle power (in the up-state of an evoked response) when easy words were presented during sleep.

Comment 9:

Strengths:

The authors investigate a research question relevant to the field, that is, which experiences are actually consolidated during sleep. To address this question, they developed an innovative task and manipulated difficulty in an elegant way.

Overall, the paper is clearly structured, and results and methods are described in an understandable way. The analysis approach is solid.

We thank the reviewer for his/her positive evaluation of our manuscript.

Weaknesses:

Comment 10:

(1) Sample size

For a between-subjects design, the sample size is too small (N = 22). The main finding (also found in the title "Difficulty in artificial word learning impacts targeted memory reactivation") is based on an independent samples t-test with 11 participants/group.

The authors explicitly mention the small sample size and the between-subjects design as a limitation in their discussion. Nevertheless, making meaningful inferences based on studies with such a small sample size is difficult, if not impossible.

We agree with the reviewer that the small sample size and the between subject comparisons represent major limitations of our study. Accordingly, we now discussed these limitations in more detail by adding alternative explanations and further suggestions for future research to overcome these limitations.        

We added the following sentences to the discussion about the limitations on p.14, ll. 465-473: 

To control for potential confounders despite the influence of difficulty in word learning on TMR, we compared parameters of sleep, the pre-sleep memory performance and the vigilance shortly before the post-sleep memory test, revealing no significant group differences (see Table

S1 and S2). Nevertheless, we cannot rule out that other individual trait factors differed between the groups, such as the individual susceptibility to TMR. To rule out these alternative explanations based on individual factors, we suggest for future research to replicate our study by conducting a within-subject design with cueing of subsets of previously learned low- and high-PP words providing all conditions within the same individuals as shown in other TMR studies (Cairney et al., 2018; Schreiner and Rasch, 2015).

Comment 11:

(2) Choice of task

though the task itself is innovative, there would have been tasks better suited to address the research question. The main disadvantage the task and the operationalisation of memory performance (d') have is that single-trial performance cannot be calculated. Consequently, choosing individual items for TMR is not possible.

Additionally, TMR of low vs. high difficulty is conducted between subjects (and independently of pre-sleep memory performance) which is a consequence of the task design.

The motivation for why this task has been used is missing in the paper.

We used a reward task combined with TMR because previous studies revealed beneficial effects of reward related information on sleep dependent memory consolidation and reactivation (Asfestani et al., 2020; Fischer and Born, 2009; Lansink et al., 2009; Sterpenich et al., 2021). In addition, we wanted to increase the motivation of the participants, as they could receive additional monetary compensation according to their learning and memory task performances. Furthermore, we designed the task, with the overall possibility to translate this task to operant conditioning in rats (see research proposal: https://data.snf.ch/grants/grant/168602). However, the task turned out to be too difficult to translate to rats, whereas we developed a different learning paradigm for the animal study (Klaassen et al., 2021) of this cross-species research project.       

We added the following sentence to the introduction on p.4, ll. 134-137:

To consider the beneficial effect of reward related information on sleep dependent memory consolidation and reactivation (Asfestani et al., 2020; Fischer and Born, 2009; Lansink et al., 2009; Sterpenich et al., 2021), we trained healthy young participants to categorize these words into rewarded and unrewarded words to gain and to avoid losses of money points.  

Reviewer #3 (Public Review):

Summary:

In this study, the authors investigated the effects of targeted memory reactivation (TMR) during sleep on memory retention for artificial words with varying levels of phonotactical similarity to real words. The authors report that the high phonotactic probability (PP) words showed a more pronounced EEG alpha decrease during encoding and were more easily learned than the low PP words. Following TMR during sleep, participants who had been cued with the high PP TMR, remembered those words better than 0, whilst no such difference was found in the other conditions. Accordingly, the authors report higher EEG spindle band power during slow-wave up-states for the high PP as compared to low PP TMR trials. Overall, the authors conclude that artificial words that are easier to learn, benefit more from TMR than those which are difficult to learn.

Comment 12 & 13:

Strengths:

(1) The authors have carefully designed the artificial stimuli to investigate the effectiveness of TMR on words that are easy to learn and difficult to learn due to their levels of similarity with prior wordsound knowledge. Their approach of varying the level of phonotactic probability enables them to have better control over phonotactical familiarity than in a natural language and are thus able to disentangle which properties of word learning contribute to TMR success.

(2) The use of EEG during wakeful encoding and sleep TMR sheds new light on the neural correlates of high PP vs. low PP both during wakeful encoding and cue-induced retrieval during sleep.

We thank the reviewer for his/her positive evaluation of our manuscript.

Weaknesses:

Comment 14:

(1) The present analyses are based on a small sample and comparisons between participants. Considering that the TMR benefits are based on changes in memory categorization between participants, it could be argued that the individuals in the high PP group were more susceptible to TMR than those in the low PP group for reasons other than the phonotactic probabilities of the stimuli (e.g., these individuals might be more attentive to sounds in the environment during sleep). While the authors acknowledge the small sample size and between-subjects comparison as a limitation, a discussion of an alternative interpretation of the data is missing.

We agree with the reviewer that the small sample size and the between subject comparisons represent major limitations of our study. We thank the reviewer for this helpful comment and now discussed these limitations in more detail by adding alternative explanations and further suggestions for future research to overcome these limitations.

We added the following sentences to the discussion on p.14, ll. 465-473: 

To control for potential confounders despite the influence of difficulty in word learning on TMR, we compared parameters of sleep, the pre-sleep memory performance and the vigilance shortly before the post-sleep memory test, revealing no significant group differences (see Table S1 and S2). Nevertheless, we cannot rule out that other individual trait factors differed between the groups, such as the individual susceptibility to TMR. To rule out these alternative explanations based on individual factors, we suggest for future research to replicate our study by conducting a within-subject design with cueing of subsets of previously learned low- and high-PP words providing all conditions within the same individuals as shown in other TMR studies (Cairney et al., 2018; Schreiner and Rasch, 2015).

Comment 15:

(2) While the one-tailed comparison between the high PP condition and 0 is significant, the ANOVA comparing the four conditions (between subjects: cued/non-cued, within-subjects: high/low PP) does not show a significant effect. With a non-significant interaction, I would consider it statistically inappropriate to conduct post-hoc tests comparing the conditions against each other. Furthermore, it is unclear whether the p-values reported for the t-tests have been corrected for multiple comparisons. Thus, these findings should be interpreted with caution.

We thank the reviewer for this comment giving us the opportunity to correct our analyses and clarify with additional description. Indeed, we investigated at first overnight changes in behavior performance within the four conditions, conducting t-tests against 0 of Δ-values of d' and c-criterion. Whereas for all our statistical analyses the p-value was set at p < 0.05 for two-tailed testing, we did not corrected the p-value of our behavior analyses for multiple comparisons. To investigate subsequently differences between conditions, we conducted additional ANOVAs. We agree with the reviewer that without significant of results of the ANOVA, post-hoc analyses should not be conducted. Taken in account as well the recommendation of reviewer 1, we included now only post-hoc pairwise comparisons when the interaction effect of the ANOVA revealed at least a trend of significance (p < 0.1). 

We removed the following post-hoc analyses from the results section on p.9, ll. 291-295: 

Additional post-hoc pairwise comparisons revealed a significant difference between the highPP cued and low-PP uncued (high-PP cued vs. low-PP uncued: t(10) = 2.43, p = 0.04), and no difference to other conditions (high-PP cued vs.: high-PP uncued t(20) = 1.28, p = 0.22; lowPP cued t(20) = 1.57, p = 0.13).  

Further, we mentioned the lack of correction for multiple comparisons as a limitation of our results in the discussion on p.13, ll. 456-458:  

The criteria of data analyses were not pre-registered and the p-values of our behavior analyses were not corrected for multiple comparisons.

We added the following sentences to the methods p.23, ll. 842-849:

To analyze overnight changes of sleep behavioral data within TMR conditions, we conducted at first dependent sample t-tests against 0 of Δ-values (post-sleep test minus pre-sleep test) of d' and c-criterion (see Fig. 3). Two-way mixed design ANOVAs were computed to compare Δvalues between TMR conditions. After confirming at least a trend of significance (p < 0.1) for the interaction effect, we conducted post-hoc pairwise comparisons by independent and dependent sample t-tests. For all behavior statistical analyses, the p-value was set at p < 0.05 for two-tailed testing. A p-value < 0.1 and > 0.05 was reported as a trend of significance.

Comment 16:

(3) With the assumption that the artificial words in the study have different levels of phonotactic similarity to prior word-sound knowledge, it was surprising to find that the phonotactic probabilities were calculated based on an American English lexicon whilst the participants were German speakers. While it may be the case that the between-language lexicons overlap, it would be reassuring to see some evidence of this, as the level of phonotactic probability is a key manipulation in the study.

We thank the reviewer pointing to the misalignment between the German-speaking participants and the used artificial words based on American English. In line with this recommendation, we added a more outlined argumentation to the manuscript about the assumption of our study that major common phonetic characteristics across both languages are still preserved.       

We now discussed these aspects on p.14, ll. 473-481:

Further, we used artificial words based on American English in combination with German speaking participants, whereas language differences of pronunciation and phoneme structures might affect word perception and memory processing (Bohn and Best, 2012). On the other hand, both languages are considered to have the same language family (Eberhard et al., 2019) and the phonological distance between English and German is quite short compared for example to Korean (Luef and Resnik, 2023). Thus, major common phonological characteristics across both languages are still preserved. In addition, our behavior analyses revealed robust word discrimination learning and distinct memory performance according to different levels of phonotactic probabilities providing evidence of successful experimental manipulation. 

Comment 17:

(4) Another manipulation in the study is that participants learn whether the words are linked to a monetary reward or not, however, the rationale for this manipulation is unclear. For instance, it is unclear whether the authors expect the reward to interact with the TMR effects.

We used a reward task combined with TMR because previous studies revealed beneficial effects of reward related information on sleep dependent memory consolidation and reactivation (Asfestani et al., 2020; Fischer and Born, 2009; Lansink et al., 2009; Sterpenich et al., 2021). In addition, we wanted to increase the motivation of the participants, as they could receive additional monetary compensation according to their learning and memory task performances. Furthermore, we designed the task, with the overall possibility to translate this task to operant conditioning in rats (see research proposal: https://data.snf.ch/grants/grant/168602). However, the task turned out to be too difficult to translate to rats, whereas we developed a different learning paradigm for the animal study (Klaassen et al., 2021) of this cross-species research project.       

We added the following sentence to the introduction on p.4, ll. 134-137:

To consider the beneficial effect of reward related information on sleep dependent memory consolidation and reactivation (Asfestani et al., 2020; Fischer and Born, 2009; Lansink et al., 2009; Sterpenich et al., 2021), we trained healthy young participants to categorize these words into rewarded and unrewarded words to gain and to avoid losses of money points.  

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Comment 18:

(1) Please clearly define all linguistics terms - and most importantly the term "phonotactics" - at first use.

We thank the reviewer for this recommendation and we added the definition of phonotactics and further reduced the diversity of linguistic terms to improve readability. 

We added the following sentences to the beginning of the introduction on p.3, ll. 72-76:

One critical characteristic of similarity to pre-existing knowledge in auditory word processing is its speech sound (phoneme) pattern. In phonology as the field of language specific phoneme structures, phonotactics determines the constraints of word phoneme composition of a specific language.

Comment 19:

(2) Some critical details about the methods should be included in the Results section to make it comprehensible. For example, the way the crucial differences between G1-4 words should be addressed in the Results, not only in Figure 1.

According to the recommendation, we added this information to the results section.  We added the following sentences to the results section on p.4, ll. 145-154:

To study the impact of difficulty in word learning on TMR, we developed a novel learning paradigm. We formed four sets of artificial words (40 words per set; see Table S3 and S4) consisting of different sequences of two vowels and two consonants. Here, we subdivided the alphabet into two groups of consonants (C1: b, c, d, f, g, h, j, k, l, m; C2: n, p, q, r, s, t, v, w, x, z) and vowels (V1: a, e, I; V2: o, u, y). Four-letter-words were created by selecting letters from the vowel and consonant groups according to four different sequences (G1:C1, V1, V2, C2; G2: C1, V1, C2, V2; G3: V1, C1, C2, V2; G4: V1, C1, V2, C2; Fig. 1a; see methods for further details). Comparison analyses between the sets revealed significant differences in phonotactic probability (PP; Fig. 1b; unpaired t-tests: G1 / G2 > G3 / G4, p < 0.005, values of Cohen’s d > 0.71).

Comment 20

(3) Was scoring done both online and then verified offline? If so, please note that.

We included now this information.  

We adjusted the method section on p.21, ll. 765-769:   

The sleep stages of NREM 1 to 3 (N1 to N3), wake, and REM sleep were scored offline and manually according to the criteria of the American Academy of Sleep Medicine (AASM) by visual inspection of the signals of the frontal, central, and occipital electrodes over 30s epochs (Iber et al., 2007). Based on offline scoring, we confirmed TMR exposure during N2 and N3 and no significant differences (p-values > 0.05) of sleep parameters between the cueing groups (see Table S2).  

Comment 21:

(4) In Figure 2, please arrange the panel letters in an easier-to-read way (e.g., label upper right panel b with a different letter).

Now we rearranged the panel letters according to the recommendation.

We adjusted Figure 2 on p.8, ll. 242-258:     

Comment 22

(5) In the first paragraph on TMR effects, please note which memory measure you are comparing (i.e., d').

We added this information according to the recommendation.  

We adjusted the sentence of the results on p.8, ll. 260-263:

To examine whether TMR during sleep impacts memory consolidation of discrimination learning with respect to learning difficulty, we calculated the overnight changes by subtracting the pre- from the post-sleep memory performance based on d'-values of the reactivated sequences (cued) and non-reactivated sequences (uncued).

Comment 23:

(6) Please show the pre-sleep and post-sleep test scores for both word categories (not only the delta). It may be best to show this as another data point in Fig 2a, but it may be helpful to also see this split between cued and uncued.

We added the pre-sleep and post-sleep test scores with the individual data points as an additional figure. 

We added the following figure to the supplementary data on p.28, ll. 936-940:  

Comment 24:

(7) In the sentence "An additional two-way mixed design ANOVA on the same values with cueing as a between-subject factor (cued vs. uncued) ...", a more exact phrasing for the last parentheses would probably be "(high-PP-Cued vs Low-PP-Cued)". Both groups were cued.

We thank the reviewer pointing this out. According to the recommendation, we corrected the descriptions of the two-way mixed design ANOVAs. In addition, we detected a mistake of wrong assignments of the conditions to ANOVAs and corrected the reported values.   

We adjusted the sentences and corrected the values on p.9, ll. 271-275 and ll. 289-291: 

An additional two-way mixed design ANOVA on the same values with the factor cueing (cued vs. uncued) as a within-subject factor and group as a between-subject factor revealed trends of significance (p < 0.1) for the interaction (cueing × group: F(1,20) = 3.47, p = 0.08) and the main effect of group (F(1,20) = 3.28, p = 0.09). The main effect of cueing was not significant (F(1,20) = 0.58, p = 0.46).

An ANOVA on c-criterion changes showed no significant effects (interaction cueing × group: F(1,20) = 2.66, p = 0.12; main effect cueing  F(1,20) = 2.08, p = 0.17; main effect group F(1,20) = 0.38, p = 0.55).

Comment 25:

(8) In the same ANOVA, please mention that there is a trend toward an interaction effect. If there wasn't one, the post-hoc comparison would be unwarranted. Please consider noting other p<0.1 pvalues as a trend as well, for consistency.

Regarding this recommendation, we included now only post-hoc pairwise comparisons after confirming at least a trend toward an interaction effect of these ANOVAs and reported consistently a p-value < 0.1 and > 0.05 as a trend of significance.

We added the following sentences to the methods p.23, ll. 844-849:

Two-way mixed design ANOVAs were computed to compare Δ-values between TMR conditions. After confirming at least a trend of significance (p < 0.1) for the interaction effect, we conducted post-hoc pairwise comparisons by independent and dependent sample t-tests. For all behavior statistical analyses, the p-value was set at p < 0.05 for two-tailed testing. A p-value < 0.1 and > 0.05 was reported as a trend of significance.

We removed the following post-hoc analyses from the results section on p.9, ll. 291-295: 

Additional post-hoc pairwise comparisons revealed a significant difference between the highPP cued and low-PP uncued (high-PP cued vs. low-PP uncued: t(10) = 2.43, p = 0.04), and no difference to other conditions (high-PP cued vs.: high-PP uncued t(20) = 1.28, p = 0.22; lowPP cued t(20) = 1.57, p = 0.13).          

Comment 26:      

(9) Please consider adding an analysis correlating spindle power with memory benefit across participants. Even if it is non-significant, it is important to report given that some studies have found such a relationship.

According to this recommendation, we conducted an additional correlation analyses.

We added the following sentences to the manuscript into the results (pp. 10-11, ll. 346-349), the discussion (p.12, ll. 413-417), and the methods (p.23, ll. 864-867):   

Whereas we found a significant group difference in spindle power nested during SW up-states,   conducting further whole sample (n = 22) correlation analyses between the individual spindle power values of the significant cluster and the overnight changes of behavior measurements revealed no significant correlations (Δ d': r = 0.16, p = 0.48; Δ c-criterion: r = 0.19, p = 0.40).

In addition to our result of the significant group difference, we failed to find significant correlations between SW nested spindle power values and overnight changes in behavior measurements, whereas previous studies reported associations of SW and spindle activities during sleep with the integration of new memories in pre-existing knowledge networks (Tamminen et al., 2013, 2010).

By using the same extracted power values (0.3 to 0.8s; 11-14Hz; Pz, P3, P4, O2, P7) per subject, we performed whole sample (n = 22) Pearson correlation analyses between these power values and the overnight changes of behavior measurements of the cued condition (Δ d' and Δ ccriterion).

Reviewer #2 (Recommendations For The Authors):

(1) Choice of task

Comment 27:      

In general, I find your task well-designed and novel. In light of your research question, however, I wonder why you chose this task. When you outlined the research question in the introduction, I expected a task similar to Schreiner et al. (2015). For example, participants have to associate high PP words with each other and low PP words. The advantage here would be that you could test the benefits of TMR in a within-subjects design (for example, cueing half of the remembered high and half of the remembered low PP words).

Please see our previous response at comment 14.    

Comment 28:

Why did you decide to introduce a reward manipulation?

Please see our previous response at comment 11.    

Comment 29:

Why did you do the cueing on a category level (cueing all high PP or all low PP words instead of single word cueing or instead of cueing 20 reward high-PP, 20 unrewarded high-PP plus 20 reward low-PP and 20 unrewarded low-PP)? Both alternatives would have provided you the option to run your statistics within participants.

Please see our previous response at comment 14.    

Comment 30:

(2) Between-subjects design and small sample size.

Why did you decide on a between-subjects design that severely reduces your power?

Why did you just collect 22 participants with such a design? Were there any reasons for this small sample size? Honestly, I think publishing a TMR study with healthy participants and such a small sample size (11 participants for some comparisons) is not advisable.

Please see our previous response at comment 14.

Comment 31:

(3) Encoding performance.

Is d' significantly above 0 in the first repetition round? I would assume that the distinction between rewarded and non-rewarded words is just possible after the first round of feedback.

Indeed, conducting t-tests against 0 revealed significantly increased d'-values in the first repetition round (2nd presentation) in both PP conditions (high-PP: 0.85 ± 0.09, t(32) = 9.17, p < 0.001; low-PP: 0.62 ± 0.09, t(32) = 6.83, p < 0.001).  

Comment 32:

(4) Encoding response options

If you want to you could make it more explicit what exactly the response options are. I assume that one button means a word has a high reward and the other button means a word has a low reward. Making it explicit increases the understanding of the results section.

Please see our previous response at comment 3.

Comment 33:           

(5) Alpha desynchronisation.

Relative change

Why did you subtract alpha power during the 1st presentation from alpha power during 2nd and 3rd presentation? You baseline-corrected already and individually included the 1st, 2nd, and 3rd repetition in your behavioural analysis.

Based on this analysis, we aimed to examine the relative change in alpha power between PP-conditions of memory-relevant word repetitions. Therefore, to extract memory relevant changes of EEG activities, the first word presentation of naive stimulus processing could serve as a more representative baseline condition covering the time-window of interest of 0.7 to 1.9 s after the stimulus onset compared to a baseline condition before stimulus onset (-1 to -0.1s). 

To explain the rational of the analyses with the baseline condition more clearly, we added this information to the results section on p.7, ll. 222-226: 

We obtained the changes in power values by subtracting the first from the second and third presentation for the high- and low-PP condition, respectively. Here, the first word presentation of naive stimulus processing served us with a more representative baseline condition covering the time-window of interest of 0.7 to 1.9 s after the stimulus onset to examine relevant changes of encoding.  

Comment 34:

(6) Alpha desynchronisation as a neural correlate of encoding depth & difficulty?

"In addition to the behavior results, these EEG results indicate differences between PP conditions in desynchronization of alpha oscillations, as an assumed neural correlate of encoding depth. In addition to the behavior results, these EEG results indicate differences between PP conditions in desynchronization of alpha oscillations, as an assumed neural correlate of encoding depth."

Given that the low-PP words are more difficult to learn, I was expecting to see higher alpha desynchronisation in the low-PP relative to the high-PP words. Could you outline in a bit more detail how your findings fit into the literature (e.g., Simon Hanslmayr did a lot of work on this)?

I would also advise you to add citations e.g., after your sentence in the quote above ("as an assumed neural correlate of encoding depth").

We thank the reviewer for the recommendation giving us the opportunity to discuss in more detail how our results relate to previous findings. 

We added additional sentences to the discussion on p.13, ll. 441-455:    

Additional studies linked alpha desynchronization to cognitive effort and cognitive load (Proskovec et al., 2019; Zhu et al., 2021). So, one could assume to observe higher alpha desynchronization in the more difficult to learn condition of low-PP compared to high-PP. On the other hand numerous studies investigating oscillatory correlates of learning and memory showed that alpha desynchronization is associated with memory across different tasks, modalities and experimental phases of encoding and retrieval (Griffiths et al., 2016, 2021, 2019a, 2019b; Hanslmayr et al., 2009; Michelmann et al., 2016). Strikingly, Griffith and colleagues (Griffiths et al., 2019a) revealed by simultaneous EEG-fMRI recordings a negative correlation between the occurrence of patterns of stimulus-specific information detected by fMRI and cortical alpha/beta suppression. Here, the authors suggested that a decrease of alpha/beta oscillations might represent the neuronal mechanism of unmasking the task-critical signal by simultaneous suppression of task-irrelevant neuronal activities to promote information processing. Following this interpretation, we assume that over the course of learning elevated memory processing of the easier to learn stimuli is associated with enhanced information processing and thus accompanied by higher cortical alpha desynchronization in comparison of the more difficult to learn stimuli.

In addition, we added the mentioned quote on p.7, ll. 239-240:

In addition to the behavior results, these EEG results indicate differences between PP conditions in desynchronization of alpha oscillations, as an assumed neural correlate of encoding depth (Griffiths et al., 2021; Hanslmayr et al., 2009).

Comment 35:

(7) Exclusion criterion.

Why did you use a d' > 0.9 as a criterion for data inclusion?

This criterion ensured that each included subject had at least in one PP-condition a d' > 1.05 of pre-sleep memory performance, which corresponds to a general accuracy rate of 70%. 

Accordingly, we adjusted these sentences of the method section on p.19, ll. 677-680: 

Data were excluded from subjects who did not reach the minimal learning performance of d' > 1.05 during the pre-sleep memory test in at least one of the two PP conditions, whereas this threshold value corresponds to accuracy rates of 70% (n = 5). In addition, we excluded one subject who showed a negative d' in one PP condition of the pre-sleep memory test (n = 1). 

Comment 36:

(8) Coherence of wording.

When you talk about your dependent variable (d') you sometimes use sensitivity. I would stick to one term.

We replaced the word sensitivity with d'.    

(9) Criterion

Comment 37:

Why do you refer to a change in criterion (Figure 3b, axis labels) as a change in memory? Do you think the criterion says something about memory?

We corrected the axis label of Figure 3b and deleted here the word memory.

Comment 38:

Additionally, why did you analyse the effect of TMR on the criterion? Do you expect the criterion to change due to sleep-dependent memory consolidation? This section would benefit from more explanation. Personally, I am very interested in your thoughts and your hypothesis (if you had one, if not that is also fine but then, make it explicit that it was an exploratory analysis).

By conducting exploratory analyses of overnight changes of the c-criterion measurements, we aimed to examine the bias of decision-making to provide comprehensive data according to the framework of the signal detection theory. Regarding the previous literature showing mainly beneficial effects of sleep on learning and memory, we focused with our hypothesis on d' and explored additionally the c-criterion.

Despite our task design with gains/hits of +10 money points and losses/FAs of -8 (instead of -10), the subjects showed already during the pre-sleep memory task significant biases towards loss avoidance in both PP conditions (t-tests against 0: high-PP: 0.44 ± 0.07, t(21) = 5.63, p < 0.001; low-PP: 0.47 ± 0.09, t(21) = 5.51, p < 0.001). As already reported in the preprint, we found an additional significant increase of c-criterion by TMR solely for the high-PP words (see Fig. 3b). Even by integrating subjects with poor pre-sleep memory performance (high-PP-cueing group: n = 15; low-PP-cueing group: n = 13), t-tests against 0 revealed a significant increase of the high-PP cueing condition (t(14) = 3.36, p = 0.005) and no significant overnight changes in the other conditions (high-PP uncued: t(12) = 1.39, p = 0.19; low-PP cued: t(12) = 1.47, p = 0.17; low-PP uncued: t(14) = -0.20, p = 0.84). These exploratory findings on c-criterion suggest potential applications of TMR to affect decision-making biases in combination with reward learning.      

We revised the manuscript mentioning the exploratory character of the c-criterion analyses of the results on p.9, ll. 282-283 and of the discussion on p.12, ll. 400-402:  

We examined next as an exploratory analysis whether TMR conditions influence biases in decision-making.

By conducting an additional exploratory analysis, we observed a significant change of the decision bias in the cueing condition of the easy to learn words and no overnight changes in the other conditions.

Comment 39:

(10) You detected SWs in the time range of 0-6 sec post sound stimulation. How was the distribution of all detected SW down-states in this time range? (You could plot a histogram for this.)

We illustrated now the detected SWs in the time range of 0 to 6 s after stimulus onset. 

We added a histogram to the supplementary section on p.30, ll. 982-986:  

Reviewer #3 (Recommendations For The Authors):

Comment 40:

(1) In line with the weakness outlined above, I would recommend including a discussion of how the between-subject comparison and small sample size could affect the results and provide alternative interpretations.

Please see our previous response at comment 14.

Comment 41:

(2) Regarding my point about statistical comparisons, I would recommend that the authors follow best practice guidelines for post-hoc tests and multiple comparisons. In Figures 3a and b, I would also recommend removing the stars indicating significance from the post-hoc tests (if this is what they reflect). Perhaps this link will be useful: https://www.statology.org/anova-post-hoc-tests/

Please see our previous response at comment 15.    

Comment 42:

(3) Furthermore, to address any doubts about the possible phonotactic probability differences between languages, I would recommend that the authors show whether the languages overlap, the level of English fluency in the German-speaking participants, and/or another way of reassuring that this is unlikely to have affected the results.

Please see our previous response at comment 7.    

Comment 43:

(4) In the introduction, I would recommend that the authors outline a clear rationale for the reward/no reward manipulation.

Please see our previous response at comment 11.    

Comment 44:

(5) Figure 1c: Please include what response options participants had, e.g., 'rewarded/not rewarded'. This would make the type of categorization clearer to the reader.

Please see our previous response at comment 3.

Comment 45:

(6) It is unclear whether the additional ANOVA conducted on the time and frequency of the identified clusters included all channels or only the channels contributing to the cluster. Consider clarifying this in the relevant methods and results. Furthermore, I would recommend labelling this as a posthoc test as this analysis was guided by an initial peak at the data and the timings, frequencies, and channels of interest were not selected a-priori.

We thank the reviewer for this recommendation and labelled the additional repeatedmeasure ANOVA as a post-hoc test. Further, we mentioned the used channels (Pz and Cz) for this analyses.

We adjusted the results section on p.7, ll. 230-233 and the methods section on p.23, ll. 858-860:            

A post-hoc repeated-measure ANOVA on alpha power changes (merged over Pz and Cz electrodes) with PP (high vs. low) and presentations (2 to 3) as within-subjects factors revealed a main effect of PP (F(1,32) = 5.42, p = 0.03, η2 = 0.15), and a significant interaction (F(1,32)  = 7.38, p = 0.01, η2 = 0.19; Fig. 2e).

After confirming the existence of a significant cluster, we conducted an additional post-hoc repeated-measure ANOVA with averaged values of the identified time and frequency range of interest and merged over the Pz and Cz electrodes (see Fig. 2e).

Comment 46:

(7) Figure 3: To better illustrate within- vs. between-subjects comparisons and promote transparency, please add individual points and lines between the within-subjects conditions.

According to this recommendation, we changed Figure 3 to add the individual data points by lines.  

We modified Figure 3 on p.9, ll. 299-303:  

Comment 47:

(8) For the SW density time-bin analyses, please include statistics for all comparisons (i.e., through 0 s to 3 s) and say whether these were corrected for multiple comparisons.

According to this recommendation, we included now statistics for all comparisons. 

We added table S6 table to the supplementary data on p.29, l.962:     

Comment 48:

(9) Consider reporting effect sizes.

We thank the reviewer for this recommendation and we added now effect sizes of significant results. 

Comment 49:

(10) For transparency and replicability, consider including a list of the four stimulus sets including their phoneme and biphone probabilities.

We included a list of the four stimulus sets with their phoneme and biphone probabilities  

We added table S3 and table S4 to the supplementary data on pp. 26-27:       

References

Asfestani MA, Brechtmann V, Santiago J, Peter A, Born J, Feld GB. 2020. Consolidation of Reward Memory during Sleep Does Not Require Dopaminergic Activation. J Cogn Neurosci 32:1688– 1703. doi:10.1162/JOCN_A_01585

Batterink LJ, Oudiette D, Reber PJ, Paller KA. 2014. Sleep facilitates learning a new linguistic rule.

Neuropsychologia 65:169–79. doi:10.1016/j.neuropsychologia.2014.10.024

Batterink LJ, Paller KA. 2017. Sleep-based memory processing facilitates grammatical generalization: Evidence from targeted memory reactivation. Brain Lang 167:83–93. doi:10.1016/J.BANDL.2015.09.003

Bohn OS, Best CT. 2012. Native-language phonetic and phonological influences on perception of American English approximants by Danish and German listeners. J Phon 40:109–128. doi:10.1016/J.WOCN.2011.08.002

Cairney SA, Guttesen A á. V, El Marj N, Staresina BP. 2018. Memory Consolidation Is Linked to Spindle-Mediated Information Processing during Sleep. Curr Biol 28:948-954.e4. doi:10.1016/j.cub.2018.01.087

Eberhard DM, Simons GF, Fennig CD. 2019. Ethnologue: Languages of the world . SIL International. Online version: http://www.ethnologue.com.

Fischer S, Born J. 2009. Anticipated reward enhances offline learning during sleep. J Exp Psychol Learn Mem Cogn 35:1586–1593. doi:10.1037/A0017256

Green DM, Swets JA. 1966. Signal detection theory and psychophysics., Signal detection theory and psychophysics. Oxford,  England: John Wiley.

Griffiths B, Mazaheri A, Debener S, Hanslmayr S. 2016. Brain oscillations track the formation of episodic memories in the real world. Neuroimage 143:256–266. doi:10.1016/j.neuroimage.2016.09.021

Griffiths BJ, Martín-Buro MC, Staresina BP, Hanslmayr S, Staudigl T. 2021. Alpha/beta power decreases during episodic memory formation predict the magnitude of alpha/beta power decreases during subsequent retrieval. Neuropsychologia 153. doi:10.1016/j.neuropsychologia.2021.107755

Griffiths BJ, Mayhew SD, Mullinger KJ, Jorge J, Charest I, Wimber M, Hanslmayr S. 2019a. Alpha/beta power decreases track the fidelity of stimulus specific information. Elife 8. doi:10.7554/eLife.49562

Griffiths BJ, Parish G, Roux F, Michelmann S, van der Plas M, Kolibius LD, Chelvarajah R, Rollings DT, Sawlani V, Hamer H, Gollwitzer S, Kreiselmeyer G, Staresina B, Wimber M, Hanslmayr S. 2019b. Directional coupling of slow and fast hippocampal gamma with neocortical alpha/beta oscillations in human episodic memory. Proc Natl Acad Sci U S A 116:21834–21842. doi:10.1073/pnas.1914180116

Hanslmayr S, Spitzer B, Bäuml K-H. 2009. Brain oscillations dissociate between semantic and nonsemantic encoding of episodic memories. Cereb Cortex 19:1631–40. doi:10.1093/cercor/bhn197

Iber C, Ancoli‐Israel S, Chesson AL, Quan SF. 2007. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Westchester, IL: American Academy of Sleep Medicine.

Klaassen AL, Heiniger A, Sánchez PV, Harvey MA, Rainer G. 2021. Ventral pallidum regulates the default mode network, controlling transitions between internally and externally guided behavior. Proc Natl Acad Sci U S A 118:1–10. doi:10.1073/pnas.2103642118

Lansink CS, Goltstein PM, Lankelma J V., McNaughton BL, Pennartz CMA. 2009. Hippocampus leads ventral striatum in replay of place-reward information. PLoS Biol 7. doi:10.1371/JOURNAL.PBIO.1000173

Luef EM, Resnik P. 2023. Phonotactic Probabilities and Sub-syllabic Segmentation in Language

Learning. Theory Pract Second Lang Acquis 9:1–31. doi:10.31261/TAPSLA.12468

Michelmann S, Bowman H, Hanslmayr S. 2016. The Temporal Signature of Memories: Identification of a General Mechanism for Dynamic Memory Replay in Humans. PLoS Biol 14:e1002528. doi:10.1371/journal.pbio.1002528

Proskovec AL, Heinrichs-Graham E, Wilson TW. 2019. Load Modulates the Alpha and Beta Oscillatory Dynamics Serving Verbal Working Memory. Neuroimage 184:256. doi:10.1016/J.NEUROIMAGE.2018.09.022

Reber AS. 1967. Implicit learning of artificial grammars. J Verbal Learning Verbal Behav 6:855–863.

doi:10.1016/S0022-5371(67)80149-X

Schreiner T, Rasch B. 2015. Boosting vocabulary learning by verbal cueing during sleep. Cereb Cortex 25:4169–4179. doi:10.1093/cercor/bhu139

Sterpenich V, van Schie MKM, Catsiyannis M, Ramyead A, Perrig S, Yang H-D, Van De Ville D, Schwartz S. 2021. Reward biases spontaneous neural reactivation during sleep. Nat Commun 2021 121 12:1–11. doi:10.1038/s41467-021-24357-5

Tamminen J, Lambon Ralph MA, Lewis PA. 2013. The role of sleep spindles and slow-wave activity in integrating new information in semantic memory. J Neurosci 33:15376–15381. doi:10.1523/JNEUROSCI.5093-12.2013

Tamminen J, Payne JD, Stickgold R, Wamsley EJ, Gaskell MG. 2010. Sleep spindle activity is associated with the integration of new memories and existing knowledge. J Neurosci 30:14356–60. doi:10.1523/JNEUROSCI.3028-10.2010

Zhu Y, Wang Q, Zhang L. 2021. Study of EEG characteristics while solving scientific problems with different mental effort. Sci Rep 11. doi:10.1038/S41598-021-03321-9

Author Response

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

In this study, the researchers aimed to investigate the cellular landscape and cell-cell interactions in cavernous tissues under diabetic conditions, specifically focusing on erectile dysfunction (ED). They employed single-cell RNA sequencing to analyze gene expression patterns in various cell types within the cavernous tissues of diabetic individuals. The researchers identified decreased expression of genes associated with collagen or extracellular matrix organization and angiogenesis in several cell types, including fibroblasts, chondrocytes, myofibroblasts, valve-related lymphatic endothelial cells, and pericytes. They also discovered a newly identified marker, LBH, that distinguishes pericytes from smooth muscle cells in mouse and human cavernous tissues. Furthermore, the study revealed that pericytes play a role in angiogenesis, adhesion, and migration by communicating with other cell types within the corpus cavernosum. However, these interactions were found to be significantly reduced under diabetic conditions. The study also investigated the role of LBH and its interactions with other proteins (CRYAB and VIM) in maintaining pericyte function and highlighted their potential involvement in regulating neurovascular regeneration. Overall, the manuscript is well-written and the study provides novel insights into the pathogenesis of ED in patients with diabetes and identifies potential therapeutic targets for further investigation.

Reviewer #2 (Public Review):

Summary: In this manuscript, the authors performed single cell RNA-sequencing of cells from the penises of healthy and diabetes mellitus model (STZ injection-based) mice, identified Lbh as a marker of penis pericytes, and report that penis-specific overexpression of Lbh is sufficient to rescue erectile function in diabetic animals. In public human single cell RNA-sea datasets, the authors report that LBH is similarly specific to pericytes and down regulated in diabetic patients. Additionally, the authors report discovery of CRYAB and VIM1 as protein interacting partners with LBH.

The authors contributions are of interest to the erectile dysfunction community and their Lbh overexpression experiments are especially interesting and well-conducted. However, claims in the manuscript regarding the specificity of Lbh as a pericyte marker, the mechanism by which Lbh overexpression rescues erectile function, cell-cell interactions impaired by diabetes, and protein-interaction partners require qualification or further evidence to justify.

Major claims and evidence:

1) Marker gene specificity and quantification: One of the authors' major contributions is the identification of Lbh as a marker of pericytes in their data. The authors present qualitative evidence for this marker gene relationship, but it is unclear from the data presented if Lbh is truly a specific marker gene for the pericyte lineage (either based on gene expression or IF presented in Fig. 2D, E). Prior results (see Tabula Muris Consortium, 2018) suggest that Lbh is widely expressed in non-pericyte cell types, so the claims presented in the manuscript may be overly broad. Even if Lbh is not a globally specific marker, the authors' subsequent intervention experiments argue that it is still an important gene worth studying.

Answer: We appreciate this comment. In our scRNAseq data for the mouse cavernosum tissues, previously known markers such as Rgs5, Pdgfrb, Cspg4, Kcnj8, Higd1b, and Cox4i2 were found to be expressed not exclusively in pericytes, while Lbh exhibited specific expression patterns in pericytes (Fig. 2 and Supplementary Fig. 5). LBH expression was easily distinguishable from α-SMA, not only in mouse cavernosum but also in dorsal artery and dorsal vein tissues within penile tissues. This distinctive expression pattern of LBH was also observed in the human cavernous pericytes (Fig. 5). Then, we examined Lbh expression patterns in various mouse tissues using the mouse single-cell atlas (Tabula Muris), although endothelial and pericyte clusters were not subclustered in most tissues from Tabula Muris. To identify pericytes, we relied on the expression pattern of known marker genes (Pecam1 for endothelial cells, Rgs5, Pdgfrb, and Cspg4 for pericytes). Lbh was expressed in pericytes of the bladder, heart and aorta, kidney, and trachea but not as specifically in penile pericytes (Supplementary Fig. 6A-D). However, it is worth noting that other known pericyte markers were also did not exhibit exclusive expression in pericytes across all the tissues we analyzed. Therefore, in certain tissues, particularly in mouse penile tissues, Lbh may be a valuable marker in conjunction with other established pericyte marker genes for distinguishing pericytes.

2) Cell-cell communication and regulon activity changes in the diabetic penis: The authors present cell-cell communication analysis and TF regulon analysis in Fig 3 and report differential activities in healthy and DM mice. These results are certainly interesting, however, no statistical analyses are performed to justify claimed changes in the disease state and no validations are performed. It is therefore challenging to interpret these results, and the relevant claims do not seem well supported.

Answer: In response to these helpful suggestions, we calculated statistical significance and performed experimental validation. CellphoneDB permutes the cluster labels of all cells 1000 times and calculates the mean(mean(molecule 1 in cluster X), mean(molecule 2 in cluster Y)) at each time for each interaction pair, for each pairwise comparison between two cell types. We only considered interactions in which the difference in means calculated by these permutations were greater than 0.25-fold between diabetes and normal. Also, we considered that the interactions with P-value < 0.05 were significant.

To assess differential regulon activities of transcription factor (SCENIC) between diabetic and normal pericytes, we utilized a generalized linear model with scaled activity scores for each cell as input. These scaled regulon activity values for angiogenesis-related TFs exhibited differences between diabetic and normal pericytes. The results of the generalized linear model revealed that Klf5, Egr1, and Junb were TFs with significantly altered regulon activities in diabetic pericytes. Experimental data indicated that the expression level of Lmo2, Junb, Elk1, and Hoxd10 was higher (Hoxd10) or lower (Lmo2, Junb, Elk1) in diabetic pericytes compared to normal pericytes (Supplementary Fig. 9). We have added the scaled regulon activity values and statistical significance in Fig. 3E.

3) Rescue of ED by Lbh overexpression: This is a striking and very interesting result that warrants attention. By simple overexpression of the pericyte marker gene Lbh, the authors report rescue of erectile function in diabetic animals. While mechanistic details are lacking, the phenomenon appears to have a large effect size and the experiments appear sophisticated and well conducted. If anything, the authors appear to underplay the magnitude of this result.

Answer: We appreciate this comment. Therefore, we have added relevant clarification in the revised manuscript discussion section to emphasize the importance of LBH overexpression on rescuing ED as follows: “To test our hypothesis, we utilized the diabetes-induced ED mouse model, commonly employed in various studies focusing on microvascular complications associated with type 1 diabetes. We observed that the overexpression of LBH in diabetic mice led to the restoration of reduced erectile function by enhancing neurovascular regeneration. However, this study primarily demonstrated the observed phenomenon without delving into the detailed mechanisms. Nonetheless, these results of LBH on erections provide us with new strategies for treating ED and should be of considerable concern.” (Please see revised ‘Discussion’)

4) Mechanistic claims for rescue of ED by Lbh overexpression: The authors claim that cell type-specific effects on MPCs are responsible for the rescue of erectile function induced by Lbh overexpression. This causal claim is unsupported by the data, which only show that Lbh overexpression influences MPC performance. In vivo, it's likely that Lbh is being over expressed by diverse cell types, any of which could be the causal driver of ED rescue. In fact, the authors report rescue of cell type abundance in endothelial cells and neuronal cells. Therefore, it cannot be concluded that MPC effects alone or in principal are responsible for ED rescue.

Answer: We agree with these claims. Therefore, we have added relevant clarifications in the discussion section of the revised manuscript. Our findings suggest that LBH can affect the function of cavernous pericytes, although we cannot definitively specify which particular cavernous cell types are affected by the overexpressed LBH, whether it be cavernous endothelial cells, smooth muscle cells, or others. Subsequent research will be required to conduct more comprehensive mechanistic investigations, such as in vitro studies using cavernous endothelial cells, smooth muscle cells, and fibroblasts to address these knowledge gaps. (Please see revised ‘Discussion’)

5) Protein interaction data: The authors claim that CRYAB and VIM1 are novel interacting partners of LBH. However, the evidence presented (2 blots in Fig. 6A,B) lack the relevant controls. It is possible that CRYAB and VIM1 are cross-reactive with the anti-LBH antibody or were not washed out completely. The abundance of bands on the Coomassie stain in Fig. 6A suggests that either event is plausible. Therefore, the evidence presented is insufficient to support the claim that CRYAB and VIM1 are protein interacting partners of LBH.

Answer: We agree with these claims. Therefore, we have added the relevant controls(Input) and performed Co-IP (IP: CRYAB or VIM, WB: LBH) to demonstrate CRYAB and VIM1 are not simply cross-reactive antigens to their LBH antibody. Our results show that we can detect the expression of CRYAB and VIM after LBH IP, and we also detect the expression of LBH after CRYAB and VIM IP. In addition, it can be seen from our results that the binding of LBH to VIM is higher than that of CRYAB. Regardless, these results indicate that the binding of CRYAB or VIM to LBH is not a random phenomenon. (Please see revised ‘Result’ and ‘Figure 6B’)

Impact: These data will trigger interest in Lbh as a target gene within the erectile dysfunction community.

Reviewer #3 (Public Review):

Bae et al. described the key roles of pericytes in cavernous tissues in diabetic erectile dysfunction using both mouse and human single-cell transcriptomic analysis. Erectile dysfunction (ED) is caused by dysfunction of the cavernous tissue and affects a significant proportion of men aged 40-70. The most common treatment for ED is phosphodiesterase 5 inhibitors; however, these are less effective in patients with diabetic ED. Therefore, there is an unmet need for a better understanding of the cavernous microenvironment, cell-cell communications in patients with diabetic ED, and the development of new therapeutic treatments to improve the quality of life.

Pericytes are mesenchymal-derived mural cells that directly interact with capillary endothelial cells (ECs). They play a vital role in the pathogenesis of erectile function as their interactions with ECs are essential for penile erection. Loss of pericytes has been associated with diabetic retinopathy, cancer, and Alzheimer's disease and has been investigated in relation to the permeability of cavernous blood vessels and neurovascular regeneration in the authors' previous studies. This manuscript explores the mechanisms underlying the effect of diabetes on pericyte dysfunction in ED. Additionally, the cellular landscape of cavernous tissues and cell type-specific transcriptional changes were carefully examined using both mouse and human single-cell RNA sequencing in diabetic ED. The novelty of this work lies in the identification of a newly identified pericyte (PC)-specific marker, LBH, in mouse and human cavernous tissues, which distinguishes pericytes from smooth muscle cells. LBH not only serves as a cavernous pericyte marker, but its expression level is also reduced in diabetic conditions. The LBH-interacting proteins (Cryab and Vim) were further identified in mouse cavernous pericytes, indicating that these signaling interactions are critical for maintaining normal pericyte function. Overall, this study demonstrates the novel marker of pericytes and highlights the critical role of pericytes in diabetic ED.

Reviewer #1 (Recommendations For The Authors):

1) The methods are poorly written. It lacks specific information on the sample size, experimental design, and data analysis methods employed. The absence of these crucial details makes it difficult to evaluate the robustness and reliability of the findings.

Answer: We agree with the reviewer’s suggestion, now we revised the methods of our manuscript, and added detailed information or references. For sample size we have added detailed information in Figure legend (Please see revised ‘Method’ , Figure Legend, and Supplementary information.)

2) The cell number in the scRNA-seq analysis is small (~12000) and some minor cell types are probably underrepresented. It is not clear whether the authors pooled the cells from different mice as one sample, or replicates in different groups have been included. It will be helpful to label different samples in the UMAP. The authors should repeat the experiments with more replicates to increase the cell number and validate the findings.

Answer: We understand the reviewer's concern, but due to the small size of mouse penile tissue, we had to pool 5 corpus cavernosum tissues for each group (using pooled samples) for scRNA-seq analysis. Moreover, owing to the unique nature of mouse penile tissue, which is highly resistant, it posed challenges for the dissolution and isolation of single cells using conventional single-cell separation methods. Consequently, we had to increase the concentration of the enzyme to finally obtain 12,894 cells. Rather than conducting a repetitive scRNAseq analysis on the same mouse model, we validated our findings in human cavernous single-cell transcriptome data. This analysis allowed us to confirm the presence of pericyte in human corpus cavernosum, specific expression of LBH in human cavernous pericytes, and the identification of relevant GO terms associated with pericyte functions (Figure 5). We have add these information in ‘Method’ (Please see revised ‘Method’).

3) Functional studies are lacking to justify how manipulating LBH expression or its interacting proteins might lead to effective therapeutic approaches for diabetic ED.

Answer: We have performed the functional study to evaluate LBH expression might lead to effective therapeutic approaches for diabetic ED as showed in Figure 4G. Assessment of intracavernous pressure (ICP) is the most representative test for evaluating erectile function. Therefore, we modulated LBH expression in the penis of diabetic mice and assessed the erectile function of the mice by intracavernous pressure. However, we have not performed ICP studies and relative in vitro studies (migration, survival experiment) to assess whether LBH-interacting proteins have the same effect.

4) Although the abstract identifies novel targets for potential interventions, such as LBH and its interacting proteins, the clinical relevance of these findings remains uncertain. The authors should include a discussion regarding the translation of these discoveries into therapeutic strategies or their potential impact on patients with diabetes and ED.

Answer: We appreciate the reviewer's suggestion and have added a discussion as per the reviewer’s recommendation (Please see revised ‘Discussion’).

5) While the study highlights the importance of pericytes in penile erection, it fails to mention the broader context of other cell types involved in the pathogenesis of ED. Neglecting to discuss potential contributions from endothelial cells, smooth muscle cells, or neural elements limits the comprehensive understanding of the cellular interactions underlying diabetic ED.

Answer: We agree with the reviewer's suggestion and have added a discussion regarding the significance of other cell populations in penile tissues, such as endothelial cells, smooth muscle cells fibroblasts, and neural elements, along with the rationale for our focus on pericytes. (Please see revised ‘Discussion’).

Reviewer #2 (Recommendations For The Authors):

We congratulate the authors on an interesting study. We were especially excited to see their Lbh overexpression results. However, we felt other claims in the paper could benefit from additional investigation, analysis, and statistical rigor. We have provided a set of suggestions for improvement below.

Major points:

1) Pericyte marker gene proposal: See public review for commentary on the following suggested experiments. The authors should perform binary classification analysis using Lbh and report the performance of this gene as a marker (e.g. using the area under the receiver operating characteristic, accuracy, precision and recall). Further, they should consider performing this analysis for all other genes in their data to determine whether Lbh is the best marker gene.

Answer: We appreciate this comment. AUC scores of Rgs5, Pln, Ednra, Npylr, Atp1b2, and Gpc3 for ability of a binary classifier to distinguish between pericyte and the other cell types in mouse penile tissues were measured by using FindMarkers function. Rgs5 had the highest AUC, but Rgs5 was also expressed in SMCs in our data. Pln, Ednra, Gpc3, and Npy1r also seemed to be candidate markers, but the literature search excluded these genes as they are also expressed in the SMCs of other tissues or different cell types. The AUC score of Lbh was over 0.7, and expression in SMC was not identified in previous studies, and ultimately, we experimentally identified that Lbh is penis pericyte specific. We have added this to the manuscript.

Author response table 1.

Robust differential expression analysis should also be performed for this gene (if not all) and the statistics should be reported, given known issues with the statistical approach used by the authors for differential expression (see: Squair 2021, 10.1038/s41467-021-25960-2). The authors' should also report the number of cells involved in these comparisons, as the number of pericytes in the data (Fig 1B) appears quite small.

Answer: We appreciate this comment. We used “MAST” to identify differentially expressed genes. This test is often used to find DEGs in single-cell RNA data. However, because the pseudobulk method has advantages over the single cell DEG method (Squair 2021, 10.1038/s41467-021-25960-2), we additionally performed DEG analysis with DESeq2 to confirm whether Lbh can distinguish pericytes from other cell types in the penile. As a result, even when tested with DESeq2, Lbh expression was significantly higher in pericytes than in other cell types in penile (adjusted p-value = 2.694475e-07 in Pericyte vs SMC, adjusted P-value = 3.700118e-58 in Pericyte vs the other cell types). Mouse penile tissue is small in size, and the number of pericytes in mouse penile tissue is relatively smaller compared to fibroblasts and chondrocytes. In our mouse penile scRNAseq data, the number of pericytes is as follows: normal: 58, diabetes: 116. Despite the limited number of cells, we were able to establish statistical significance in our analyses.

Immunostaining results in Fig. 2D, E should likewise be quantified. At present, it's unclear that LBH and aSMA are mutually exclusive as claimed. The authors should also investigate Lbh expression in public single cell genomics data, rather than performing candidate gene literature searches. For example, the Tabula Muris suggests Lbh is expressed widely outside pericytes.

Answer: For Figure 2D and E, the aim of these analyses was to assess the distribution of LBH and other cellular markers to see if they overlap and if they can be distinguished. We think that some of the overlapping staining in the tissue may be caused by multilayered cellular structures, so staining within cells would be more convincing. Therefore, we quantified the percentage of LBH- or α-SMA-expressed pericytes and relative expression in smooth muscle cells in cell staining (Supplementary Fig. 5E). We found that only 3% of smooth muscle cells expressed LBH, 67% of mouse cavernous pericytes (MCPs) expressed α-SMA, and more than 97% of MCPs expressed LBH. Therefore, these results may illustrate the specific expression of LBH in MCPs. These information was added as ‘Supplementary Fig. 5E’ (Please see revised ‘Supplementary information’). We also examined Lbh expression patterns in various mouse tissues using the public mouse single-cell atlas (Tabula Muris), and provided a detailed response in reviewer 2’s public review 1.

Even if Lbh is not the best marker, the authors' intervention experiment still motivates study of the gene, but these analyses would help contextualize the result for readers.

2) Statistical anslyses for cell-cell communication and TF regulon analysis: See public review for context on these comments. The authors should perform statistical tests to evaluate the significance of differences detected for each of these analysis. For example, generalized linear models can be used to assess the significance of TF regulon activity scores from SCENIC, and permutation tests can be used to measure the significance of cell-cell interaction score changes. Without these statistical tests, it's challenging for a reader to interpret whether the results reported are meaningful or within the realm of experimental noise.

Answer: We appreciate this comment. We calculated statistical significance TF regulon analyses as suggested by the reviewer and described a detailed statistical calculation method for cell-cell communication. We provided a detailed response in reviewer 2’s public review 2.

3) Mechanism of ED rescue by Lbh overexpression: To support this claim, the authors would need to perform an experiment where Lbh is over expressed specifically in MPCs (using e.g. a specific promoter on their LTV construct, or a transgenic line with a cell type-specific Cre-Lox system). Absent these data, the claim should be removed.

Answer: We agree with the reviewer's suggestion and we have reworked the claim that ‘LBH overexpression is affected by pericytes during ED recovery’ and have added relevant clarification in the Discussion section to clearly state that LBH overexpression may affect many cavernosum cells, such as cavernous endothelial cells, smooth muscle cells, fibroblasts, and pericytes (Please see revised ‘Result’ and ‘Discussion’)

4) Protein interaction claims: This experiment would require that the authors perform a similar pull-down with LBH KO cells and or a reciprocal Co-IP (e.g. IP: CRYAB or VIM1, WB: LBH) to demonstrate CRYAB and VIM1 are not simply cross-reactive antigens to their LBH antibody. Further, these experiments appear to only have a single replicate for each condition. The authors should either remove associated claims, or perform a Co-IP experiment with the relevant controls with sufficient replication.

Answer: We agree with the claims. Therefore, we have included the necessary controls (Input) and performed Co-IP (IP: CRYAB or VIM1, WB: LBH) to demonstrate that CRYAB and VIM1 are not simply cross-reactive antigens to their LBH antibody. Our results show that we can detect the expression of CRYAB and VIM after LBH IP, and we also detect the expression of LBH after CRYAB and VIM IP. In addition, it can be seen from our results that the binding of LBH to VIM is higher than that of CRYAB. Regardless, these results indicate that the binding of CRYAB or VIM to LBH is not a random phenomenon. Additionally, all IP experiments were replicated at least three times. (Please see revised ‘Result’ and ‘Figure 6B’)

Minor Points:

• The reference "especially in men" on line 56 seems odd given that only males can experience penile erectile dysfunction.

Answer: We agree with the reviewer's suggestion and have removed the description 'especially male' (Please see revised ‘Introduction’)

• Line 109, it's unclear what genes showed altered expression in Schwann cells.

Answer: We apologize for the confusion. There was no significant differentially expressed genes between normal and diabetes in Schwann cells. We revised this part in the manuscript. (Schwann cells showed an increased expression compared to normal cells in diabetes, though not significant. In Schwann cells, there were no significant DEGs between diabetic and normal cells.)

• It would be helpful for readers to see an analysis of the cell types that are transduced in the Lbh overexpression experiment in vivo. At present, some pericyte specificity is implied, but not demonstrated.

Answer: We appreciate this comment. Our findings suggest that LBH can affect the function of cavernous pericytes, although we cannot definitively conclude which specific-cavernous cell types are affected by the overexpressed LBH, whether it be cavernous endothelial cells, smooth muscle cells, or others. Subsequent research will be required to conduct more comprehensive mechanistic investigations, such as in vitro studies using cavernous endothelial cells, smooth muscle cells, and fibroblasts to address these knowledge gaps. These were also mentioned in the manuscript.

• To improve clarity and enhance readability, define abbreviations before their initial usage in the text. For instance, in the second paragraph of the Introduction, the abbreviation 'ECs' is used without prior definition. It can be inferred that it is referring to endothelial cells, mentioned in parentheses in the subsequent sentence.

Answer: We agree with the reviewer's suggestion to expand acronyms and ensure that all acronyms are defined in the revised manuscript before they are used for the first time in the text (Please see revised Manuscript).

• It is important to include relevant references that align with the content being discussed. For example, in the Introduction, pericytes are described as being involved in various processes such as angiogenesis, vasoconstriction, and permeability. The text refers to a single reverence, a review by Gerhardt and Besholtz, which primarily focuses on pericyte's role in regulating angiogenesis. Adding additional sources, such as the review by Bergers and Song (Neuro Oncol., 2005) is recommended.

Answer: We agree with the reviewer's suggestion, and have added the reference as reviewer recommended (Please see revised Manuscript and reference).

• Figure 3E: it is stated that a panel of 53 angiogenesis factors were tested, it is stated that only MMP3 showed increased expression. However, various unlabeled spots appear to show changed expression patterns. It would be helpful to show a summary graph with the relative intensities of the full array of factors tested.

Answer: We agree with the reviewer’s suggestion, now we showed all spots density in angiogenesis array as Supplementary Table 1. The condition of the spots we selected was that the expression density was at least above 1500, and the change ratio was greater than 1.2. (Please see revised ‘Supplementary information’)

Reviewer #3 (Recommendations For The Authors):

Detailed statistical power calculation

Data availability statement( were both mouse and human scRNA deposited in GEO with a taken and when will they be released to the public?)

Answer: Human scRNA data have been deposited in GEO under accession number GSE206528. Our mouse scRNA dataset has been uploaded to KoNA and is available for download (https://www.kobic.re.kr/kona/review?encrypt_url=amlod2FucGFya3xLQUQyMzAxMDEz)

Major concerns about this work

1) The single cell RNAseq data collected for mouse diabetic ED(Fig 1B), FB are the most abundant cell population compared to PC, EC, SMC and other clusters. The rationale for studying FB clusters (in Figure 1, D-F) instead of PC cluster is unclear. Which cluster DEG did the authors annotate for Fig 1G-H?

Answer: We understand the reviewer's suggestion and confusion. Although other major cell populations in penile tissue such as smooth muscle cells, endothelial cell, and fibroblasts have been extensively studied, pericytes have mainly been investigated in the context of the central nervous system (CNS). For example, in the CNS, pericytes are involved in maintaining the integrity of the brain's blood-brain barrier (BBB) [PMID: 27916653], regulating blood flow at capillary junctions [PMID: 33051294], and promoting neuroinflammatory processes [PMID: 31316352], whose dysfunction is considered an important factor in the progression of vascular diseases such as Alzheimer's disease [PMID: 24946075]. But little is known about the role of pericytes in penile tissue [PMID: 35865945; PMID: 36009395; PMID: 26044953]. In order to explore the role of pericytes in repairing the corpus cavernosum vascular and neural tissues damaged by DM, we focused on pericytes, which are multipotent perivascular cells that contribute to the generation and repair of various tissues in response to injury. Although recent studies have shown that pericytes are involved in physiological mechanisms of erection, little is known about their detailed mechanisms. We have also added this rationale in discussion.

Single cell level study has not been conducted in mouse penile tissues. Therefore, before delving into pericytes, we aimed to identify overall transcriptome differences between normal and diabetic conditions in mouse penile tissues. We presented the analyses of FB, which make up the largest proportion among the cell types in the mouse penis, in Fig. 1D-F. The analysis of other cell types is provided in Supplementary Fig. 1-4. Fig. 1G-H are GO terms for Fibroblasts clusters. We added this information in the figure.

2) Fig 2 is the critical data to show Lbh is a cavernous PC specific marker. More PC violin plots to identify PC cluster such as Cspg4, Kcnj8, Higd1b, Cox4i2 and more SMC violin plots to identify SMC cluster such as Acta2, Myh11, Tagln, Actg2 should be used for inclusion and exclusion of PC( the same concern applied to human scRNAseq in Fig 5B).

Answer: We appreciate this comment. We examined the expression of other marker genes of pericytes and SMCs. Although some marker genes were rarely expressed in the mouse penis data (Kcnj8, Higd1b), the expression of marker genes tended to be relatively high in each cluster. The expression of Cspg4 and Cox4i2 was higher in pericytes than in SMCs, while the expression of Acta2, Myh11,and Tagln was higher in SMCs than in pericytes. Actag2 was specifically expressed in SMCs. Through the gene set enrichment test as well as the expression of known cell type marker genes, we identified that the annotation of pericyte and SMC was appropriate (Fig. 2B and Fig. 5C). We added the violin plots of these marker genes in Supplementary Fig. 5.

Author response image 1.

(Mouse)

In human penis data, ACTA2 and MYH11 were expressed in SMCs, pericytes, and myofibroblasts, as in the previous paper [PMID: 35879305]. Among pericyte markers, the number of cells expressing KCNJ8 and HIGD1B was small. The cluster we annotated as pericyte was double positive for pericyte markers CSPG4 and COX4I2. ACTG2, a marker for SMC, was expressed more highly in SMC than in pericytes and myofibroblasts. As in the mouse penis data, we identified that the annotation of each cell type was appropriate through the gene set enrichment test in the human penis data. We added the violin plots of CSPG4, COX4I2, and ACTG2 in Supplementary Fig. 11.

Author response image 2.

(Human)

When exploring Lbh expression levels in "Database of gene expression in adult mouse brain and lung vascular and perivascular cells" from https://betsholtzlab.org/VascularSingleCells/database.html, Lbh is not uniquely expressed in PC, suggesting its tissue-specific expression level. This difference should be discussed in the Discussion section.

Answer: We appreciate this valuable comment. For the answer to this comment, we extensively analyzed Lbh expression patterns in various mouse tissues using the public mouse single-cell atlas (Tabula Muris) as also suggested by Reviewer 2. Please see our detailed response in reviewer 2’s public review 1.

3) In prior studies on PC morphology and location (PMID: 21839917), they reside in capillaries (diameter less than 10um) or distal vessels (diameter less than 25um) and have oval cell body and long processes. Due to the non-specificity of Pdgfrb, SMC are positive for Pdgfrb staining (this has been shown in many publications that SMC are Pdgfrb+; unfortunately, NG2 antibody also stains for both PC and SMC). Therefore, the LBH immunostaining (in Fig 2D and 2E of large-sized vessels) are very likely for SMC identity, not PC. PC should be in close contact with CD31+ ECs in healthy conditions. The LBH immunostaining of PC in both mouse and human tissues (Fig 4) must be replaced and better characterized.

Answer: We agree with the reviewer's suggestion. As it is widely known, peicytes are primarily located in capillaries, where they surround endothelial cells of blood vessels. However, recent discoveries have identified cells with pericyte-like characteristics in the walls of large blood vessels, challenging the traditional concept [PMID: 27268036]. In our study, we observed minimal overlap in staining between LBH and α-SMA, suggesting that the cells expressing LBH were not smooth muscle cells but possibly pericyte-like cells in large vessels. In small vessels within the bladder, kidney, and even the aorta, we found LBH-expressing cells surrounding CD31-expressing vessels, consistent with the known characteristics of pericytes. Further research is needed to comprehend the differences in LBH expression and its characteristics in both large and small blood vessels. We have added discussions and references for this issue (Please see revised ‘Discussion’ and ‘Reference’)

4) How do mouse cavernous pericytes isolate? How is purity?

Answer: As the reviewer points out, we isolated mouse spongiform pericytes following our and other previously published methods. We used pigment epithelium-derived factor (PEDF), which removes non-pericytic cells [PMID: 30929324, 23493068]. Although there are no purity study results such as FACS, other staining results thoroughly support the notion that this method yields pericytes with a notably high level of purity. (Please see ‘Method’ section).

5) Can mouse scRNAseq cell-cell communication in Fig 3 be reproducible in human scRNAseq cell-cell communication? The results in human ED are more clinically significant than in mouse data.

Answer: In human scRNAseq data, the difference between angiogenesis-related interactions between normal and diabetes was not as significant as that in mouse data. Because the cell type composition of the human and mouse penis is not completely identical, there are limitations in comparing cell-cell interactions. However, in the human penis data, some interactions related to angiogenesis between pericytes and other cell types were decreased in diabetes compared to normal (boxed parts).

Author response image 3.

6) Fibroblasts also express Vim. Murine PC VIM/CRYAB( should be written as Vim/Cryab as mouse proteins) direct interaction with Lbh is unclear from Lbh IP as Fig 6A red boxes showed a wide range of sizes. Where is the band for Lbh? Do human PC LBH interact with VIM/CRYAB?

Answer: We agree with the reviewer's comment. VIM is a type III intermediate filament protein expressed in many cell types. We have added the relevant controls (Input) and performed Co-IP (IP: CRYAB or VIM, WB: LBH) to demonstrate CRYAB and VIM are not simply cross-reactive antigens to their LBH antibody. In western blot study, the LBH band was expressed between 35 kDa-48 kDa. From Figure 6A, we detected CRYAB in band 1 and VIM in bands 2 and 3. This may be due to the formation of dimers or multimers by VIM. We did not use human PCs for IP studies because IP requires large amounts of protein, making IP studies using human pericyte challenging. Nevertheless, the interaction between LBH and CRYAB in humans has been reported through fluorescent resonance energy transfer assay and affinity chromatography technology assay [PMID:34000384, PMID:20587334].

7) In Fig 6H and I, why does CRYAB expression significantly reduce in vitro and in vivo under diabetic conditions, whereas VIM expression significantly increases?

Answer: As the reviewer pointed out, and we have discussed on this issue in the manuscript, CRYAB is known to promote angiogenesis. Diabetes reduces CRYAB expression, so angiogenesis may be impaired. Furthermore, since VIM is a multifunctional protein, it interacts with several other proteins with multiple functions under various pathophysiological conditions. There are many relevant literatures showing that VIM expression is increased under diabetic conditions [PMID: 28348116 and PMID: 32557212]. And VIM deficiency protects against obesity and insulin resistance in patients with type 2 diabetes. Therefore, we hypothesize that exogenous LBH may have the ability to bind to the increased VIM in diabetic conditions and inactivate the effects of VIM. Thereby achieving the protective effect. This needs to be proved in further studies.

8) The therapeutic strategies targeting (Lbh-Cryab-Vim) on mouse diabetic ED model is not investigated and need to be further validated and discussed.

Answer: As the reviewers pointed out, in this study, we did not evaluate the targeted therapeutic strategy for LBH-CRYAB-VIM in a mouse diabetic ED model. We only identified the binding potential of these three proteins. Evaluation of this treatment strategy requires further study. For example, we can employ shRNA lentivirus, either alone or in combination, to downregulate CRYABexpression [PMID: 31612679] in normal mice, utilize a lentiviral vector CMV-GFP-puro-vimentin to overexpress Vimentin [PMID: 36912679], and then treat it with LBH to evaluate whether the LBH effect still exists (in vivo erectile function study and in vitro angiogenesis assay). We include this information in the Discussion section as a limitation of this study (Please see revised ‘Discussion’).

9) The Discussion of current knowledge of pericytes in diabetic ED and other diseases and the significance of this study as well as clinical implications, should be expanded.

Answer: As the reviewers pointed out, we have expanded the current knowledge of pericytes in diabetic ED and other diseases (CNS disease) and clinical implications as follows: “Although other major cell populations in penile tissue such as smooth muscle cells, endothelial cell, and fibroblasts have been extensively studied, pericytes have mainly been investigated in the context of the central nervous system (CNS). For example, in the CNS, pericytes are involved in maintaining the integrity of the brain's blood-brain barrier (BBB), regulating blood flow at capillary junctions, and promoting neuroinflammatory processes, whose dysfunction is considered an important factor in the progression of vascular diseases such as Alzheimer's disease. But little is known about the role of pericytes in penile tissue.” (Please see revised ‘Discussion’).

10) How many clinical samples were used? How many times did each experiment repeat?

Answer: As the reviewers pointed out, the clinical samples’ information was added in ‘method’ section. A total four human samples were used in this study (‘human corpus cavernosum tissues were obtained from two patients with congenital penile curvature (59-year-old and 47-year-old) who had normal erectile function during reconstructive penile surgery and two patients with diabetic ED (69-year-old and 56-year-old) during penile prosthesis implantation.’). For in vivo study, we quantified four different fields from human samples.

Minor concerns

1) Fig 1A, why normal mouse's body size is the same as DM?

Answer: As the reviewer pointed out, in Figure 1A, while the size of normal mice and DM mice may not appear significantly different, there are indeed notable difference in body weight and size. The normal mice body weigh we used was about 30 grams, while DM mice body weigh was generally less than 24 grams. We found that we missed information on physiological and metabolic parameters from in vivo studies (ICP function study). Therefore, we have added it in Supplementary Table 2 (Please see revised ‘Supplementary information’)

2) The label and negative, and positive controls for Fig 6B are missing.

Answer: We thank for pointing out this. We have added the relevant controls (Input) and performed Co-IP (IP: CRYAB or VIM1, WB: LBH) to demonstrate CRYAB and VIM1 are not simply cross-reactive antigens to their LBH antibody and all IP was replicated for at least 3 times. (Please see revised ‘Result’ and ‘Figure 6B’)

3) The limitation of this study and future work should be discussed.

Answer: As the reviewer pointed out, we have added the limitation of this study and future direction in the discussion section (Please see revised ‘Discussion’).

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors report an fMRI investigation of the neural mechanisms by which selective attention allows capacity-limited perceptual systems to preferentially represent task-relevant visual stimuli. Specifically, they examine competitive interactions between two simultaneously-presented items from different categories, to reveal how task-directed attention to one of them modulates the activity of brain regions that respond to both. The specific hypothesis is that attention will bias responses to be more like those elicited by the relevant object presented on its own, and further that this modulation will be stronger for more dissimilar stimulus pairs. This pattern was confirmed in univariate analyses that measured the mass response of a priori regions of interest, as well as multivariate analyses that considered the patterns of evoked activity within the same regions. The authors follow these neuroimaging results with a simulation study that favours a "tuning" mechanism of attention (enhanced responses to highly effective stimuli, and suppression for ineffective stimuli) to explain this pattern.

Strengths:

The manuscript clearly articulates a core issue in the cognitive neuroscience of attention, namely the need to understand how limited perceptual systems cope with complex environments in the service of the observer's goals. The use of a priori regions of interest, and the inclusion of both univariate and multivariate analyses as well as a simple model, are further strengths. The authors carefully derive clear indices of attentional effects (for both univariate and multivariate analyses) which makes explication of their findings easy to follow.

Weaknesses:

There are some relatively minor weaknesses in presentation, where the motivation behind some of the procedural decisions could be clearer. There are some apparently paradoxical findings reported -- namely, cases in which the univariate response to pairs of stimuli is greater than to the preferred stimulus alone -- that are not addressed. It is possible that some of the main findings may be attributable to range effects: notwithstanding the paradox just noted, it seems that a floor effect should minimise the range of possible attentional modulation of the responses to two highly similar stimuli. One possible limitation of the modelled results is that they do not reveal any attentional modulation at all under the assumptions of the gain model, for any pair of conditions, implying that as implemented the model may not be correctly capturing the assumptions of that hypothesis.

We thank the reviewer for the constructive comments. In response, in the current version of the manuscript we have improved the presentation. We further discuss how the response in paired conditions is in some cases higher than the response to the preferred stimulus in this letter. For this, we provide a vector illustration, and a supplementary figure of the sum of weights to show that the weights of isolated-stimulus responses for each category pair are not bound to the similarity of the two isolated responses.

Regarding the simulation results, we have clarified that the univariate effect of attention is not the attentional modulation itself, but the change in the amount of attentional modulation in the two paired conditions. We provide an explanation for this in this letter below, and have changed the term “attentional modulation” to “univariate shift” in the manuscript to avoid the confusion.

Reviewer #2 (Public Review):

Summary:

In an fMRI study requiring participants to attend to one or another object category, either when the object was presented in isolation or with another object superimposed, the authors compared measured univariate and multivariate activation from object-selective and early visual cortex to predictions derived from response gain and tuning sharpening models. They observed a consistent result across higher-level visual cortex that more-divergent responses to isolated stimuli from category pairs predicted a greater modulation by attention when attending to a single stimulus from the category pair presented simultaneously, and argue via simulations that this must be explained by tuning sharpening for object categories.

Strengths:

- Interesting experiment design & approach - testing how category similarity impacts neural modulations induced by attention is an important question, and the experimental approach is principled and clever.

- Examination of both univariate and multivariate signals is an important analysis strategy.

- The acquired dataset will be useful for future modeling studies.

Weaknesses:

- The experimental design does not allow for a neutral 'baseline' estimate of neural responses to stimulus categories absent attention (e.g., attend fixation), nor of the combination of the stimulus categories. This seems critical for interpreting results (e.g., how should readers understand univariate results like that plotted in Fig. 4C-D, where the univariate response is greater for 2 stimuli than one, but the analyses are based on a shift between each extreme activation level?).

We are happy to clarify our research rationale. We aimed to compare responses in paired conditions when the stimuli were kept constant while varying the attentional target. After we showed that the change in the attentional target resulted in a response change , we compared the amount of this response change to different stimulus category pairs to investigate the effect of representation similarity between the target and the distractor on the response modulation caused by attentional shift. While an estimate of the neural responses in the absence of attention might be useful for other modeling studies, it would not provide us with more information than the current data to answer the question of this study.

Regarding the univariate results in Fig. 4C-D (and other equivalent ROI results in the revised version) and our analyses, we did not impose any limit on the estimated weights of the two isolated responses in the paired response and thus the sum of the two weights could be any number. We however see that the naming of “weighted average”, which implies a sum of weights being capped at one, has been misleading . We have now changed the name of this model to “linear combination” to avoid confusion

Previous studies (Reddy et al., 2009, Doostani et al., 2023) using a similar approach have shown a related results pattern: the response to multiple stimuli is higher than the average, but lower than the sum of the isolated responses, which is exactly what our results suggest. We have added discussion on this topic in the Results section in lines 409-413 for clarification:

“Note that the response in paired conditions can be higher or lower than the response to the isolated more preferred stimulus (condition Mat), depending on the voxel response to the two presented stimuli, as previously reported (Doostani et al. 2023). This is consistent with previous studies reporting the response to multiple stimuli to be higher than the average, but lower than the sum of the response to isolated stimuli (Reddy et al. 2009).”

We are not sure what the reviewer means by “each extreme activation level”. Our analyses are based on all four conditions. The two isolated conditions are used to calculate the distance measures and the two paired conditions are used for calculating the shift index. Please note that either the isolated or the paired conditions could show the highest response and we seeboth cases in our data. For example, as shown in Figure 4A in EBA, the isolated Body condition and the paired BodyatCar condition show the highest activation levels for the Body-Car pair, whereas in Figure 4C, the two paired conditions (BodyatCat and BodyCatat) elicit the highest response.

- Related, simulations assume there exists some non-attended baseline state of each individual object representation, yet this isn't measured, and the way it's inferred to drive the simulations isn't clearly described.

We agree that the simulations assume a non-attended baseline state, and that we did not measure that state empirically. We needed this non-attended response in the simulations to test which attention mechanism led to the observed results. Thus, we generated the non-attended response using the data reported in previous neural studies of object recognition and attention in the visual cortex (Ni et al., 2012, Bao and Tsao, 2018). Note that the simulations are checking for the profile of the modulations based on category distance. Thus, they do not need to exactly match the real isolated responses in order to show the effect of gain and tuning shift on the results. We include the clarification and the range of neural responses and attention parameters used in the simulations in the revised manuscript in lines 327-333:

“To examine which attentional mechanism leads to the effects observed in the empirical data, we generated the neural response to unattended object stimuli as a baseline response in the absence of attention, using the data reported by neural studies of object recognition in the visual cortex (Ni et al., 2012, Bao and Tsao, 2018). Then, using an attention parameter for each neuron and different attentional mechanisms, we simulated the response of each neuron to the different task conditions in our experiment. Finally, we assessed the population response by averaging neural responses.”

- Some of the simulation results seem to be algebraic (univariate; Fig. 7; multivariate, gain model; Fig. 8)

This is correct. We have used algebraic equations for the effect of attention on neural responses in the simulations. In fact, thinking about the two models of gain and tuning shift leads to the algebraic equations, which in turn logically leads to the observed results, if no noise is added to the data. The simulations are helpful for visualizing these logical conclusions. Also, after assigning different noise levels to each condition for each neuron, the results are not algebraic anymore which is shown in updated Figure 7 and Figure 8.

- Cross-validation does not seem to be employed - strong/weak categories seem to be assigned based on the same data used for computing DVs of interest - to minimize the potential for circularity in analyses, it would be better to define preferred categories using separate data from that used to quantify - perhaps using a cross-validation scheme? This appears to be implemented in Reddy et al. (2009), a paper implementing a similar multivariate method and cited by the authors (their ref 6).

Thank you for pointing out the missing details about how we used cross-validation. In the univariate analysis, we did use cross validation, defining preferred categories and calculating category distance on one half of the data and calculating the univariate shift on the other half of the data. Similarly, we employed cross-validation for the multivariate analysis by using one half of the data to calculate the multivariate distance between category pairs, and the other half of the data to calculate the weight shift for each category pair. We have now added this methodological information in the revised manuscript.

- Multivariate distance metric - why is correlation/cosine similarity used instead of something like Euclidean or Mahalanobis distance? Correlation/cosine similarity is scale-invariant, so changes in the magnitude of the vector would not change distance, despite this likely being an important data attribute to consider.

Since we are considering response patterns as vectors in each ROI, there is no major difference between the two measures for similarity. Using euclidean distance as a measure of distance (i.e. inverse of similarity) we observed the same relationship between weight shift and category euclidean distance. There was a positive correlation between weight shift and the euclidean category distance in all ROIs ( ps < 0.01, ts > 2.9) except for V1 (p = 0.5, t = 0.66). We include this information in the revised manuscript in the Results section lines 513-515:

“We also calculated category distance based on the euclidean distance between response patterns of category pairs and observed a similarly positive correlation between the weight shift and the euclidean category distance in all ROIs (ps < 0.01, ts >2.9) except V1 ( p = 0.5, t = 0.66).”

- Details about simulations implemented (and their algebraic results in some cases) make it challenging to interpret or understand these results. E.g., the noise properties of the simulated data aren't disclosed, nor are precise (or approximate) values used for simulating attentional modulations.

We clarify that the average response to each category was based on previous neurophysiology studies (Ni et al., 2012, Bao and Tsao, 2018). The attentional parameter was also chosen based on previous neurophysiology (Ni et al., 2012) and human fMRI (Doostani et al., 2023) studies of visual attention by randomly assigning a value in the range from 1 to 10. We have included the details in the Methods section in lines 357-366:

“We simulated the action of the response gain model and the tuning sharpening model using numerical simulations. We composed a neural population of 4⨯105 neurons in equal proportions body-, car-, cat- or house-selective. Each neuron also responded to object categories other than its preferred category, but to a lesser degree and with variation. We chose neural responses to each stimulus from a normal distribution with the mean of 30 spikes/s and standard deviation of 10 and each neuron was randomly assigned an attention factor in the range between 1 and 10 using a uniform distribution. These values are comparable with the values reported in neural studies of attention and object recognition in the ventral visual cortex (Ni et al. 2012, Bao and Tsao 2018). We also added poisson noise to the response of each neuron (Britten et al. 1993), assigned randomly for each condition of each neuron.”

- Eye movements do not seem to be controlled nor measured. Could it be possible that some stimulus pairs result in more discriminable patterns of eye movements? Could this be ruled out by some aspect of the results?

Subjects were instructed to direct their gaze towards the fixation point. Given the variation in the pose and orientation of the stimuli, it is unlikely that eye movements would help with the task. Eye movements have been controlled in previous experiments with individual stimulus presentation (Xu and Vaziri-Pashkam, 2019) and across attentional tasks in which colored dots were superimposed on the stimuli (Vaziri-Pashkam and Xu, 2017) and no significant difference for eye movement across categories or conditions was observed. As such, we do not think that eye movements would play a role in the results we are observing here.

- A central, and untested/verified, assumption is that the multivariate activation pattern associated with 2 overlapping stimuli (with one attended) can be modeled as a weighted combination of the activation pattern associated with the individual stimuli. There are hints in the univariate data (e.g., Fig. 4C; 4D) that this might not be justified, which somewhat calls into question the interpretability of the multivariate results.

If the reviewer is referring to the higher response in the paired compared to the isolated conditions, as explained above, we have not forced any limit on the sum of the estimated weights to equal 1 or 2. Therefore, our model is an estimation of a linear combination of the two multivariate patterns in the isolated conditions. In fact, Leila Reddy et al. (reference 6) reported that while the combination is closer to a weighted average than to a weighted sum, the sum of the weights are on average larger than 1. In Figure 4C and 4D the responses in the paired conditions are higher than either of the isolated-condition responses. This suggests that the weights for the linear combination of isolated responses in the multivariate analysis should add up to larger than one. This is what we find in our results. We have added a supplementary figure to Figure 6, depicting the sum of weights for different category pairs in all ROIs. The figure illustrates that in each ROI, the sum of weights are greater than 1 for some category pairs. It is however noteworthy that we normalized the weights in each condition by the sum of weights to calculate the weight shift in our analysis. The amount of the weight shift was therefore not affected by the absolute value of the weights.

- Throughout the manuscript, the authors consistently refer to "tuning sharpening", an idea that's almost always used to reference changes in the width of tuning curves for specific feature dimensions (e.g., motion direction; hue; orientation; spatial position). Here, the authors are assaying tuning to the category (across exemplars of the category). The link between these concepts could be strengthened to improve the clarity of the manuscript.

The reviewer brings up an excellent point. Whereas tuning curves have been extensively used for feature dimensions such as stimulus orientation or motion direction, here, we used the term to describe the variation in a neuron’s response to different object stimuli.

With a finite set of object categories, as is the case in the current study, the neural response in object space is discrete, rather than a continuous curve illustrated for features such as stimulus orientation. However, since more preferred and less preferred features (objects in this case) can still be defined, we illustrated the neural response using a hypothetical curve in object space in Figure 3 to show how it relates with other stimulus features. Therefore, here, tuning sharpening refers to the fact that the response to the more preferred object categories has been enhanced while the response to the less preferred stimulus categories is suppressed.

We clarify this point in the revised manuscript in the Discussion section lines 649-659:

“While tuning curves are commonly used for feature dimensions such as stimulus orientation or motion direction, here, we used the term to describe the variation in a neuron’s response to different object stimuli. With a finite set of object categories, as is the case in the current study, the neural response in object space is discrete, rather than a continuous curve illustrated for features such as stimulus orientation. The neuron might have tuning for a particular feature such as curvature or spikiness (Bao et al., 2020) that is present to different degrees in our object stimuli in a continuous way, but we are not measuring this directly. Nevertheless, since more preferred and less preferred features (objects in this case) can still be defined, we illustrate the neural response using a hypothetical curve in object space. As such, here, tuning sharpening refers to the fact that the response to the more preferred object categories has been enhanced while the response to the less preferred stimulus categories is suppressed.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

a. The authors should address the apparent paradox noted above (and report whether it is seen in other regions of interest as well). On what model would the response to any pair of stimuli exceed that of the response to the preferred stimulus alone? This implies some kind of Gestalt interaction whereby the combined pair generates a percept that is even more effective for the voxels in question than the "most preferred" one?

The response to a pair of stimuli can exceed the response to each of the stimuli presented in isolation if the voxel is responsive to both stimuli and as long as the voxel has not reached its saturation level. This phenomenon has been reported in many previous studies (Zoccolan et al., 2005, Reddy et al., 2009, Ni et al., 2012, Doostani et al., 2023) and can be modeled using a linear combination model which does not limit the weights of the isolated responses to equal 1 (Doostani et al., 2023). Note that the “most preferred” stimulus does not necessarily saturate the voxel response, thus the response to two stimuli could be more effective based on voxel responsiveness to the second stimulus.

As for the current study, the labels “more preferred” and “less preferred” are only relatively defined (as explained in the Methods section), meaning that the more preferred stimulus is not necessarily the most preferred stimulus for the voxels. Furthermore, the presented stimuli are semi-transparent and presented with low-contrast, which moves the responses further away from the saturation level. Based on reported evidence for multiple-stimulus responses, responses to single stimuli are in many cases sublinearly added to yield the multiple-stimulus response (Zoccolan et al., 2005, Reddy et al., 2009, Doostani et al., 2023). This means that the multiple-stimulus response is lower than the sum of the isolated responses and not lower than each of the isolated responses. Therefore, it is not paradoxical to observe higher responses in paired conditions compared to the isolated conditions. We observe similar results in other ROIs, which we provide as supplementary figures to Figure 4 in the revised manuscript.

We address this observation and similar reports in previous studies in the Results section of the revised manuscript in lines 409-413:

“Note that the response in paired conditions can be higher or lower than the response to the isolated more preferred stimulus (condition Mat), depending on the voxel preference for the two presented stimuli, as previously reported (Doostani et al., 2023). This is consistent with previous studies reporting the response to multiple stimuli to be higher than the average, but lower than the sum of the response to isolated stimuli (Reddy et al., 2009).”

b. Paradox aside, I wondered to what extent the results are in part explained by range limits. Take two categories that evoke a highly similar response (either mean over a full ROI, or in the multivariate sense). That imposes a range limit such that attentional modulation, if it works the way we think it does, could only move responses within that narrow range. In contrast, the starting point for two highly dissimilar categories leaves room in principle for more modulation.

We do not believe that the results can be explained by range limits because responses in paired conditions are not limited by the isolated responses, as can be observed in Figure 4. However, to rule out the possibility of the similarity between responses in isolated conditions affecting the range within which responses in paired conditions can change, we turned to the multivariate analysis. We used the weight shift measure as the change in the weight of each stimulus with the change in the attentional target. In this method, no matter how close the two isolated vectors are, the response to the pair could still have a whole range of different weights of the isolated responses. We have plotted an example illustration of two-dimensional vectors for better clarification. Here, the vectors Vxat and Vyat denote the responses to the isolated x and y stimuli, respectively, and the vector Pxaty denotes the response to the paired condition in which stimulus x is attended. The weights a1 and a2 are illustrated in the figure, which are equal to regression coefficients if we solve the equation Pxaty \= [a1 a2] [x y]’. While the weight values depend on the amplitude of and the angle between the three vectors, they are not limited by a lower angle between Vxat and Vyat.

We have updated Figure 2 in the manuscript to avoid the confusion. We have also added a figure including the sum of weights for different category pairs in different regions, showing that the sum of weights are not dependent on the similarity between the two stimuli. The conclusions based on the weight shift are therefore not confounded by the similarity between the two stimuli.

c. Finally, related to the previous point, while including V1 is a good control, I wonder if it is getting a "fair" test here, because the range of responses to the four categories in this region, in terms of (dis)similarity, seems compressed relative to the other categories.

We believe that V1 is getting a fair test because the single-subject range of category distance in V1 is similar to LO, as can be observed Author response image 1_:_

Author response image 1.

Range of category distance in each ROI averaged across participants

The reason that V1 is showing a more compressed distance range on the average plot is that the category distance in V1 is not consistent among participants. Although the average plots are shown in Figure 5 and Figure 6, we tested statistical significance in each ROI based on single-subject correlation coefficients.

Please also note that a more compressed range of dissimilarity does not necessarily lead to a less strong effect of category distance on the effect of attention. For instance, while LO shows a more compressed dissimilarity range for the presented categories compared to the other object selective regions, it shows the highest correlation between weight shift and category distance. Furthermore, as illustrated in Figure 5, no significant correlation is observed between univariate shift and category distance in V1, even though the range of the univariate distance in V1 is similar to LO and pFs, where we observed a significant correlation between category distance and univariate shift.

d. In general, the manuscript does a very good job explaining the methods of the study in a way that would allow replication. In some places, the authors could be clearer about the reasoning behind those methodological choices. For example: - How was the sample size determined?

Estimating conservatively based on the smallest amount of attentional modulation we observed in a previous study (Doostani et al., 2023), we chose a medium effect size (0.3). For a power of 0.8, the minimum number of participants should be 16. We have added the explanation to the Methods section in lines 78-81:

“We estimated the number of participants conservatively based on the smallest amount of attentional modulation observed in our previous study (Doostani et al., 2023). For a medium effect size of 0.3 and a power of 0.8, we needed a minimum number of 16 participants.”

- Why did the authors choose those four categories? What was the evidence that would suggest these would span the range of similarities needed here?

We chose these four categories based on a previous behavioral study reporting the average reaction time of participants when detecting a target from one category among distractors from another category (Xu and Vaziri-Pashkam, 2019). Ideally the experiment should include as many object categories as possible. However, since we were limited by the duration of the experiment, the number of conditions had to be controlled, leading to a maximum of 4 object categories. We chose two animate and two inanimate object categories to include categories that are more similar and more different based on previous behavioral results (Xu and Vaziri-Pashkam, 2019). We included body and house categories because they are both among the categories to which highly responsive regions exist in the cortex. We chose the two remaining categories based on their similarity to body and house stimuli. In this way, for each category there was another category that elicited similar cortical responses, and two categories that elicited different responses. While we acknowledge that the chosen categories do not fully span the range of similarities, they provide an observable variety of similarities in different ROIs which we find acceptable for the purposes of our study.

We include this information in the Methods section of the revised manuscript in lines 89-94:

“We included body and house categories because there are regions in the brain that are highly responsive and unresponsive to each of these categories, which provided us with a range of responsiveness in the visual cortex. We chose the two remaining categories based on previous behavioral results to include categories that provided us with a range of similarities (Xu and Vaziri-Pashkam, 2019). Thus, for each category there was a range of responsiveness in the brain and a range of similarity with the other categories.”

- Why did the authors present the stimuli at the same location? This procedure has been adopted in previous studies, but of course, it does also move the stimulus situation away from the real-world examples of cluttered scenes that motivate the Introduction.

We presented the stimuli at the same location because we aimed to study the mechanism of object-based attention and this experimental design helped us isolate it from spatial attention. We do not think that our design moves the stimulus situation away from real-world examples in such a way that our results are not generalizable. We include real-world instances, as well as a discussion on this point, in the Discussion section of the revised manuscript, in lines 611-620:

“Although examples of superimposed cluttered stimuli are not very common in everyday life, they still do occur in certain situations, for example reading text on the cellphone screen in the presence of reflection and glare on the screen or looking at the street through a patterned window. Such instances recruit object-based attention which was the aim of this study, whereas in more common cases in which attended and unattended objects occupy different locations in space, both space-based and object-based attention may work together to resolve the competition between different stimuli. Here we chose to move away from usual everyday scenarios to study the effect of object-based attention in isolation. Future studies can reveal the effect of target-distractor similarity, i.e. proximity in space, on space-based attention and how the effects caused by object-based and space-based attention interact.”

- While I'm not concerned about this (all relevant comparisons were within-participants) was there an initial attempt to compare data quality from the two different scanners?

We compared the SNR values of the two groups of participants and observed no significant difference between these values (ps > 0.34, ts < 0.97). We have added this information to the Methods section.

Regarding the observed effect, we performed a t-test between the results of the participants from the two scanners. For the univariate results, the observed correlation between univariate attentional modulation and category distance was not significantly different for participants of the two scanners in any ROIs (ps > 0.07 , ts < 1.9). For the multivariate results, the observed correlation between the weight shift and multivariate category distance was not significantly different in any ROIs (ps > 0.48 , ts < 0.71) except for V1 (p-value = 0.015 , t-value = 2.75).

We include a sentence about the comparison of the SNR values in the preprocessing section in the revised manuscript.

e. There are a couple of analysis steps that could be applied to the existing data that might strengthen the findings. For one, the authors have adopted a liberal criterion of p < 0.001 uncorrected to include voxels within each ROI. Why, and to what extent is the general pattern of findings robust over more selective thresholds? Also, there are additional regions that are selective for bodies (fusiform body area) and scenes (occipital place area and retrosplenial cortex). Including these areas might provide more diversity of selectivity patterns (e.g. different responses to non-preferred categories) that would provide further tests of the hypothesis.

We selected this threshold to allow for selection of a reasonable number of voxels in each hemisphere across all participants. To check whether the effect is robust over more selective thresholds, we exemplarily redefined the left EBA region using p < 0.0001 and p < 0.00001 and observed that the weight shift effect remained equivalent. We have made a note of this analysis in the Results section. As for the additional regions suggested by the reviewer, we chose not to include them because they could not be consistently defined in both hemispheres of all participants. Please note that the current ROIs also show different responses to non-preferred categories (e.g. in LO and pFs). We include this information in the Methods section in lines 206-207:

“We selected this threshold to allow for selection of a reasonable number of voxels in each hemisphere across all participants.”

And in the Results section in lines 509-512:

“We performed the analysis including only voxels that had a significantly positive GLM coefficient across the runs and observed the same results. Moreover, to check whether the effect is robust over more selective thresholds for ROI definition, we redefined the left EBA region with p < 0.0001 and p < 0.00001 criteria. We observed a similar weight shift effect for both criteria.”

f. One point the authors might address is the potential effect of blocking the paired conditions. If I understood right, the irrelevant item in each paired display was from the same category throughout a block. To what extent might this knowledge shape the way participants attend to the task-relevant item (e.g. by highlighting to them certain spatial frequencies or contours that might be useful in making that particular pairwise distinction)? In other words, are there theoretical reasons to expect different effects if the irrelevant category is not predictable?

We believe that the participants’ knowledge about the distractor does not significantly affect our results because our results are in agreement with previous behavioral data (Cohen et al., 2014, Xu and Vaziri-Pashkam, 2019), in which the distractor could not be predicted. These reports suggest there is a theoretical reason to expect similar effects if the participants could not predict the distractor. To directly test this, one would need to perform an fMRI experiment using an event-related design, an interesting venue for future research.

We have made a note of this point in the Discussion section of the revised manuscript in lines 621-626:

“Please note that we used a blocked design in which the target and distractor categories could be predicted across each block. While it is possible that the current design has led to an enhancement of the observed effect, previous behavioral data (Cohen et al., 2014, Xu and Vaziri-Pashkam, 2019) have reported the same effect in experiments in which the distractor was not predictable. To study the effect of predictability on fMRI responses, however, an event-related design is more appropriate, an interesting venue for future fMRI studies.”

g. The authors could provide behavioural data as a function of the specific category pairs. There is a clear prediction here about which pairs should be more or less difficult.

We provide the behavioral data as a supplementary figure to Figure 1 in the revised manuscript. We however do not see differences in behavior for the different category paris. This is so because our fMRI task was designed in a way to make sure the participants could properly attend to the target for all conditions. The task was rather easy across all conditions and due to the ceiling effect, there was no significant difference between behavioral performance for different category pairs. However, the effect of category pair on behavior has been previously tested and reported in a visual search paradigm with the same categories (Xu and Vaziri-Pashkam, 2019), which was in fact the basis for our choice of categories in this study (as explained in response to point “d” above).

h. Figure 4 shows data for EBA in detail; it would be helpful to have a similar presentation of the data for the other ROIs as well.

We provide data for all ROIs as figure supplements 1-4 to Figure 4 in the revised manuscript.

i. For the pFs and LOC ROIs, it would be helpful to have an indication of what proportion of voxels was most/least responsive to each of the four categories. Was this a relatively even balance, or generally favouring one of the categories?

In LO, the proportion of voxels most responsive to each of the four categories was relatively even for Body (31%) and House (32%) stimuli, which was higher than the proportion of Car- and Cat-preferring voxels (18% and 19%, respectively). In pFs, 40% of the voxels were house-selective, while the proportion was relatively even for voxels most responsive to bodies, cars, and houses with 21%, 17%, and 22% of the voxels, respectively. We include the percentage of voxels most responsive to each of the four categories in each ROI as Appendix 1-table 1.

j. Were the stimuli in the localisers the same as in the main experiment?

No, we used different sets of stimuli for the localizers and the main experiment. We have added the information in line 146 of the Methods section.

Reviewer #2 (Recommendations For The Authors):

(1) Why are specific ROIs chosen? Perhaps some discussion motivating these choices, and addressing the possible overlap between these and retinotopic regions (based on other studies, or atlases - Wang et al, 2015) would be useful.

Considering that we used object categories, we decided to look at general object-selective regions (LO, pFS) as well as regions that are highly selective for specific categories (EBA, PPA). We also looked at the primary visual cortex as a control region. We have added this clarification in the Methods section lines 128-133:

“Considering that we used object categories, we investigated five different regions of interest (ROIs): the object-selective areas lateral occipital cortex (LO) and posterior fusiform (pFs) as general object-selective regions, the body-selective extrastriate body area (EBA) and the scene-selective parahippocampal place area (PPA) as regions that are highly selective for specific categories, and the primary visual cortex (V1) as a control region. We chose these regions because they could all be consistently defined in both hemispheres of all participants and included a large number of voxels.”

(2) The authors should consider including data on the relative prevalence of voxels preferring each category for each ROI (and/or the mean activation level across voxels for each category for each ROI). If some ROIs have very few voxels preferring some categories, there's a chance the observed results are a bit noisy when sorting based on those categories (e.g., if a ROI has essentially no response to a given pair of categories, then there's not likely to be much attentional modulation detectable, because the ROI isn't driven by those categories to begin with).

We thank the reviewer for the insightful comment.

We include the percentage of voxels most responsive to each of the four categories in each ROI in the Appendix ( Appendix 1-table 1, please see the answer to point “i” of the first reviewer).

We also provide a table of average activity across voxels for each category in all ROIs as Appendix 1-table 2.

As shown in the table, voxels show positive activity for all categories in all ROIs except for PPA, where voxels show no response to body and cat stimuli. This might explain why we observed a marginally significant correlation between weight shift and category distance in PPA only. As the reviewer mentions, since this region does not respond to body and cat stimuli, we do not observe a significant change in response due to the shift in attention for some pairs. We include the table in the Appendix and add the explanation to the Results section of the revised manuscript in lines 506-508:

_“_Less significant results in PPA might arise from the fact that PPA shows no response to body and cat stimuli and little response to car stimuli (Appendix 1-table 2). Therefore, it is not possible to observe the effect of attention for all category pairs.”

a. Related - would it make sense to screen voxels for inclusion in analysis based on above-basely activation for one or both of the categories? [could, for example, imagine you're accidentally measuring from the motor cortex - you'd be able to perform this analysis, but it would be largely nonsensical because there's no established response to the stimuli in either isolated or combined states].

We performed all the analyses including only voxels that had a significantly positive GLM coefficient across the runs and the results remained the same. We have added the explanation in the Results section in line 509-510.

(3) Behavioral performance is compared against chance level, but it doesn't seem that 50% is chance for the detection task. The authors write on page 4 that the 1-back repetition occurred between 2-3 times per block, so it doesn't seem to be the case that each stimulus had a 50% chance of being a repetition of the previous one.

We apologize for the mistake in our report. We have reported the detection rate for the target-present trials (2-3 per block), not the behavioral performance across all trials. We have modified the sentence in the Results section.

(4) Authors mention that the stimuli are identical for 2-stimulus trials where each category is attended (for a given pair) - but the cue is different, and the cue appears as a centrally-fixated word for 1 s. Is this incorporated into the GLM? I can't imagine this would have much impact, but the strict statement that the goals of the participant are the only thing differentiating trials with otherwise-identical stimuli isn't quite true.

The word cue was not incorporated as a separate predictor into the GLM. As the reviewer notes, the signals related to the cue and stimuli are mixed. But given that the cues are brief and in the form of words rather than images, they are unlikely to have an effect on the response in the regions of interest.

To be more accurate, we have included the clarification in the Methods section in lines 181-182:

“We did not enter the cue to the GLM as a predictor. The obtained voxel-wise coefficients for each condition are thus related to the cue and the stimuli presented in that condition.”

And in the Results section in lines 425-428 :

“It is important to note that since the cue was not separately modeled in the GLM, the signals related to the cue and the stimuli were mixed. However, given that the cues were brief and presented in the form of words, they are unlikely to have an effect on the responses observed in the higher-level ROIs.”

(5) Eq 5: I expected there to be some comparison of a and b directly as ratios (e.g., a_1 > b_1, as shown in Fig. 2). The equations used here should be walked through more carefully - it's very hard to understand what this analysis is actually accomplishing. I'm not sure I follow the explanation of relative weights given by the authors, nor how that maps onto the delta_W quantity in Equation 5.

We provide a direct comparison of a and b, as well as a more thorough clarification of the analysis, in the Methods section in lines 274-276:

“We first projected the paired vector on the plane defined by the isolated vectors (Figure 2A) and then determined the weight of each isolated vector in the projected vector (Figure 2B).”

And in lines 286-297:

“A higher a1 compared to a2 indicates that the paired response pattern is more similar to Vxat compared to Vyat, and vice versa. For instance, if we calculate the weights of the Body and Car stimuli in the paired response related to the simultaneous presentation of both stimuli, we can write in the LO region: VBodyatCar \= 0.81 VBody + 0.31 VCar, VBodyCarat \= 0.43 VBody + 0.68 VCar. Note that these weights are averaged across participants. As can be observed, in the presence of both body and car stimuli, the weight of each stimulus is higher when attended compared to the case when it is unattended. In other words, when attention shifts from body to car stimuli, the weight of the isolated body response (VBody) decreases in the paired response. We can therefore observe that the response in the paired condition is more similar to the isolated body response pattern when body stimuli are attended and more similar to the isolated car response pattern when car stimuli are attended.”

And lines 303-306:

“As shown here, even when body stimuli are attended, the effect of the unattended car stimuli is still present in the response, shown in the weight of the isolated car response (0.31). However, this weight increases when attention shifts towards car stimuli (0.68 in the attended case).”

We also provide more detailed clarification for the 𝛥w and the relative weights in lines 309-324:

“To examine whether this increase in the weight of the attended stimulus was constant or depended on the similarity of the two stimuli in cortical representation, we defined the weight shift as the multivariate effect of attention:

𝛥w = a1/(a1+a2) – b1/(b1+b2)                                                                                          (5)

Here, a1, a2, b1,and b2 are the weights of the isolated responses, estimated using Equation 4. We calculate the weight of the isolated x response once when attention is directed towards x (a1), and a second time when attention is directed towards y (b1). In each case, we calculate the relative weight of the isolated x in the paired response by dividing the weight of the isolated x by the sum of weights of x and y (a1+a2 when attention is directed towards x, and b1+b2 when attention is directed towards y). We then define the weight shift, Δw, as the change in the relative weight of the isolated x response in the paired response when attention shifts from x to y. A higher Δw for a category pair indicates that attention is more efficient in removing the effect of the unattended stimulus in the pair. We used relative weights as a normalized measure to compensate for the difference in the sum of weights for different category pairs. Thus, using the normalized measure, we calculated the share of each stimulus in the paired response. For instance, considering the Body-Car pair, the share of the body stimulus in the paired response was equal to 0.72 and 0.38, when body stimuli were attended and unattended, respectively. We then calculated the change in the share of each stimulus caused by the shift in attention using a simple subtraction ( Equation 5: Δw=0.34 for the above example of the Body-Car pair in LO) and used this measure to compare between different pairs.”

We hope that this clarification makes it easier to understand the multivariate analysis and the weight shift calculation in Equation 5.

We additionally provide the values of the weights (a1, b1, a2, and b2 ) for each category pair averaged across participants as Appendix 1 -table 4.

(6) For multivariate analyses (Fig. 6A-E), x axis is normalized (pattern distance based on Pearson correlation), while the delta_W does not seem to be similarly normalized.

We calculated ΔW by dividing the weights in each condition by the sum of weights in that condition. Thus, we use relative weights which are always in the range of 0 to 1, and ΔW is thus always in the range of -1 to 1. This means that both axes are normalized. Note that even if one axis were not normalized, the relationship between the independent and the dependent variables would remain the same despite the change in the range of the axis.

(7) Simulating additional scenarios like attention to both categories just increasing the mean response would be helpful - is this how one would capture results like those shown in some panels of Fig. 4?

We did not have a condition in which participants were asked to attend to both categories. Therefore it was not useful for our simulations to include such a scenario. Please also note that the goal of our simulations is not to capture the exact amount of attentional modulation, but to investigate the effect of target-distractor similarity on the change in attentional modulation (univariate shift and weight shift).

As for the results in some panels of Figure 4, we have explained the reason underlying higher responses in paired conditions compared to isolated conditions) in response to the “weaknesses” section of the second reviewer. We hope that these points satisfy the reviewer’s concern regarding the results in Figure 4 and our simulations.

(8) Lines 271-276 - the "latter" and "former" are backwards here I think.

We believe that the sentence was correct, but confusing.. We have rephrased the sentence to avoid the confusion in lines 371-376 of the revised manuscript:

“We modeled two neural populations: a general object-selective population in which each voxel shows preference to a particular category and voxels with different preferences are mixed in with each other (similar to LO and pFS), and a category-selective population in which all voxels have a similar preference for a particular category (similar to EBA and PPA).”

(9) Line 314 - "body-car" pair is mentioned twice in describing the non-significant result in PPA ROI.

Thank you for catching the typo. We have changed the second Body-Car to Body-Cat.

(10) Fig. 5 and Fig. 6 - I was expecting to see a plot that demonstrated variability across subjects rather than across category pairs. Would it be possible to show the distribution of each pair's datapoints across subjects, perhaps by coloring all (e.g.) body-car datapoints one color, all body-cat datapoints another, etc? This would also help readers better understand how category preferences (which differ across ROIs) impact the results.

We demonstrated variability across category pairs rather than subjects because we aimed to investigate how the variation in the similarity between categories (i.e. category distance) affected the univariate and multivariate effects of attention. The variability across subjects is reflected in the error bars in the bar plots of Figure 5 and Figure 6.

Here we show the distribution of each category pair’s data points across subjects by using a different color for each pair:

Author response image 2.

Univariate shift versus category distance including single-subject data points in all ROIs.

Author response image 3.

Weight shift versus category distance including single-subject data points in all ROIs.

As can be observed in the figures, category preference has little impact on the results. Rather, the similarity in the preference (in the univariate case) or the response pattern (in the multivariate case) to the two presented categories is what impacts the amount of the univariate shift and the weight shift, respectively. For instance, in EBA we observe a low amount of attentional shift both for the Body-Cat pair, with two stimuli for which the ROI is highly selective, and the Car-House pair, including stimuli to which the region shows little response. A similar pattern is observed in the object-selective regions LO and pFs which show high responses to all stimulus categories.

We believe that the figures including the data points related to all subjects are not strongly informative. However, we agree that using different colors for each category pair helps the readers better understand that category preference has little impact on the results in different ROIs. We therefore present the colored version of Figure 5 and Figure 6 in the revised manuscript, with a different color for each category pair.

(11) Fig. 5 and Fig. 6 use R^2 as a dependent variable across participants to conclude a positive relationship. While the positive relationship is clear in the scatterplots, which depict averages across participants for each category pair, it could still be the case that there are a substantial number of participants with negative (but predictive, thus high positive R^2) slopes. For completeness and transparency, the authors should illustrate the average slope or regression coefficient for each of these analyses.

We concluded the positive relationship and calculated the significance in Figure 5 and Figure 6 using the correlation r rather than r.^2 This is why the result was not significantly positive in V1. We acknowledge that the use of r-squared in the bar plot leads to confusion. We have therefore changed the bar plots to show the correlation coefficient instead of the r-squared. Furthermore, we have added a table of the correlation coefficient for all participants in all ROIs for the univariate and weight shift analyses supplemental to Figure 5 and Figure 6, respectively.

(12) No statement about data or analysis code availability is provided

Thanks for pointing this out. The fMRI data is available on OSF. We have added a statement about it in the Data Availability section of the revised manuscript in line 669.

Author response:

The following is the authors’ response to the original reviews.

Reviewer#1:

Comment #1: It is unclear how the fraction of NK cell populations is quantified in the spatial-seq datasets. Figures display spatial data with expression scores, but the method for calculating the score and determining NK cell presence in tumor tissue is ambiguous. Clarification is needed on whether the identification relied solely on visual inspection or if quantitative analyses using other criteria were conducted.

Thank you for your questions. We removed the background and made the accordingly modifications according to your demand. We used the AddModuleScore function in Seurat to quantify the main immune subpopulations in spatial-seq using the gene sets identified in single-cell-seq. Additionally, the tumor and non-tumor region was identified by immunohistochemistry as well as cell clusters in spatial-seq, it is rough that we can't quantify the NK cell presence in each region precisely. The consolation is that the differences of NK cell presence in tumor and non-tumor region is observable by visual inspection. The methodology has been supplemented in the revised manuscript (line 190-193).

Comment #2: The authors do not provide a clear definition of "resting" NK cells. It remains unclear whether they refer to a senescent state or a non-matured NK cell population. Furthermore, the criteria used to define resting and activated cells based on the expression of KIR2DL4, GPR183, GRP171, CD69, IFNG, GZMK, TTC38, CD160, and PLEKNF1 in Figure 4 are not well-defined. The expression patterns of these genes in Figure 4D are not distinct, and it is unclear which combination of genes was used to classify the populations. Clarification is needed on whether the presence of GZMK alone defines resting NK cells, or if the presence of any of the described genes (GZMK, TTC38, or CD160) is sufficient. Additionally, the method used for this classification, whether visual or algorithm-based, should be described.

Thank you for your question. The resting and activated NK cells was defined by the preferential expression of the described resting genes (AZU, BPI, CAMP, CD160,CD2, CDHR1, CEACAM8, DEFA4, ELANE, GFI1, GZMK, KLRC4, MGAM, MS4A3, NME8, PLEKHF1, TEP1, TRBC1, TTC38, ZNF135) and activated NK genes (APOBEC3G, APOL6, CCL4, CCND2, CD69, CDK6, CSF2, DPP4, FASLG, GPR171, GPR18, GRAP2, IFNG, KIR2DL4, KIR2DS4, LTA, LTB, NCR3, OSM, PTGER2, SOCS1, TNFSF14) in CIBERSORT. Actually, these marker genes were not specifically expressed in a single NK cells subset. On the other hand, combined with further flow cytometric analysis verification, the resting NK cell tend to be a decidual-like NK cells and tumor- infiltrated NK cells with higher expression of CD9, CD49a and PD-1.

Comment #3: Criteria used to define high or low NK cell presence/infiltration in Figure 5 are not described in the main text or figure legend. Since, the claim that the presence of the resting or activated NK cells predicts cancer prognosis is based on this figure, this needs to be clearly described.

Thank you for your questions. The activated and resting NK cell percentage in TCGA and GSE29623 was determined by CIBERSORT. Additionally, the infiltration of activated and resting NK cell was also determined by the AddModuleScore function using the gene sets of activated and resting NK cell identified in single-cell-seq, the differences of activated and resting NK cell presence in tumor and non-tumor region is also determined by visual inspection. We have amended in the main text and figure legend in the revised manuscript.

Comment #4: The absence of FMO controls for KIR2DL4 or GZMK and the lack of increase in GZMK expression during co-culture with tumour lines raises concerns since GZMK was used as a defining feature of resting NK cells.

Thank you for your questions. We did a new batch of flow experiments and FMO controls of all the markers used in the experiments were set up to define the precise positive gate locations.

Author response image 1.

The positive gate locations of CD56, GZMK, KIR2DL4, CD9, CD49a, PD-1 defined according to the FMO control.

Comment #5: All the co-cultures were performed with tumour cell line only and no healthy cells, such as human foreskin fibroblasts, were used as control. In the absence of a non-tumour cell line, it is very difficult to draw any conclusions. Furthermore, to claim that resting or activated NK cells are responsible for tumour migration or proliferation, it is important to at least isolate resting and activated NK cells ex vivo and culture with tumour lines, instead of NK cell lines.

Thank you for your questions. According to your suggestion, NK cells were co-cultured with human foreskin fibroblasts, the phenotype was identified by Flow cytometry. When co-cultured with HFF in direct contact (CN group), NK cells were also tending towards tissue infiltration state (high expression of CD9). However, the domestication effect is significantly reduced compared to co-culturing with tumor cells. Additionally, unlike supernatant of CNS group (NK and HCT were in contact) from NK and HCT co-culture system could significantly increase the migration of fresh HCT, fresh HCT underwent a limited increase (no statistical significance was found) in migration when cultured in the supernatant from the co-culture system in which NK and HFF were in contact (CNS group), but not when co-cultures were performed in the cell supernatant (SNS group) and fresh medium (MNS group). Finally, we tried to isolate resting and activated NK cells from fresh colon cancer surgical specimen. Unfortunately, the NK cells were too few to perform further functional experiments such as migration and proliferation.

Author response image 2.

Phenotype switch of NK cells in different co-cultured system and the corresponding NK cell-mediated effect on cell migration of fresh colon cancer cell (HCT-116). A-B: NK cells underwent phenotype switch (high expression of CD9) when cocultured with HCT and HFF, the phenotype switch was more obvious when co-cultured with HCT. CN: NK cells cocultured with HCT/HFF; SN: NK cells cocultured with supernatant of HCT/HFF; MN: NK cells cocultured in fresh medium. C-E: Transwell assay showed the only tumor co-cultured NK mediated the inductive effect on cell migration of colon cancer cell (HCT-116). CNS: Colon cancer cells were cultured in the supernatant from co-culture system that NK and HCT/HFF were cultured in direct contact; SNS: Colon cancer cells were cultured in the supernatant from co-culture system that NK cocultured with supernatant of HCT/HFF; MNS: Colon cancer cells were cultured in the fresh medium.

Comment #6: It seems that flow cytometric analyses and GZMK and KIR2DL4 staining were performed without cell permeabilization. Could authors confirm if this is accurate, or if they performed intracellular staining instead?

Thank you for your questions. For GZMK, which known as the secretory protein, flow cytometric analyses were performed both with (Fig.3) and without cell fixation and permeabilization, no significant differences were found among each group. The difference is that GZMK was nearly all negative without fixation and permeabilization while it is all positive with fixation and permeabilization. Conditions of flow cytometry analyses for GZMK may need further optimization or GZMK may not be a suitable flow cytometric marker for resting NK cells. On the other hand, for membrane protein such as CD56, CD9, CD49a, KIR2DL4, PD-1, staining was performed without cell permeabilization.

Author response image 3.

Phenotype switch (CD56+, GZMK+) of NK cells was analyzed by FACS after fixation and permeabilization in different co-cultured groups. CN: NK cells cocultured with colon cancer cells; SN: NK cells cocultured with supernatant of cancer cells; MN: NK cells cocultured in fresh medium.

Comment #7: The identity of the published datasets used for analysis is not provided, and references are not cited in the results section.

Thank you for your questions. We are sorry for the neglect of our previous work. We have added the information in the revised manuscript (section of Materials and Methods) (Line 123-128).

Comment #8: References are difficult to locate, as the main text follows APA style while the reference section is organized numerically with no clear order.

Thank you for your questions. We have modified the format of the references in the revised manuscript.

Comment #9: Figure 3 shows volcano plots showing DEG genes between tumor and healthy tissue NK cells are not described clearly, and authors did not discuss the significance of these genes, highlighted in the plot.

Thank you for your questions. Volcano plots of Figure 3 showed the DEGs between colon cancer with metastasis and without metastasis in TCGA database. We focused on the genes which were enriched in the pathway of “Natural killer cell mediated cytotoxicity” and found nearly all the genes enriched in the pathway were down-regulated in the colon cancer with metastasis. We have modified the description in the result section and added the description of importance of these genes in the discussion section in the revise manuscript (Line 322-326).

Comment #10: The meaning of "M0" and "M1" in Figures 5A and 5B is unclear and should be defined in the text.

Thank you for your questions. "M0" and "M1" in Figure 5A and 5B means “colon cancer without metastasis” and “colon cancer with metastasis”, respectively. We have modified in the revise manuscript (Line 350-354).

Comment #11: Terms such as "dynamic remodelling of NK cells" and "landscape of NK cells" are used without explanation, necessitating clarification of their meaning.

Thank you for your questions. We have modified in the revise manuscript (Line 331-334).

Comment #12: In vitro assays are described vaguely, making it difficult for readers to understand. More clarity is needed in describing these assays.

Thank you for your questions. We have added clarification in the revise manuscript (Line 205-211).

Reviewer #2:

Comment #1: This manuscript investigates the role of the abundant NK cells that are observed in colon cancer liver metastasis using sequencing and spatial approaches in an effort to clarify the pro and anti-tumorigenic properties of NK cells. This descriptive study characterises different categories of NK cells in tumor and tumor-adjacent tissues and some correlations. An attempt has been made using pseudotime trajectory analysis but no models around how these NK cells might be regulated are provided.

Thank you for your questions. The single-cell sequencing data enrolled in this study are CD45 positive immune cells and do not involve tumor cells, cellular communication analysis between NK cells and tumor cells cannot be conducted. The change process of NK can only be predicted through pseudotime trajectory analysis. Our hypothesis is that tumor cells domesticate NK cells into a tumor- infiltrated NK cells through direct contact, and flow cytometry experiments have also confirmed that tumor cells can only have such domestication through direct contact with NK cells (with prominent high expression of CD9). However, the detailed mechanism remained unclear.

Comment #2: A small number of patients are analyzed in this study. The descriptive gene markers, while interesting, need to be further validated to understand how strong this analysis might be and its potential application.

Thank you for your questions. The sample size included in this study is indeed a bit small, which is also a limitation of our study. However, this is the only large sample single-cell sequencing dataset could be found that includes primary colon cancer tissues, paired paratumor normal colon tissues, paired liver metastatic cancer tissue, and paired paratumor normal liver tissues. We will expand the sample size to further verify the current conclusion in subsequent experiments. In addition, the marker genes of different NK groups used in this study refer to the CIBERSORT's classification of activated NK cells and resting NK cells, which is a widely recognized indicator. We will verify the expression and clinical application value of the screened genes in tissues in subsequent studies.

Comment #3: Figure 1C and other figures throughout the paper. It is not clear how marker genes were selected.

Thank you for your questions. The marker genes displayed in the Figure.3C were the highly variable genes of each cell group as well as the marker genes of each immune cells, such as T cells (CD3D, CD3E), NK cells (NKG7, KLRD1), monocytes (LYZ, S100A8, S100A9), B cells (CD79A), plasma cells (JCHAIN, IGHA1, IGHA2), Neutrophils (CXCL8, FCGR3B).

Comment #4: Figure 1E. P and T have not been defined. Lines should not connect the datasets as they are independent assessments.

Thank you for your questions. P and T means paratumor normal tissues and tumor tissues, respectively. Which have been added in the caption of Figure 1E. Additionally, the single cell sequencing samples included in the study were paired, with primary colon cancer tissues, paired normal tissues adjacent to colon cancer, paired liver metastatic cancer tissue, and paired normal liver tissues from 20 colon cancer patients with liver metastasis, paired test analysis was thus performed.

Comment #5: Figure 2C. It is unclear what ST-P1 means. This is not a particularly informative figure.

Thank you for your questions. We are sorry that it was our annotation error. Actually, it is the spatial transcriptome of the primary colon cancer tissue and liver metastasis tissue of four patients. We have made the modifications in the revised manuscript.

Comment #6: Multiple figures - abbreviations are used but not provided in the legend. They occur in the text but are not directly related to the figures where they are used to label axes or groups.

Thank you for your questions. We have rechecked and made corresponding modifications in the revised manuscript.

Comment #6: Patients: it is not clear what other drugs patients have been exposed to or basic data (sex, age, underlying conditions etc)

Thank you for your questions. The baseline data of the patient of SC dataset and ST dataset were showed in the Table.1 and Table.2 followed, respectively. They were not presented before as no patients characteristics related analysis was performed in the current study.

Author response table 1.

The baseline data of patient from single cell sequencing database.

Author response table 2.

The baseline data of patient from spatial transcriptome database.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript, the authors investigated the dynamics of a neural network model characterized by sparsely connected clusters of neuronal ensembles. They found that such a network could intrinsically generate sequence preplay and place maps, with properties like those observed in the real-world data. Strengths of the study include the computational model and data analysis supporting the hippocampal network mechanisms underlying sequence preplay of future experiences and place maps.

Previous models of replay or theta sequences focused on circuit plasticity and usually required a pre-existing place map input from the external environment via upstream structures. However, those models failed to explain how networks support rapid sequential coding of novel environments or simply transferred the question to the upstream structure. On the contrary, the current proposed model required minimal spatial inputs and was aimed at elucidating how a preconfigured structure gave rise to preplay, thereby facilitating the sequential encoding of future novel environments.

In this model, the fundamental units for spatial representation were clusters within the network. Sequential representation was achieved through the balance of cluster isolation and their partial overlap. Isolation resulted in a self-reinforced assembly representation, ensuring stable spatial coding. On the other hand, overlap-induced activation transitions across clusters, enabling sequential coding.

This study is important when considering that previous models mainly focused on plasticity and experience-related learning, while this model provided us with insights into how network architecture could support rapid sequential coding with large capacity, upon which learning could occur efficiently with modest modification via plasticity.

I found this research very inspiring and, below, I provide some comments aimed at improving the manuscript. Some of these comments may extend beyond the scope of the current study, but I believe they raise important questions that should be addressed in this line of research.

(1) The expression 'randomly clustered networks' needs to be explained in more detail given that in its current form risks to indicate that the network might be randomly organized (i.e., not organized). In particular, a clustered network with future functionality based on its current clustering is not random but rather pre-configured into those clusters. What the authors likely meant to say, while using the said expression in the title and text, is that clustering is not induced by an experience in the environment, which will only be later mapped using those clusters. While this organization might indeed appear as randomly clustered when referenced to a future novel experience, it might be non-random when referenced to the prior (unaccounted) activity of the network. Related to this, network organization based on similar yet distinct experiences (e.g., on parallel linear tracks as in Liu, Sibille, Dragoi, Neuron 2021) could explain/configure, in part, the hippocampal CA1 network organization that would appear otherwise 'randomly clustered' when referenced to a future novel experience.

As suggested by the reviewer, we have revised the text to clarify that the random clustering is random with respect to any future, novel environment (lines 111-114 and 710-712).

Lines 111-114: “To reconcile these experimental results, we propose a model of intrinsic sequence generation based on randomly clustered recurrent connectivity, wherein place cells are connected within multiple overlapping clusters that are random with respect to any future, novel environment.”

Lines 710-712: “Our results suggest that the preexisting hippocampal dynamics supporting preplay may reflect general properties arising from randomly clustered connectivity, where the randomness is with respect to any future, novel experience.”

The cause of clustering could be prior experiences (e.g. Bourjaily and Miller, 2011) or developmental programming (e.g. Perin et al., 2011; Druckmann et al., 2014; Huszar et al., 2022), and we have modified lines 116 and 714-718 to state this.

Lines 116: Added citation of “Perin et al., 2011”

Lines 714-718: “Synaptic plasticity in the recurrent connections of CA3 may primarily serve to reinforce and stabilize intrinsic dynamics, which could be established through a combination of developmental programming (Perin et al., 2011; Druckmann et al., 2014; Huszar et al., 2022) and past experiences (Bourjaily and Miller, 2011), rather than creating spatial maps de novo.”

We thank the reviewer for suggesting that the results of Liu et al., 2021 strengthen the support for our modeling motivations. We agree, and we now cite their finding that the hippocampal representations of novel environments emerged rapidly but were initially generic and showed greater discriminability from other environments with repeated experience in the environment (lines 130-134).

Lines 130-134: “Further, such preexisting clusters may help explain the correlations that have been found in otherwise seemingly random remapping (Kinsky et al., 2018; Whittington et al., 2020) and support the rapid hippocampal representations of novel environments that are initially generic and become refined with experience (Liu et al., 2021).”

(2) The authors should elaborate more on how the said 'randomly clustered networks' generate beyond chance-level preplay. Specifically, why was there preplay stronger than the time-bin shuffle? There are at least two potential explanations:

(1) When the activation of clusters lasts for several decoding time bins, temporal shuffle breaks the continuity of one cluster's activation, thus leading to less sequential decoding results. In that case, the preplay might mainly outperform the shuffle when there are fewer clusters activating in a PBE. For example, activation of two clusters must be sequential (either A to B or B to A), while time bin shuffle could lead to non-sequential activations such as a-b-a-b-a-b where a and b are components of A and B;

(2) There is a preferred connection between clusters based on the size of overlap across clusters. For example, if pair A-B and B-C have stronger overlap than A-C, then cluster sequences A-B-C and C-B-A are more likely to occur than others (such as A-C-B) across brain states. In that case, authors should present the distribution of overlap across clusters, and whether the sequences during run and sleep match the magnitude of overlap. During run simulation in the model, as clusters randomly receive a weak location cue bias, the activation sequence might not exactly match the overlap of clusters due to the external drive. In that case, the strength of location cue bias (4% in the current setup) could change the balance between the internal drive and external drive of the representation. How does that parameter influence the preplay incidence or quality?

Explanation 1 is correct: Our cluster-activation analyses (Figure 5) showed that the parameter values that generate preplay correspond to the parameter regions that support sustained cluster activity over multiple decoding time bins, which led us to the conclusion of the reviewer’s first proposed explanation.

We have now added additional analyses supporting the conclusion that cluster-wise activity is the main driver of preplay rather than individual cell-identity (Figures 6 and 7). In Figure 6 we show that cluster-identity alone is sufficient to produce significant preplay by performing decoding after shuffling cell identity within clusters, and in Figure 7 we show that this result holds true when considering the sequence of spiking activity within population bursts rather than the spatial decoding.

Lines 495-515: The pattern of preplay significance across the parameter grid in Figure 4f shows that preplay only occurs with modest cluster overlap, and the results of Figure 5 show that this corresponds to the parameter region that supports transient, isolated cluster-activation. This raises the question of whether cluster-identity is sufficient to explain preplay. To test this, we took the sleep simulation population burst events from the fiducial parameter set and performed decoding after shuffling cell identity in three different ways. We found that when the identity of all cells within a network are randomly permuted the resulting median preplay correlation shift is centered about zero (t-test 95% confidence interval, -0.2018 to 0.0012) and preplay is not significant (distribution of p-values is consistent with a uniform distribution over 0 to 1, chi-square goodness-of-fit test p=0.4436, chi-square statistic=2.68; Figure 6a). However, performing decoding after randomly shuffling cell identity between cells that share membership in a cluster does result in statistically significant preplay for all shuffle replicates, although the magnitude of the median correlation shift is reduced for all shuffle replicates (Figure 6b). The shuffle in Figure 6b does not fully preserve cell’s cluster identity because a cell that is in multiple clusters may be shuffled with a cell in either a single cluster or with a cell in multiple clusters that are not identical. Performing decoding after doing within-cluster shuffling of only cells that are in a single cluster results in preplay statistics that are not statistically different from the unshuffled statistics (t-test relative to median shift of un-shuffled decoding, p=0.1724, 95% confidence interval of -0.0028 to 0.0150 relative to the reference value; Figure 6c). Together these results demonstrate that cluster-identity is sufficient to produce preplay.

Lines 531-551: While cluster-identity is sufficient to produce preplay (Figure 6b), the shuffle of Figure 6c is incomplete in that cells belonging to more than one cluster are not shuffled. Together, these two shuffles leave room for the possibility that individual cell-identity may contribute to the production of preplay. It might be the case that some cells fire earlier than others, both on the track and within events. To test the contribution of individual cells to preplay, we calculated for all cells in all networks of the fiducial parameter point their mean relative spike rank and tested if this is correlated with the location of their mean place field density on the track (Figure 7). We find that there is no relationship between a cell’s mean relative within-event spike rank and its mean place field density on the track (Figure 7a). This is the case when the relative rank is calculated over the entire network (Figure 7, “Within-network”) and when the relative rank is calculated only with respect to cells with the same cluster membership (Figure 7, “Within-cluster”). However, because preplay events can proceed in either track direction, averaging over all events would average out the sequence order of these two opposite directions. We performed the same correlation but after reversing the spike order for events with a negative slope in the decoded trajectory (Figure 7b). To test the significance of this correlation, we performed a bootstrap significance test by comparing the slope of the linear regression to the slope that results when performing the same analysis after shuffling cell identities in the same manner as in Figure 6. We found that the linear regression slope is greater than expected relative to all three shuffling methods for both the within-network mean relative rank correlation (Figure 6c) and the within-cluster mean relative rank correlation (Figure 6d).

Lines 980-1000:

“Cell identity shuffled decoding

We performed Bayesian decoding on the fiducial parameter set after shuffling cell identities in three different manners (Figures 6 and 7). To shuffle cells in a cluster-independent manner (“Across-network shuffle”), we randomly shuffled the identity of cells during the sleep simulations. To shuffle cells within clusters (“Within-cluster shuffle”), we randomly shuffled cell identity only between cells that shared membership in at least one cluster. To shuffle cells within only single clusters (“Within-single-cluster shuffle”), we shuffled cells in the same manner as the within-cluster shuffle but excluded any cells from the shuffle that were in multiple clusters.

To test for a correlation between spike rank during sleep PBEs and the order of place fields on the track (Figure 7), we calculated for each excitatory cell in each network of the fiducial parameter set its mean relative spike rank and correlated that with the location of its mean place field density on the track (Figure 7a). To account for event directionality, we calculated the mean relative rank after inverting the rank within events that had a negatively sloped decoded trajectory (Figure 7b). We calculated mean relative rank for each cell relative to all cells in the network (“Within-network mean relative rank”) and relative to only cells that shared cluster membership with the cell (“Within-cluster mean relative rank”). We then compared the slope of the linear regression between mean relative rank and place field location against the slope that results when applying the same analysis to each of the three methods of cell identify shuffles for both the within-network regression (Figure 7c) and the within-cluster regression (Figure 7d).”

We also now show that the sequence of cluster-activation in events with 3 active clusters does not match the sequence of cluster biases on the track above chance levels and that events with fewer active clusters have the largest increase in median weighted decode correlation (Figure 5—figure supplement 1), showing that the reviewer’s second explanation is not the case.

Lines 466-477: “The results of Figure 5 suggest that cluster-wise activation may be crucial to preplay. One possibility is that the random overlap of clusters in the network spontaneously produces biases in sequences of cluster activation which can be mapped onto any given environment. To test this, we looked at the pattern of cluster activations within events. We found that sequences of three active clusters were not more likely to match the track sequence than chance (Figure 5—figure supplement 1a). This suggests that preplay is not dependent on a particular biased pattern in the sequence of cluster activation. We then we asked if the number of clusters that were active influenced preplay quality. We split the preplay events by the number of clusters that were active during each event and found that the median preplay shift relative to shuffled events with the same number of active clusters decreased with the number of active clusters (Spearman’s rank correlation, p=0.0019, =-0.13; Figure 5—figure supplement 1b).”

Lines 1025-1044:

“Active cluster analysis

To quantify cluster activation (figure 5), we calculated the population rate for each cluster individually as the mean firing rate of all excitatory cells belonging to the cluster smoothed with a Gaussian kernel (15 ms standard deviation). A cluster was defined as ‘active’ if at any point its population rate exceeded twice that of any other cluster during a PBE. The active clusters’ duration of activation was defined as the duration for which it was the most active cluster.

To test whether the sequence of activation in events with three active clusters matched the sequence of place fields on the track, we performed a bootstrap significance test (Figure 5—figure supplement 1). For all events from the fiducial parameter set that had three active clusters, we calculated the fraction in which the sequence of the active clusters matched the sequence of the clusters’ left vs right bias on the track in either direction. We then compared this fraction to the distribution expected from randomly sampling sequences of three clusters without replacement.

To determine if there was a relationship between the number of active clusters within an event and it’s preplay quality we performed a Spearman’s rank correlation between the number of active clusters and the normalized absolute weighted correlation across all events at the fiducial parameter set. The absolute weighted correlations were z-scored based on the absolute weighted correlations of the time-bin shuffled events that had the same number of active clusters.”

We also now add control simulations showing that without the cluster-dependent bias the population burst events no longer significantly decode as preplay (Figure 4—figure supplement 4e).

(3) The manuscript is focused on presenting that a randomly clustered network can generate preplay and place maps with properties similar to experimental observations. An equally interesting question is how preplay supports spatial coding. If preplay is an intrinsic dynamic feature of this network, then it would be good to study whether this network outperforms other networks (randomly connected or ring lattice) in terms of spatial coding (encoding speed, encoding capacity, tuning stability, tuning quality, etc.)

We agree that this is an interesting future direction, but we see it as outside the scope of the current work. There are two interesting avenues of future work: 1) Our current model does not include any plasticity mechanisms, but a future model could study the effects of synaptic plasticity during preplay on long-term network dynamics, and 2) Our current model does not include alternative approaches to constructing the recurrent network, but future studies could systematically compare the spatial coding properties of alternative types of recurrent networks.

(4) The manuscript mentions the small-world connectivity several times, but the concept still appears too abstract and how the small-world index (SWI) contributes to place fields or preplay is not sufficiently discussed.

For a more general audience in the field of neuroscience, it would be helpful to include example graphs with high and low SWI. For example, you can show a ring lattice graph and indicate that there are long paths between points at opposite sides of the ring; show randomly connected graphs indicating there are no local clustered structures, and show clustered graphs with several hubs establishing long-range connections to reduce pair-wise distance.

How this SWI contributes to preplay is also not clear. Figure 6 showed preplay is correlated with SWI, but maybe the correlation is caused by both of them being correlated with cluster participation. The balance between cluster overlap and cluster isolation is well discussed. In the Discussion, the authors mention "...Such a balance in cluster overlap produces networks with small-world characteristics (Watts and Strogatz, 1998) as quantified by a small-world index..." (Lines 560-561). I believe the statement is not entirely appropriate, a network similar to ring lattice can still have the balance of cluster isolation and cluster overlap, while it will have small SWI due to a long path across some node pairs. Both cluster structure and long-range connection could contribute to SWI. The authors only discuss the necessity of cluster structure, but why is the long-range connection important should also be discussed. I guess long-range connection could make the network more flexible (clusters are closer to each other) and thus increase the potential repertoire.

We agree that the manuscript would benefit from a more concrete explanation of the small-world index. We have added a figure illustrating different types of networks and their corresponding SWI (Figure 1—figure supplement 1) and a corresponding description in the main text (lines 228-234).

Lines 228-234: “A ring lattice network (Figure 1—figure supplement 1a) exhibits high clustering but long path lengths between nodes on opposite sides of the ring. In contrast, a randomly connected network (Figure 1—figure supplement 1c) has short path lengths but lacks local clustered structure. A network with small world structure, such as a Watts-Strogatz network (Watts and Strogatz, 1998) or our randomly clustered model (Figure 1—figure supplement 1b), combines both clustered connectivity and short path lengths. In our clustered networks, for a fixed connection probability the SWI increases with more clusters and lower cluster participation…”

We note that while our most successful clustered networks are indeed those with small-world characteristics, there are other ways of producing small-world networks which may not show good place fields or preplay. We have modified lines 690-692 to clarify that that statement is specific to our model.

Lines 690-692: “In our clustered network structure, such a balance in cluster overlap produces networks with small-world characteristics (Watts and Strogatz, 1998) as quantified by a small-world index (SWI, Figure 1g; Neal, 2015; Neal, 2017).”

(5) What drives PBE during sleep? Seems like the main difference between sleep and run states is the magnitude of excitatory and inhibitory inputs controlled by scaling factors. If there are bursts (PBE) in sleep, do you also observe those during run? Does the network automatically generate PBE in a regime of strong excitation and weak inhibition (neural bifurcation)?

During sleep simulations, the PBEs are spontaneously generated by the recurrent connections in the network. The constant-rate Poisson inputs drive low-rate stochastic spiking in the recurrent network, which then randomly generates population events when there is sufficient internal activity to transiently drive additional spiking within the network.

During run simulations, the spatially-tuned inputs drive greater activity in a subset of the cells at a given point on the track, which in turn suppress the other excitatory cells through the feedback inhibition.

We have added a brief explanation of this in the text in lines 281-284.

Lines 281-284: “During simulated sleep, sparse, stochastic spiking spontaneously generates sufficient excitement within the recurrent network to produce population burst events resembling preplay (Figure 2d-f)”

(6) Is the concept of 'cluster' similar to 'assemblies', as in Peyrache et al, 2010; Farooq et al, 2019? Does a classic assembly analysis during run reveal cluster structures?

Our clusters correspond to functional assemblies in that cells that share a cluster membership have more-similar place fields and are more likely to reactivate together during population burst events. In the figure to the right, we show for an example network at the fiducial parameter set the Pearson correlation between all pairs of place fields split by whether the cells share membership in a cluster (blue) or do not (red).

Author response image 1.

We expect an assembly analysis would identify assemblies similarly to the experimental data, but we see this additional analysis as a future direction. We have added a description of this correspondence in the text at lines 134-137.

Lines 134-137: “Such clustered connectivity likely underlies the functional assemblies that have been observed in hippocampus, wherein groups of recorded cells have correlated activity that can be identified through independent component analysis (Peyrache et al., 2010; Farooq et al., 2019).”

(7) Can the capacity of the clustered network to express preplay for multiple distinct future experiences be estimated in relation to current network activity, as in Dragoi and Tonegawa, PNAS 2013?

We agree this is an interesting opportunity to compare the results of our model to what has been previously found experimentally. We report here preliminary results supporting this as an interesting future direction.

Author response image 2.

We performed a similar analysis to that reported in Figure 3C of Dragoi and Tonegawa, 2013. We determined the statistical significance of each event individually for each of the two environments by testing whether the decoded event’s absolute weighted correlation exceeded that 99th percentile of the corresponding shuffle events. We then fit a linear regression to the fraction of events that were significant for each of the two tracks and that were significant to either of the two tracks (left panel of above figure). We then estimated the track capacity as the number of tracks at the point where the linear regression reached 100% of the network capacity. We find that applying this analysis to our fiducial parameter set returns an estimate of ~8.6 tracks (Dragoi and Tonegawa, 2013, found ~15 tracks).

We performed this same analysis for each parameter point in our main parameter grid (right panel of above figure). The parameter region that produces significant preplay (Figure 4f) corresponds to the region that has a track capacity of approximately 8-25 tracks. In the parameter grid region that does not produce preplay, the estimated track capacity approaches the high values that this analysis would produce when applied to events that are significant only at the false-positive rate. This analysis is based on the assumption that each preplay event would significantly correspond to at least one future event. Interesting interpretation issues arise when applying this analysis to parameter regions that do not produce statistically significant preplay, which we leave to future directions to address.

We note two differences between our analysis here and that in Dragoi and Tonegawa, 2013. First, their track capacity analysis was performed on spike sequences rather than decoded spatial sequences, which is the focus of our manuscript. Second, they recorded rats exploring three novel tracks, while in our manuscript we only simulated two novel tracks, which reduces the accuracy of our linear extrapolation of track capacity.

Reviewer #2 (Public Review):

Summary:

The authors show that a spiking network model with clustered neurons produces intrinsic spike sequences when driven with a ramping input, which are recapitulated in the absence of input. This behavior is only seen for some network parameters (neuron cluster participation and number of clusters in the network), which correspond to those that produce a small world network. By changing the strength of ramping input to each network cluster, the network can show different sequences.

Strengths:

A strength of the paper is the direct comparison between the properties of the model and neural data.

Weaknesses:

My main critiques of the paper relate to the form of the input to the network.

First, because the input is the same across trials (i.e. all traversals are the same duration/velocity), there is no ability to distinguish a representation of space from a representation of time elapsed since the beginning of the trial. The authors should test what happens e.g. with traversals in which the animal travels at different speeds, and in which the animal's speed is not constant across the entire track, and then confirm that the resulting tuning curves are a better representation of position or duration.

We thank the reviewer for pointing out this important limitation. We see extensive testing of the time vs space coding properties of this network as a future direction, but we have performed simulations that demonstrate the robustness of place field coding to variations in traversal speeds and added the results as a supplemental figure (Figure 3—figure supplement 1).

Lines 332-336: “To verify that our simulated place cells were more strongly coding for spatial location than for elapsed time, we performed simulations with additional track traversals at different speeds and compared the resulting place fields and time fields in the same cells. We find that there is significantly greater place information than time information (Figure 3—figure supplement 1).

Lines 835-841: “To compare coding for place vs time, we performed repeated simulations for the same networks at the fiducial parameter point with 1.0x and 2.0x of the original track traversal speed. We then combined all trials for both speed conditions to calculate both place fields and time fields for each cell from the same linear track traversal simulations. The place fields were calculated as described below (average firing rate within each of the fifty 2-cm long spatial bins across the track) and the time fields were similarly calculated but for fifty 40-ms time bins across the initial two seconds of all track traversals.”

Second, it's unclear how much the results depend on the choice of a one-dimensional environment with ramping input. While this is an elegant idealization that allows the authors to explore the representation and replay properties of their model, it is a strong and highly non-physiological constraint. The authors should verify that their results do not depend on this idealization. Specifically, I would suggest the authors also test the spatial coding properties of their network in 2-dimensional environments, and with different kinds of input that have a range of degrees of spatial tuning and physiological plausibility. A method for systematically producing input with varying degrees of spatial tuning in both 1D and 2D environments has been previously used in (Fang et al 2023, eLife, see Figures 4 and 5), which could be readily adapted for the current study; and behaviorally plausible trajectories in 2D can be produced using the RatInABox package (George et al 2022, bioRxiv), which can also generate e.g. grid cell-like activity that could be used as physiologically plausible input to the network.

We agree that testing the robustness of our results to variations in feedforward input is important. We have added new simulation results (Figure 4—figure supplement 4) showing that the existence of preplay in our model is robust to variations in the form of input.

Testing the model in a 2D environment is an interesting future direction, but we see it as outside the scope of the current work. To our knowledge there are no experimental findings of preplay in 2D environments, but this presents an interesting opportunity for future modeling studies.

Lines 413-420: To test the robustness of our results to variations in input types, we simulated alternative forms of spatially modulated feedforward inputs. We found that with no parameter tuning or further modifications to the network, the model generates robust preplay with variations on the spatial inputs, including inputs of three linearly varying cues (Figure 4—figure supplement 4a) and two stepped cues (Figure 4—figure supplement 4b-c). The network is impaired in its ability to produce preplay with binary step location cues (Figure 4—figure supplement 4d), when there is no cluster bias (Figure 4—figure supplement 4e), and at greater values of cluster participation (Figure 4—figure supplement 4f).

Finally, I was left wondering how the cells' spatial tuning relates to their cluster membership, and how the capacity of the network (number of different environments/locations that can be represented) relates to the number of clusters. It seems that if clusters of cells tend to code for nearby locations in the environment (as predicted by the results of Figure 5), then the number of encodable locations would be limited (by the number of clusters). Further, there should be a strong tendency for cells in the same cluster to encode overlapping locations in different environments, which is not seen in experimental data.

Thank you for making this important point and giving us the opportunity to clarify. We do find that subsets of cells with identical cluster membership have correlated place fields, but as we show in Figure 9b (original Figure 7b) the network place map as a whole shows low remapping correlations across environments, which is consistent with experimental data (Hampson et al., 1996; Pavlides, et al., 2019).

Our model includes a relatively small number of cells and clusters compared to CA3, and with a more realistic number of clusters, the level of correlation across network place maps should reduce even further in our model network. The reason for a low level of correlation in the model is because cluster membership is combinatorial, whereby cells that share membership in one cluster can also belong to separate/distinct other clusters, rendering their activity less correlated than might be anticipated.

We have added text at lines 627-630 clarifying these points.

Lines 628-631: “Cells that share membership in a cluster will have some amount of correlation in their remapping due to the cluster-dependent cue bias, which is consistent with experimental results (Hampson et al., 1996; Pavlides et al., 2019), but the combinatorial nature of cluster membership renders the overall place field map correlations low (Figure 9b).”

Reviewer #3 (Public Review):

Summary:

This work offers a novel perspective on the question of how hippocampal networks can adaptively generate different spatial maps and replays/preplays of the corresponding place cells, without any such maps pre-existing in the network architecture or its inputs. Unlike previous modeling attempts, the authors do not pre-tune their model neurons to any particular place fields. Instead, they build a random, moderately-clustered network of excitatory (and some inhibitory) cells, similar to CA3 architecture. By simulating spatial exploration through border-cell-like synaptic inputs, the model generates place cells for different "environments" without the need to reconfigure its synaptic connectivity or introduce plasticity. By simulating sleep-like random synaptic inputs, the model generates sequential activations of cells, mimicking preplays. These "preplays" require small-world connectivity, so that weakly connected cell clusters are activated in sequence. Using a set of electrophysiological recordings from CA1, the authors confirm that the modeled place cells and replays share many features with real ones. In summary, the model demonstrates that spontaneous activity within a small-world structured network can generate place cells and replays without the need for pre-configured maps.

Strengths:

This work addresses an important question in hippocampal dynamics. Namely, how can hippocampal networks quickly generate new place cells when a novel environment is introduced? And how can these place cells preplay their sequences even before the environment is experienced? Previous models required pre-existing spatial representations to be artificially introduced, limiting their adaptability to new environments. Other models depended on synaptic plasticity rules which made remapping slower than what is seen in recordings. This modeling work proposes that quickly-adaptive intrinsic spiking sequences (preplays) and spatially tuned spiking (place cells) can be generated in a network through randomly clustered recurrent connectivity and border-cell inputs, avoiding the need for pre-set spatial maps or plasticity rules. The proposal that small-world architecture is key for place cells and preplays to adapt to new spatial environments is novel and of potential interest to the computational and experimental community.

The authors do a good job of thoroughly examining some of the features of their model, with a strong focus on excitatory cell connectivity. Perhaps the most valuable conclusion is that replays require the successive activation of different cell clusters. Small-world architecture is the optimal regime for such a controlled succession of activated clusters.

The use of pre-existing electrophysiological data adds particular value to the model. The authors convincingly show that the simulated place cells and preplay events share many important features with those recorded in CA1 (though CA3 ones are similar).

Weaknesses:

To generate place cell-like activity during a simulated traversal of a linear environment, the authors drive the network with a combination of linearly increasing/decreasing synaptic inputs, mimicking border cell-like inputs. These inputs presumably stem from the entorhinal cortex (though this is not discussed). The authors do not explore how the model would behave when these inputs are replaced by or combined with grid cell inputs which would be more physiologically realistic.

We chose the linearly varying spatial inputs as the minimal model of providing spatial input to the network so that we could focus on the dynamics of the recurrent connections. We agree our results will be strengthened by testing alternative types of border-like input. We show in Figure 4—figure supplement 4that our preplay results are robust to several variations in the location-cue inputs. However, given that a sub-goal of our model was to show that place fields could arise in locations at which no neurons receive a peak in external input, whereas combining input from multiple grid cells produces peaked place-field like input, adding grid cell input (and the many other types of potential hippocampal input) is beyond the scope of the paper.

Even though the authors claim that no spatially-tuned information is needed for the model to generate place cells, there is a small location-cue bias added to the cells, depending on the cluster(s) they belong to. Even though this input is relatively weak, it could potentially be driving the sequential activation of clusters and therefore the preplays and place cells. In that case, the claim for non-spatially tuned inputs seems weak. This detail is hidden in the Methods section and not discussed further. How does the model behave without this added bias input?

We apologize for a lack of clarity if we have caused confusion about the type of inputs and if we implied an absence of spatially-tuned information in the network. In order for place fields to appear the network must receive spatial information, which we model as linearly-varying cues and illustrate in Figure 1b and describe in the caption (original lines 156-157), Results (original lines 189-190 & 497-499), and Methods (original lines 671-683). Such input is not place-field like, as the small bias to any cell linearly decreases from one boundary of the track or the other.

The cluster-dependent bias, which is also described in the same lines (Figure 1 caption (original lines 156-157), Results (original lines 189-190 & 497-499), and Methods (original lines 671-683)), only affects the strength of the spatial cues that are present during simulated run periods. Crucially, this cluster-dependent bias is absent during sleep simulations when preplay occurs, which is why preplay can equally correlate with place field sequences in any context.

We have modified the text (lines 207-210, 218, and 824-827) to clarify these points. We have also added results from a control simulation (Figure 4—figure supplement 4e) showing that preplay is not generated in the absence of the cluster-dependent bias.

Lines 207-210: “This bias causes cells that share cluster memberships to have more similar place fields during the simulated run period, but, crucially, this bias is not present during sleep simulations so that there is no environment-specific information present when the network generates preplay.”

Lines 218: “Second, to incorporate cluster-dependent correlations in place fields, a small…”

Lines 824-827: “The addition of this bias produced correlations in cells’ spatial tunings based on cluster membership, but, importantly, this bias was not present during the sleep simulations, and it did not lead to high correlations of place-field maps between environments (Figure 9b).”

Unlike excitation, inhibition is modeled in a very uniform way (uniform connection probability with all E cells, no I-I connections, no border-cell inputs). This goes against a long literature on the precise coordination of multiple inhibitory subnetworks, with different interneuron subtypes playing different roles (e.g. output-suppressing perisomatic inhibition vs input-gating dendritic inhibition). Even though no model is meant to capture every detail of a real neuronal circuit, expanding on the role of inhibition in this clustered architecture would greatly strengthen this work.

This is an interesting future direction, but we see it as outside the scope of our current work. While inhibitory microcircuits are certainly important physiologically, we focus here on a minimal model that produces the desired place cell activity and preplay, as measured in excitatory cells. We have added a brief discussion of this to the manuscript.

Lines 733-739: “Additionally, the in vivo microcircuitry of CA3 is complex and includes aspects such as nonlinear dendritic computations and a variety of inhibitory cell types (Rebola et al., 2017). This microcircuitry is crucial for explaining certain aspects of hippocampal function, such as ripple and gamma oscillogenesis (Ramirez-Villegas et al., 2017), but here we have focused on a minimal model that is sufficient to produce place cell spiking activity that is consistent with experimentally measured place field and preplay statistics.”

For the modeling insights to be physiologically plausible, it is important to show that CA3 connectivity (which the model mimics) shares the proposed small-world architecture. The authors discuss the existence of this architecture in various brain regions but not in CA3, which is traditionally thought of and modeled as a random or fully connected recurrent excitatory network. A thorough discussion of CA3 connectivity would strengthen this work.

We agree this is an important point that is missing, and we have modified lines 114-116 to address the clustered connectivity reported in CA3.

Lines 114-116: “Such clustering is a common motif across the brain, including the CA3 region of the hippocampus (Guzman et al., 2016) as well as cortex (Song et al., 2005), …”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) Based on Figure 3, the place fields are not uniformly distributed in the maze. Meanwhile, based on Figure 1b and Methods, the total input seems to be uniform across the maze. Why does the uniform total external input lead to nonuniform network activities?

While the total input to the network is constant across the maze, the input to any individual cell can peak only at either end of the track. All excitatory cells receive input from both the left-cue and the right-cue with different input strengths. By chance and due to the cluster-dependent bias some cells will have stronger input from one cue than the other and will therefore be more likely to have a place field toward that side of the track. However, no cell receives a peak of input in the center of the track. We have modified lines 141-143 to clarify this.

Lines 141-143: “While the total input to the network is constant as a function of position, each cell only receives a peak in its spatially linearly varying feedforward input at one end of the track.”

(2) I find these sentences confusing: "...we expected that the set of spiking events that significantly decode to linear trajectories in one environment (Figure 4) should decode with a similar fidelity in another environment..." (Lines 513-515) and "As expected... but not with the place fields of trajectories from different environments (Figure 7c)" (Line 517-520). What is the expectation for cross-environment decoding? Should they be similar or different? Also, in Figure 7c, the example is not fully convincing. In the figure caption, it states that decoding is significant in the top row but not in the bottom row, but they look similar across rows.

Original lines 513-515 refer to the entire set of events, while original lines 517-520 refer to one example event. The sleep events are simulated without any track-specific information present, so the degree to which preplay occurs when decoding based on the place fields of a specific future track should be independent of any particular track when considering the entire set of decoded PBEs, as shown in Figure 9d (original Figure 7). However, because there is strong remapping across tracks (Figure 9b), an individual event that shows a strong decoded trajectory based on the place fields of one track (Figure 9c, top row) should show chance levels of a decoded trajectory when decoded with the place fields of an alternative track (Figure 9c, bottom row).

We have revised lines 643-650 for clarity, and we have added statistics for the events shown in Figure 9c.

Lines 644-651: “Since the place field map correlations are high for trajectories on the same track and near zero for trajectories on different tracks, any individual event would be expected to have similar decoded trajectories when decoding based on the place fields from different trajectories in the same environment and dissimilar decoded trajectories when decoding based on place fields from different environments. A given event with a strong decoded trajectory based on the place fields of one environment would then be expected to have a weaker decoded trajectory when decoded with place fields from an alternative environment (Figure 9c).

Lines 604-608: “(c) An example event with a statistically significant trajectory when decoded with place fields from Env. 1 left (absolute correlation at the 99th percentile of time-bin shuffles) but not when decoded with place fields of the other trajectories (78th, 45th, and 63rd percentiles, for Env. 1 right, Env. 2 left, and Env. 2 right, respectively). shows a significant trajectory when it is decoded with place fields from one environment (top row), but not when it is decoded with place fields from another environment (bottom row). “

(3) In Methods, the equation at line 610, E in the last term should be E_ext.

We modeled the feedforward inputs as excitatory connections with the same reversal potential as the recurrent excitatory connections, so  is the proper value.

(4) Equation line 617 states that conductances follow exponential decay, but the initial conductances of g_I.g_E and g_SRA are not specified.

We have added a description of the initial values in lines 760-764.

Lines 760-764: “Initial feed-forward input conductances were set to values approximating their steady-state values by randomly selecting values from a Gaussian with a mean of   and a standard deviation of . Initial values of the recurrent conductances and the SRA conductance were set to zero.”

(5) In the parameter table below line 647, W_E-E, W_E-I, and W_I-E are not described in the text.

We have clarified in lines 757-760 that the step increase in conductance corresponds to these parameter values.

Lines 757-760: “A step increase in conductance occurs at the time of each spike by an amount corresponding to the connection strength for each synapse ( for E-to-E connections, for E-to-I connections, and  for I-to-E connections), or by  for .”

(6) On line 660, "...Each environment and the sleep session had unique context cue input weights...". Does that mean that within a sleep session, the network received the same context input? How strongly are the sleep dynamics driven by that context input rather than by intrinsic dynamics? Usually, sleep activity is high dimensional, what would happen if the input during sleep is more stochastic?

Yes, within a sleep session each network receives a single set of context inputs, which are implemented as independent Poisson spike trains (so being independent, in small time-windows the dimensionality is equal to the number of neurons). The effects of any particular set of sleep context cue inputs should be minor, since the standard deviation of the input weights, , is small. Further, because the preplay analysis is performed across many networks at each parameter point, the observation of preplay is independent of any particular realization of either the recurrent network or the sleep context inputs.

Further exploring the effects of more biophysically realistic neural dynamics during simulated sleep is an interesting future direction.

(7) One bracket is missing in the denominator in line 831.

We have fixed this error.

Line 1005: “)” -> “()”

Reviewer #2 (Recommendations For The Authors):

- I would suggest the authors cite Chenkov et al 2017, PLOS Comp Bio, in which "replay" sequences were produced in clustered networks, and discuss how their work differs.

We have included a contrast of our model to that of Chenkov et al., 2017 in lines 73-78.

Lines 73-78: “Related to replay models based on place-field distance-dependent connectivity is the broader class of synfire-chain-like models. In these models, neurons (or clusters of neurons) are connected in a 1-dimensional feed-forward manner (Diesmann et al., 1999; Chenkov et al., 2017). The classic idea of a synfire-chain has been extended to included recurrent connections, such as by Chenkov et al., 2017, however such models still rely on an underlying 1-dimensional sequence of activity propagation.”

- Figure legend 2e says "replay", should be "preplay".

We have fixed this error.

Line 255: “(e) Example preplay event…”

- How much does the context cue affect the result? e.g. Is sleep notably different with different sleep context cues?

As discussed above in our response to Reviewer 1, the context cue weights have a small standard deviation, , which means that differences in the effects of different realizations of the context inputs are small. Different sets of context cues will cause cells to have slightly higher or lower spiking rates during sleep simulations, but because there is no correlation between the sleep context cue and the place field simulations there should be no effect on preplay quality.

- Figure 4 should include a control with a single cluster.

We thank the reviewer for this suggestion and have added additional control simulations.

In our model, the recurrent structure of a network with a single cluster is equivalent to a cluster-less random network. Additionally, any network where cluster participation equals the number of clusters is equivalent to a cluster-less random network, since all neurons belong to all clusters and can therefore potentially connect to any other neuron. Such a condition corresponds to a diagonal boundary where the number of clusters equals the cluster participation, which occurs at higher values of cluster participation than we had shown in our primary parameter grid.

We now include simulation results that extend to this boundary, corresponding to cluster-less networks (Figure 4—figure supplement 4f). Networks at these parameter points do not show preplay. See our earlier response for the new text associated with Figure 4—figure supplement 4.

- The results of Figure 4 are very noisy. I would recommend increasing the sampling, both in terms of the number of population events in each condition and the number of conditions.

We have run simulations for longer durations (300 seconds) and with more networks (20) to produce more accurate empirical values for the statistics calculated across the parameter grids in Figures 3 and 4. Our additional simulations (Figure 4—figure supplement 4) provide support that the parameter region of preplay significance is reliable.

Lines 831-833: “For the parameter grids in Figures 3 and 4 we simulated 20 networks with 300 s long sleep sessions in order to get more precise empirical estimates of the simulation statistics.”

- It's not entirely clear what's different between the analysis described in lines 334-353, and the preplay analysis in Figure 2. In general, the description of this result was difficult to follow, as it included a lot of text that would be better served in the methods.

In Figure 2 we first introduce the Bayesian decoding method, but it is not until Figure 4 that the shuffle-based significance testing is first introduced. We have simplified the description of the shuffle comparison in lines 371-375 and now refer the reader to the methods for details.

Lines 371-375: “We find significant preplay in both our reference experimental data set (Shin et al., 2019; Figure 4a, b; see Figure 4—figure supplement 1 for example events) and our model (Figure 4c, d) when analyzed by the same methods as Farooq et al., 2019, wherein the significance of preplay is determined relative to time-bin shuffled events (see Methods). For each detected event we calculated its absolute weighted correlation. We then generated 100 time-bin shuffles of each event, and for each shuffle recalculated the absolute weighted correlation to generate a null distribution of absolute weighted correlations.”

- Many of the figures have low text resolution (e.g. Figure 6).

We have now fixed this.

- How does the clustered small world network compare to e.g. a small world ring network as used in Watts and Strogatz 1998?

As described in our above response to Reviewer 1's fourth point, we have added a supplementary figure (Figure 1—figure supplement 1, with corresponding text) comparing our model with the Watts-Strogatz model.

Reviewer #3 (Recommendations For The Authors):

Figure 5 would benefit from a plot of the overlap of activated clusters per event.

In our cluster activation analysis in Figure 5, we defined a cluster as “active” if at any point in the event its population rate was twice that of any other clusters’. We used this definition—which permits no overlap of activated clusters—rather than a definition based on a z-scoring of the rate, because we determined that preplay required periods of spiking dominated by individual clusters.

Author response image 3.

The choice of such a definition is supported by our observation that most spiking activity within an event is dominated by whichever cluster is most active at each point in time. In the left panel of the above figure we show the distribution of the average fraction of spikes within each event that came from the most active cluster at each point in time. The right panel shows the distribution of the average across time within each event of the ratio of the population activity rate of the most active cluster to the second most active cluster. The data for both panels comes from all events at the fiducial parameter set.

Author response image 4.

Rather than overlapping at a given moment in time, clusters might have overlap in their probability of being active at some point within an event. We do find that there is a small but significant correlation in cluster co-activation. For each network we calculated the activation correlation across events for each pair of clusters (example network show in the left panel). We compared the distribution of resulting absolute correlations against the values that results after shuffling the correlations between cluster activations (right panel, all correlations for all networks from the fiducial parameter point).

Figures 4e/f are referred to as 4c/d in the text (pg 14).

We have fixed this error.

Lines 400-412: “4c” -> “4e” and “4d” -> “4f”

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

The Notch signaling pathway plays an important role in many developmental and disease processes. Although well-studied there remain many puzzling aspects. One is the fact that as well as activating the receptor through trans-activation, the transmembrane ligands can interact with receptors present in the same cell. These cis-interactions are usually inhibitory, but in some cases, as in the assays used here, they may also be activating. With a total of 6 ligands and 4 receptors, there is potentially a wide array of possible outcomes when different combinations are co-expressed in vivo. Here the authors set out to make a systematic analysis of the qualitative and quantitative differences in the signaling output from different receptor-ligand combinations, generating sets of "signaling" (ligand expressing) and "receiving" (receptor +/- ligand expressing cells).

The readout of pathway activity is transcriptional, relying on the fusion of GAL4 in the intracellular part of the receptor. Positive ligand interactions result in the proteolytic release of Gal4 that turns on the expression of H2B-citrine. As an indicator of ligand and receptor expression levels, they are linked via TA to H2B mCherry and H2B mTurq expression respectively. The authors also manipulate the expression of the glycosyltransferase Lunatic-Fringe (LFng) that modifies the EGF repeats in the extracellular domains impacting their interactions. The testing of multiple ligand-receptor combinations at varying expression levels is a tour de force, with over 50 stable cell lines generated, and yields valuable insights although as a whole, the results are quite complex.

Strengths:

Taking a reductionist approach to testing systematically differences in the signaling strength, binding strength, and cis-interactions from the different ligands in the context of the Notch1 and Notch 2 receptors (they justify well the choice of players to test via this approach) produces a baseline understanding of the different properties and leads to some unexpected and interesting findings. Notably:

-                Jag1 ligand expressing cells failed to activate Notch1 receptor although were capable of activating Notch2. Conversely, Jag2 cells elicited the strongest activation of both receptors. The results with

Jag1 are surprising also because it exhibits some of the strongest binding to plate-bound ligands. The failure to activate Notch1 has major functional significance and it will be important in the future to understand the mechanistic basis.

-                Jagged ligands have the strongest cis-inhibitory effects and the receptors differ in their sensitivity to cis-inhibition by Dll ligands. These observations are in keeping with earlier in vivo and cell culture studies. More referencing of those would better place the work in context but it nicely supports and extends previous studies that were conducted in different ways.

-                Responses to most trans-activating ligands showed a degree of ultrasensitivity but this was not the case for cis-interactions where effects were more linear. This has implications for the way the two mechanisms operate and for how the signaling levels will be impacted by ligand expression levels.

-                Qualitatively similar results are obtained in a second cell line, suggesting they reflect fundamental properties of the ligands/receptors.

We appreciate the positive and constructive feedback.

Weaknesses:

One weakness is that the methods used to quantify the expression of ligands and receptors rely on the co-translation of tagged nuclear H2B proteins. These may not accurately capture surface levels/correctly modified transmembrane proteins. In general, the multiple conditions tested partly compensate for the concerns - for example, as Jag1 cells do activate Notch2 even if they do not activate Notch1 some Jag1 must be getting to the surface. But even with Notch2, Jag1 activities are on the lower side, making it important to clarify, especially given the different outcomes with the plated ligands. Similarly, is the fact that all ligands "signalled strongest to Notch2" an inherent property or due to differences in surface levels of Notch 2 compared to Notch1? The results would be considerably strengthened by calibration of the ligand/receptor levels (and ideally their sub-cellular localizations). Assessing the membrane protein levels would be relatively straightforward to perform on some of the basic conditions because their ligand constructs contain Flag tags, making it plausible to relate surface protein to H2B, and there are antibodies available for Notch1 and Notch2.

We agree that mCherry fluorescence does not provide a direct readout of active surface ligand levels. As the reviewer points out, the ability of Jag1 to activate Notch2 demonstrates that expressed Jag1 is competent for signaling. Further, in some cases, Jag1-Notch2 activation can be comparable to Dll1-Notch2 activation (Figure 2A). Following the reviewer’s suggestion, we performed a Western blot for multiple expression levels for each of three surface ligands (Dll1, Dll4, Jag1) (Figure 2—figure supplement 2). This blot revealed a signal for surface expression of Jag1. Interpretation is complicated by the expected dependence of the efficiency of surface protein purification on the number of primary amines in the protein, which varies among these ligands, and qualitatively correlates with the staining intensity. While this makes quantitative interpretation difficult, this result further supports the notion that Jag1 is present on the cell surface. Finally, we note that high signaling activity need not, in general, directly correlate with surface expression levels. In fact, one study showed an example in which increased ligand activity occurred with decreased basal ligand surface levels (Antfolk et al., 2017). While one would ideally like to know all parameters of the system, including surface protein levels, rates of recycling, etc. the perspective taken here is that the net effect of these many post-translational processing steps can be subsumed into the overall relationship between the expression of the protein (which, in our case, is read out by the co-translational reporter) and its activity, which is relevant for the behavior of developmental circuits, among other systems. To address this comment, we now explicitly mention the limitation of mCherry as a proxy for surface protein, and add a reference to previous work highlighting the relationship between surface levels and ligand activity.

In terms of the dependence of signaling on Notch levels, the metric of signaling activity used here is explicitly normalized by the mTurquoise co-translational reporter of Notch expression to account for differences in receptor expression across receiver clones. We have added a new figure to show the variation in expression (Figure 1—figure supplement 1A) and to demonstrate this normalization (Figure 1—figure supplement 5). Having said that, as the reviewer correctly points out, we cannot directly address the dependence on surface receptor levels with mTurquoise alone. To address this comment, we have added a figure that shows cotranslational and surface receptor expression for a subset of our receiver clones (Figure 1—figure supplement 1B). Although antibody binding strengths may vary, it appears unlikely that higher surface levels could explain most ligands’ preferential activation of Notch2 over Notch1, since Notch2 levels were lower than Notch1 levels in both surface expression and cotranslational expression.

Cis-activation as a mode of signaling has only emerged from these synthetic cell culture assays raising questions about its physiological relevance. Cis-activation is only seen at the higher ligand (Dll1, Dll4) levels, how physiological are the expression levels of the ligands/receptors in these assays? Is it likely that this would make a major contribution in vivo? Is it possible that the cells convert themselves into "signaling" and "receiving" sub-populations within the culture by post-translational mechanism? Again some analysis of the ligand/receptors in the cultures would be a valuable addition to show whether or not there are major heterogeneities.

The cis-activation results in this paper are, as the reviewer points out, conducted in synthetic cell culture assays. Cis-activation is observed across a large dynamic range of ligand expression, possibly including non-physiologically high levels. However, our previous work (Nandagopal et al, eLife 2019) showed that cis-activation does not require over-expression, as it occurred in unmodified Caco-2 and NMuMG cells with their endogenous ligand and receptor expression levels. As shown here in Figure 4B, cis-activation for Notch2 increases monotonically and is substantial even at intermediate ligand concentrations. In other cases, cis-activation is maximal at intermediate concentrations. We agree that the in vivo role remains unclear, and is difficult to determine due to the typical close contacts among cells in tissues. Therefore, these assays do not speak to in vivo relevance. Note that we can, however, rule out the possibility of trans signaling between well-mixed cell populations at these densities (Figure 4A).

It is hard to appreciate how much cell-to-cell variability in the "output" there is. For example, low "outputs" could arise from fewer cells becoming activated or from all cells being activated less. As presented, only the latter is considered. That may be already evident in their data, but not easy for the reader to distinguish from the way they are presented. For example, in many of the graphs, data have been processed through multiple steps of normalization. Some discussion/consideration of this point is needed.

We agree that in different experiments changes in a mean response can reflect changes in fraction of activated cells, or level of activation or some combination of both. In this work, most assays were conducted by flow cytometry, which provides a full distribution of cellular responses. We provided distributions for some experiments in the supplementary figures (i.e., Figure 4—figure supplement 1, and Figure 5—figure supplement 4). The sheer number of experiments and samples prevents us from displaying all underlying histograms. Therefore, we have provided all flow data sets in an extensive archive that is publicly available on data.caltech.edu (https://doi.org/10.22002/gjjkn-wrj28).

Impact:

Overall, cataloging the outcomes from the different ligand-receptor combinations, both in cis and trans, yields a valuable baseline for those investigating their functional roles in different contexts. There is still a long way to go before it will be possible to make a predictive model for outcomes based on expression levels, but this work gives an idea about the landscape and the complexities. This is especially important now that signaling relationships are frequently hypothesized based on single-cell transcriptomic data. The results presented here demonstrate that the relationships are not straightforward when multiple players are involved.

We appreciate this concise impact summary, and agree with its conclusions.

Reviewer #2 (Public Review):

Summary:

In this manuscript, the authors extend their previous studies on trans-activation, cis-inhibition (PMID: 25255098), and cis-activation (PMID: 30628888) of the Notch pathway. Here they create a large number of cell lines using CHO-K1 and C2C12 cells expressing either Notch1-Gal4 or Notch2-Gal4 receptors which express a fluorescent protein upon receptor activation (receiver cells). For cis-inhibition and cis-activation assays, these cells were engineered to express one of the four canonical Notch ligands (Dll1, Dll4, Jag1, Jag2) under tetracycline control. Some of the receiver cells were also transfected with a Lunatic fringe (Lfng) plasmid to produce cells with a range of Lfng expression levels. Sender cells expressing all of the canonical ligands were also produced. Cells were mixed in a variety of co-culture assays to highlight trans-activation, cis-activation, and cis-inhibition. All four ligands were able to trans-activate Notch1 and Notch 2, except Jag1 did not transactivate Notch1. Lfng enhanced trans-activation of both Notch receptors by Dll1 and Dll2, and inhibited Notch1 activation by Jag2 and Notch2 activation by both Jag 1 and Jag2. Cis-expression of all four ligands was predominantly inhibitory, but Dll1 and Dll4 showed strong cis-activation of Notch2. Interestingly, cis-ligands preferentially inhibited trans-activation by the same ligand, with varying effects on other trans-ligands.

Strengths:

This represents the most comprehensive and rigorous analysis of the effects of canonical ligands on cis- and trans-activation, and cis-inhibition, of Notch1 and Notch2 in the presence or absence of Lfng so far. Studying cis-inhibition and cis-activation is difficult in vivo due to the presence of multiple Notch ligands and receptors (and Fringes) that often occur in single cells. The methods described here are a step towards generating cells expressing more complex arrays of ligands, receptors, and Fringes to better mimic in vivo effects on Notch function.

In addition, the fact that their transactivation results with most ligands on Notch1 and 2 in the presence or absence of Lfng were largely consistent with previous publications provides confidence that the author's assays are working properly.

We appreciate the thoughtful comments and feedback.

Weaknesses:

It was unusual that the engineered CHO cells expressing Notch1-Gal4 were not activated at all by co-culture with Jag1-expressing CHO cells. Many previous reports have shown that Jag1 can activate Notch1 in co-culture assays, including when Notch1 was expressed in CHO cells. Interestingly, when the authors used Jag1-Fc in a plate coating assay, it did activate Notch1 and could be inhibited by the expression of Lfng.

In our assays, we do in fact also see some signaling of Jag1 to Notch1, especially when dLfng is coexpressed (Figure 2—figure supplement 4, formerly Figure 2—figure supplement 3). While these levels are lower than those observed for other ligand-receptor combinations, they are significantly elevated compared to baseline. In specific natural contexts, it will be important to determine whether the weak but non-zero Jag1-Notch1 signaling acts negatively to suppress signaling from other ligands, or provides weak but potentially functionally important levels of signaling. Evidence for both modes exists in the literature. To address this, we have expanded the discussion of Jag1-Notch1 signaling and added references to other work on Jag1-Notch1 signaling to the Discussion section.

The cell surface level of the ligands was determined by flow cytometry of a co-translated fluorescent protein. Some calibration of the actual cell surface levels with the fluorescent protein would strengthen the results.

This issue was also raised by Reviewers #1 and #3. Please see responses to Reviewer #1, above.

Reviewer #3 (Public Review):

Summary:

This manuscript reports a comprehensive analysis of Notch-Delta/Jagged signaling inclusive of the human Notch1 and Notch2 receptors and DLL1, DLL4, JAG1, and JAG2 ligands. Measurements

encompassed signaling activity for ligand trans-activation, cis-activation, cis-inhibition, and activity modulation by Lfng. The most striking observations of the study are that JAG1 has no detectable activity as a Notch1 ligand when presented on a cell (though it does have activity when immobilized on a surface), even though it is an effective cis-inhibitor of Notch1 signaling by other ligands, and that DLL1 and DLL4 exhibit cis-activating activity for Notch1 and especially for Notch2. Notwithstanding the artificiality of the system and some of its shortcomings, the results should nevertheless be a valuable resource for the Notch signaling community.

Strengths:

(1)  The work is systematic and comprehensive, addressing questions that are of importance to the community of researchers investigating mammalian Notch proteins, their activation by ligands, and the modulation of ligand activity by LFng.

(2)  A quantitative and thorough analysis of the data is presented.

Weaknesses:

(1) The manuscript is primarily descriptive and does not delve into the underlying, mechanistic origin or source of the different ligand activities.

We agree that the goals of this paper were largely to discover the range of signaling modes that occur. A mechanistic analysis would be beyond the scope of this work, but we agree it is an important next step.

(2) The amount of ligand or receptor expressed is inferred from the flow cytometry signal of a co-translated fluorescent protein-histone fusion, and is not directly measured. The work would be more compelling if the amount of ligand present on the cell surface were directly measured with anti-ligand antibodies, rather than inferred from measurements of the fluorescent protein-histone fusion.

This issue was also raised by Reviewers #1 and #2. Please see responses to Reviewer #1, above.

(3) It would be helpful to see plots of the raw activity data before transformation and normalization, because the plots present data after several processing steps, and it is not clear how the processed data relate to the original values determined in each measurement.

We included examples showing how raw data is processed in Figure 4—figure supplement 1 and Figure 5—figure supplement 4. The sheer number of experiments precludes including similar figures for all data sets. However, all raw and processed data and data analysis code is publicly available at (https://doi.org/10.22002/gjjkn-wrj28).

(4) The authors use sparse plating of engineered cells with parental (no ligand or receptor-expressing cell to measure cis activation). However, the cells divide within the cultured period of 22-24 h and can potentially trans-activate each other.

If measured cis-activation signal arises solely from trans-activation, then the measured cis-activation signal per cell should increase with cell density, since trans-activation per cell does depend on cell density (Figure 4A). However, for the strongest cis-activators (Dll1- and Dll4-Notch2), signaling magnitude is similar when these cells are cultured sparsely or at confluence, which would otherwise allow efficient trans signaling (Figure 5A). Thus, for Dll1- and Dll4-Notch2 receivers, total signaling strength per cell depends little or not at all on the opportunity to signal intercellularly. Moreover, cis-activation signal for the Dll1- and Dll4-Notch2 combinations exceeded the maximum trans-signaling levels we could achieve for the same receivers when cis-ligand was suppressed (Figure 4B). These results argue that cis interactions dominate signaling in this context. However, we have not ruled out the possibility that trans-signaling between sister cells after division contributes to the comparatively weak cis-activation observed for Notch1 receivers.

Reviewer #1 (Recommendations For The Authors):

As outlined in the public review, there is a question of whether the nuclear H2B accurately reflects the surface levels of the transmembrane proteins (ligand and receptor). Clearly, it would not be feasible to check levels in all of the experimental conditions, but some baseline conditions should be analyzed.

We addressed this above.

Reviewer #2 (Recommendations For The Authors):

(1)  As mentioned above, it was unusual that Jag1 did not activate Notch1 in co-culture assays, but did activate Notch1 in plate-coating assays. The authors should add some text to the Discussion to explain why they think this is happening in their engineered cells. One possibility is that the CHO cells express Manic fringe (Mfng) which is known to reduce Jag1-Notch1 activation. Data for Mfng levels in CHO cells were not included in Supplemental Table 2. Knocking down all three Fringes in CHO cells might increase Jag1-Notch1 activation.

This is already addressed in a sentence in the results: “Strikingly, while Jag1 sender cells failed to activate Notch1 receivers above background (Figure 2D), plate-bound Jag1-ext-Fc activated Notch1 only ~3-fold less efficiently than it activated Notch2 (Figure 3B-D). This suggests that the natural endocytic activation mechanism, or potential differences in tertiary structure between the expressed and recombinant Jag1 extracellular domains, could play roles in preventing Jag1-Notch1 signaling in coculture.” Regarding the point about Mfng, we added a note to Supplementary Table about other CHO-K1 expression data.

(2) Figure 1-supplemental figure 1: Both the Notch1-Jag1 and Notch1-Jag2 cells show high expression of Jag1 in low 4epi, but any higher concentration reduces to control levels. How much of a problem is this for interpreting your data?

This was not the ideal behavior, but by binning cells by co-translational reporters for ligand expression, we were able to obtain enough cells in intermediate bins. (Note: Figure 1—figure supplement 1 is now Figure 1—figure supplement 2.)

(3)  Figure 1C legend: Are these stably-expressing cells or Tet-off cells? Please state in legend.

The figure legend has been updated.

(4)  Figure 1E: How long is the knockdown of Rfng and Lfng effective? Does it affect the expression of Lfng later?

siRNA effects generally last for at least 72-96 hours, so we do not anticipate this being an issue.

(5) Page 9: "Lfng significantly decreased trans-activation of both receptors by Jag1 (>2.5-fold)". If there is no Jag1-Notch1 activation, how can Lfng decrease trans-activation?

We added a note in the main text to clarify that while Jag1-Notch1 signaling is relatively low, it can still be detectably decreased.

(6) Figure 4A legend: Please define what "2.5k ea senders and Rec" means. In the text, it says "To focus on cis-interactions alone, we then cultured receiver cells at low density, amid an excess of wildtype CHO-K1 cells" (page 14).

This was clarified in the text.

(7)  Page 14: "By contrast, Notch2 was cis-activated by both Dll1 and Dll4, to levels exceeding those produced by trans-activation by high-Dll1 senders (Figure 4B, lower left)." Where is the trans-activation data? 4B, lower right?

We updated this reference in the main text.

(8)  Page 16: "For Notch2-Dll1 and Notch2-Dll4, single cell reporter activities correlated with cis-ligand expression, regardless of whether cells were pre-induced at a high or low culture density (Figure 4D)." It appears that Notch2-Dll1 has lower Notch activation at sparse culture than confluent.

We agree that the level signaling is lower in sparse compared to confluent on average. This is explained by the sensitivity of the Tet-OFF promoter to culture density (Figure 4—figure supplement 2). However, the key point of this experiment is the positive correlation, which is consistent with cis-activation, and inconsistent with the pre-generation of NEXT hypothesis diagrammed in Figure 4C, which would not be expected to produce such a correlation.

(9a) For the creation of the C2C12-Nkd cells: Has genomic sequencing been done to confirm editing of Notch2 and Jag1 loci?

We confirmed the knockdown but did not do genomic sequencing.

(9b) The gel in Figure 7-Supplement 1C is not adequate for showing loss of Jag1. It should be repeated.

In this case, we have only the single gel. We added a note in figure legend that no duplicate was performed.

(10) Figure 7A: Which Fringes are expressed in C2C12 cells? You should provide a rationale for knocking down just Rfng.

Figure 7—figure supplement 1A shows the levels of expression in C2C12. Note that Mfng is not highlighted because its levels were undetectable.

(11) Figure 7-Supplement 1D: This is confusing. Notch2 levels are not reduced in the left panel, and Notch1 and Notch2 levels are not reduced in the right panel?

C2C12-Nkd cells exhibit reduced levels of Notch1 and Notch3. This can be seen in Figure 7—figure supplement 1A. Panel D presents the results of additional siRNA knockdown, performed to prevent subsequent up-regulation of Notch1 and Notch3 during the assay. These knockdown results were variable, as shown. The Notch2 siRNA knockdown was not essential for these experiments, but performed despite very low levels of Notch2 to begin with. In the revision, we have added this note to the Methods.

Reviewer #3 (Recommendations For The Authors):

(1) The results section of the manuscript is very dense and difficult to follow, as are the figure legends.

We appreciate the criticism, and regret that it is not easier to read in its current form.

(2) The authors could emphasize areas of concordance with published results (where available) to place their artificial, engineered system into a better biological context. Are there any examples of studies in whole organisms where cis-activation plays a role?

We are not aware of examples of cis-activation in whole organisms at this point.

(3) How do the authors rationalize the different responses of Notch1 to cell-presented Jag1 as opposed to immobilized Jag1, where its signal strength is second in rank order on a molar basis?

This comment was addressed above in response to the first recommendation from Reviewer #2.

It is also difficult to understand Figure 2_—_figure Supplement 3B, in which it appears that Jag1 induces a Notch1 reporter response when LFng is knocked down (dLfng), and how those data relate to the inactive response to Jag1 shown in the main figures.

The issue here is a difference of normalization. Figure 2A in the main text is normalized to the sender expression level, i.e. relative signaling strength. By contrast, Figure 2—figure supplement 4B (previously Figure 2—figure supplement 3B) shows absolute signaling activity, which can appear higher because it does not normalize for ligand expression. For Jag1-Notch1 signaling in particular, substantial signaling required very high levels of Jag1. We have added a new figure to demonstrate these two types of normalization (Figure 2—figure supplement 1A).

See the Authr response image 1 below for a direct comparison of these two normalization modes using data from both Figure 2A and Figure 2—figure supplement 4B. Note how the Jag1-Notch1 signaling activities that are nonzero in the top plot go to zero in the bottom plot as a result of normalizing the values to ligand expression.

Author response image 1. Comparison of normalization modes in Figure 2A and Figure 2—figure supplement 4B (formerly 3B). Normalized trans-activation signaling activities for different ligand-receptor combinations (with dLfng only), either with further normalization to ligand expression (bottom row) or without further normalization (top row). Normalized signaling activity is defined as reporter activity (mCitrine, A.U.) divided by cotranslational receptor expression (mTurq2, A.U.), normalized to the strongest biological replicate-averaged signaling activity across all ligand-receptor-Lfng combinations in this experiment. Saturated data points, defined here as those with normalized signaling activity over 0.75 in both dLfng and Lfng conditions, were excluded. Colors indicate the identity of the trans-ligand expressed by cocultured sender cells. Error bars denote bootstrapped 95% confidence intervals (Methods), in this case sampled from the number of biological replicates given in the legend—n1 (for Notch1) or n2 (for Notch2). See Methods and Figure 2A caption for more details. Note that the only difference between this figure and the new Figure 2—figure supplement 1A is that this figure additionally includes the Jag1-high data from Figure 2—figure supplement 4B.

Author response:

The following is the authors’ response to the current reviews.

Reviewer #1 (Public review):

Summary:

In this manuscript, Herrmannova et al explore changes in translation upon individual depletion of three subunits of the eIF3 complex (d, e and h) in mammalian cells. The authors provide a detailed analysis of regulated transcripts, followed by validation by RT-qPCR and/or Western blot of targets of interest, as well as GO and KKEG pathway analysis. The authors confirm prior observations that eIF3, despite being a general translation initiation factor, functions in mRNA-specific regulation, and that eIF3 is important for translation re-initiation. They show that global effects of eIF3e and eIF3d depletion on translation and cell growth are concordant. Their results support and extend previous reports suggesting that both factors control translation of 5'TOP mRNAs. Interestingly, they identify MAPK pathway components as a group of targets coordinately regulated by eIF3 d/e. The authors also discuss discrepancies with other reports analyzing eIF3e function.

Strengths:

Altogether, a solid analysis of eIF3 d/e/h-mediated translation regulation of specific transcripts. The data will be useful for scientists working in the Translation field.

Weaknesses:

The authors could have explored in more detail some of their novel observations, as well as their impact on cell behavior.

The manuscript has improved with the new corrections. I appreciate the authors' attention to the minor comments, which have been fully solved. The authors have not, however, provided additional experimental evidence that uORF-mediated translation of Raf-1 mRNA depends on an intact eIF3 complex, nor have they addressed the consequences of such regulation for cell physiology. While I understand that this is a subject of follow-up research, the authors could have at least included their explanations/ speculations regarding major comments 2-4, which in my opinion could have been useful for the reader.

Our explanations/speculations regarding major comments 2 and 3 were included in the Discussion. We apologize for this misunderstanding as we thought that we were supposed to explain our ideas only in the responses. We did not discuss the comment 4, however, as we are really not sure what is the true effect and did not want to go into wild speculations in our manuscript. We thank this reviewer for his insightful comments and understanding.

The following is the authors’ response to the original reviews.

Reviewer #1 (Recommendations For The Authors):

Major comments:

(1) The authors report the potential translational regulation of Raf kinase by re-initiation. It would be interesting to show that Raf is indeed regulated by uORF-mediated translation, and that this is dependent on an intact eIF3 complex. Analyzing the potential consequences of Raf1 regulation for cancer cell proliferation or apoptosis would be a plus.

We agree that this is an interesting and likely possibility. In fact, another clue that translation of Raf1 is regulated by uORFs comes from Bohlen et al. 2023 (PMID: 36869665) where they showed that RAF1 translation is dependent on PRRC2 proteins (that promote leaky scanning through these uORFs). We noted in the discussion that our results from eIF3d/e/hKD and the PRRC2A/B/CKD partly overlap. It is a subject of our follow-up research to investigate whether eIF3 and PRRC2 co-operate together to regulate translation of this important mRNA. 

(2) The authors show that eIF3 d/e -but not 3h- has an effect on cell proliferation. First, this indicates that proliferation does not fully correlate with eIF3 integrity. Depletion of eIF3d does not affect the integrity of eIF3, yet the effects on proliferation are similar to those of eIF3e. What is the possibility that changes in proliferation reflect functions of eIF3d outside the eIF3 complex? What could be the real consequences of disturbing eIF3 integrity for the mammalian cell? Please, discuss.

Yes, proliferation does not fully correlate with eIF3 integrity. Downregulation of eIF3 subunits that lead to disintegration of eIF3 YLC core (a, b, c, g, i) have more detrimental effect on growth and translation than downregulation of the peripheral subunits (e, k, l, f, h, m). Our previous studies (Wagner et al. 2016, PMID: 27924037 and Herrmannová et al. 2020, PMID: 31863585) indicate that the YLC core of eIF3 can partially support translation even without its peripheral subunits. In this respect eIF3d (as a peripheral subunit) is an amazing exception, suggesting it may have some specialized function(s). Whether this function resides outside of the eIF3 complex or not we do not know, but do not think so. Mainly because in the absence of eIF3e – its interaction partner, eIF3d gets rapidly degraded. Therefore, it is not very likely that eIF3d exists alone outside of eIF3 complex with moonlighting functions elsewhere. We think that eIF3d, as a head-interacting subunit close to an important head ribosomal protein RACK1 (a landing pad for regulatory proteins), is a target of signaling pathways, which may make it important for translation of specific mRNAs. In support is these thoughts, eIF3d (in the context of entire eIF3) together with DAP5 were shown to promote translation by an alternate capdependent (eIF4F-independent) mechanism (Lee et al. 2016, PMID: 27462815; de la Parra et al. 2018, PMID:30076308). In addition, the eIF3d function (also in the context of entire eIF3) was proved to be regulated by stress-triggered phosphorylation (Lamper et al. 2020, PMID: 33184215). 

(3) Figure 6D: Surprisingly, reduced levels of ERK1/2 upon eIF3d/e-KD are compensated by increased phosphorylation of ERK1/2 and net activation of c-Jun. Please comment on the functional consequences of buffering mechanisms that the cell deploys in order to counteract compromised eIF3 function. Why would the cell activate precisely the MAPK pathway to compensate for a compromised eIF3 function?

This we do not know. We can only speculate that when translation is compromised, cells try to counteract it in two ways: 1) they produce more ribosomes to increase translational rates and 2) activate MAPK signaling to send pro-growth signals, which can in the end further boost ribosome biogenesis.

(4) Regarding DAP-sensitive transcripts, can the authors discuss in more detail the role of eIF3d in alternative cap-dependent translation versus re-initiation? Are these transcripts being translated by a canonical cap- and uORF-dependent mechanism or by an alternative capdependent mechanism?

This is indeed not an easy question. On one hand, it was shown that DAP5 facilitates translation re-initiation after uORF translation in a canonical cap-dependent manner. This mechanism is essential for translation of the main coding sequence (CDS) in mRNAs with structured 5' leaders and multiple uORFs. (Weber et al. 2022, PMID: 36473845; David et al., 2022, PMID: 35961752). On the other hand, DAP5 was proposed to promote alternative, eIF4F-independent but cap-dependent translation, as it can substitute the function of the eIF4F complex in cooperation with eIF3d (de la Parra et al., 2018, PMID: 30076308; Volta et al., 2021 34848685). Overall, these observations paint a very complex picture for us to propose a clear scenario of what is going on between these two proteins on individual mRNAs. We speculate that both mechanisms are taking place and that the specific mechanism of translation initiation differs for differently arranged mRNAs.

Minor comments:

(5) Figure S2C: why is there a strong reduction of the stop codon peak for 3d and 3h KDs?

We have checked the Ribowaltz profiles of all replicates (in the Supplementary data we are showing only a representative replicate I) and the stop codon peak differs a lot among the replicates. We think that this way of plotting was optimized for calculation and visualization of P-sites and triplet periodicity and thus is not suitable for this type of comparison among samples. Therefore, we have performed our own analysis where the 5’ ends of reads are used instead of P-sites and triplicates are averaged and normalized to CDS (see below please), so that all samples can be compared directly in one plot (same as Fig. S13A but for stop codon). We can see that the stop codon peak really differs and is the smallest for eIF3hKD. However, these changes are in the range of 20% and we are not sure about their biological significance. We therefore refrain from drawing any conclusions. In general, reduced stop codon peak may signal faster termination or increased stop codon readthrough, but the latter should be accompanied by an increased ribosome density in the 3’UTR, which is not the case. A defect in termination efficiency would be manifested by an increased stop codon peak, instead.

Author response image 1.

(6) Figures 5 and S8: Adding a vertical line at 'zero' in all cumulative plots will help the reader understand the author's interpretation of the data. 

We have added a dashed grey vertical line at zero as requested. However, for interpretation of these plots, the reader should focus on the colored curve and whether it is shifted in respect to the grey curve (background) or not. Shift to the right indicates increased expression, while shift to the left indicates decreased expression. The reported p-value then indicates the statistical significance of the shift.

(7) The entire Figure 2 are controls that can go to Supplementary Material. The clustering of Figure S3B could be shown in the main Figure, as it is a very easy read-out of the consistent effects of the KDs of the different eIF3 subunits under analysis.

We have moved the entire Figure 2 to Supplementary Material as suggested (the original panels can be found as Supplementary Figures 1B, 1C and 3A). Figure S3B is now the main Figure 2E. 

(8) There are 3 replicates for Ribo-Seq and four for RNA-Seq. Were these not carried out in parallel, as it is usually done in Ribo-seq experiments? Why is there an extra replicate for RNASeq?

Yes, the three replicates were carried out in parallel. We have decided to add the fourth replicate in RNA-Seq to increase the data robustness as the RNA-Seq is used for normalization of FP to calculate the TE, which was our main analyzed metrics in this article. We had the option to add the fourth replicate as we originally prepared five biological replicates for all samples, but after performing the control experiments, we selected only the 3 best replicates for the Ribo-Seq library preparation and sequencing.  

(9) Please, add another sheet in Table S2 with the names of all genes that change only at the translation (RPF) levels.

As requested, we have added three extra sheets (one for each downregulation) for differential FP with Padjusted <0.05 in the Spreadsheet S2. We also provide a complete unfiltered differential expression data (sheet named “all data”), so that readers can filter out any relevant data based on their interest.

(10) Page 5, bottom: ' ...we showed that the expression of all 12 eIF3 subunits is interconnected such that perturbance of the expression of one subunit results in the down-regulation of entire modules...'. This is not true for eIF3d, as shown in Fig1B and mentioned in Results.

This reviewer is correct. By this generalized statement, we were trying to summarize our previous results from Wagner et al., 2014, PMID: 24912683; Wagner et al.,2016, PMID: 27924037 and Herrmannova et al.,2020, PMID: 31863585. The eIF3d downregulation is the only exception that does not affect expression of any other eIF3 subunit. Therefore, we have rewritten this paragraph accordingly: “We recently reported a comprehensive in vivo analysis of the modular dynamics of the human eIF3 complex (Wagner et al, 2020; Wagner et al, 2014; Wagner et al., 2016). Using a systematic individual downregulation strategy, we showed that the expression of all 12 eIF3 subunits is interconnected such that perturbance of the expression of one subunit results in the down-regulation of entire modules leading to the formation of partial eIF3 subcomplexes with limited functionality (Herrmannova et al, 2020). eIF3d is the only exception in this respect, as its downregulation does not influence expression of any other eIF3 subunit.”

(11) Page 10, bottom: ' The PCA plot and hierarchical clustering... These results suggest that eIF3h depletion impacts the translatome differentially than depletion of eIF3e or eIF3d.' This is already obvious in the polysome profiles of Figure S2C.

We agree that this result is surely not surprising given the polysome profile and growth phenotype analyses of eIF3hKD. But still, we think that the PCA plot and hierarchical clustering results represent valuable controls. Nonetheless, we rephrased this section to note that this result agrees with the polysome profiles analysis: “The PCA plot and hierarchical clustering (Figure 2A and Supplementary Figure 4A) showed clustering of the samples into two main groups: Ribo-Seq and RNA-seq, and also into two subgroups; NT and eIF3hKD samples clustered on one side and eIF3eKD and eIF3dKD samples on the other. These results suggest that the eIF3h depletion has a much milder impact on the translatome than depletion of eIF3e or eIF3d, which agrees with the growth phenotype and polysome profile analyses (Supplementary Figure 1A and 1D).”

(12) Page 12: ' As for the eIF3dKD "unique upregulated" DTEGs, we identified one interesting and unique KEGG pathway, the ABC transporters (Supplementary Figure 5A, in green).' This sentence is confusing, as there are more pathways that are significant in this group, so it is unclear why the authors consider it 'unique'.

The eIF3dKD “unique upregulated” group comprises genes with increased TE only in eIF3dKD but not in eIF3eKD or eIF3hKD (500 genes, Fig 2G). All these 500 genes were examined for enrichment in the KEGG pathways, and the top 10 significant pathways were reported (Fig S6A). However, 8 out of these 10 pathways were also significantly enriched in other gene groups examined (e.g. eIF3d/eIF3e common). Therefore, the two remaining pathways (“ABC transporters” and “Other types of O-glycan biosynthesis”) are truly unique for eIF3dKD. We wanted to highlight the ABC transporters group in particular because we find it rather interesting (for the reasons mentioned in the article). We have corrected the sentence in question to avoid confusion: “Among the eIF3dKD “unique upregulated” DTEGs, we identified one interesting KEGG pathway, the ABC transporters, which did not show up in other gene groups (Supplementary Figure 6A, in green). A total of 12 different ABC transporters had elevated TE (9 of them are unique to eIF3dKD, while 3 were also found in eIF3eKD), 6 of which (ABCC1-5, ABCC10) belong to the C subfamily, known to confer multidrug resistance with alternative designation as multidrug resistance protein (MRP1-5, MRP7) (Sodani et al, 2012).

Interestingly, all six of these ABCC transporters were upregulated solely at the translational level (Supplementary Spreadsheet S2).”    

(13) Note typo ('Various') in Figure 4A.

Corrected

(14) The introduction could be shortened.

This is a very subjective requirement. In fact, when this manuscript was reviewed in NAR, we were asked by two reviewers to expand it substantially. Because a number of various research topics come together in this work, e.g. translational regulation, the eIF3 structure and function, MAPK/ERK signaling, we are convinced that all of them demand a comprehensive introduction for non-experts in each of these topics. Therefore, with all due respect to this reviewer, we did not ultimately shorten it.

Reviewer #2 (Recommendations For The Authors):

- In Figure 2, it would be useful to know why eIF3d is destabilized by eIF3e knockdown - is it protein degradation and why do the eIF3d/e knockdowns not more completely phenocopy each other when there is the same reduction to eIF3d as in the eIF3d knockdown sample?

Yes, we do think that protein degradation lies behind the eIF3d destabilization in the eIF3eKD, but we have not yet directly demonstrated this. However, we have shown that eIF3d mRNA levels are not altered in eIF3eKD and that Ribo-Seq data indicate no change in TE or FP for eIF3d-encoding mRNA in eIF3eKD. Nonetheless, it is important to note (and we discuss it in the article) that eIF3d levels in eIF3dKD are lower than eIF3d levels in eIF3eKD (please see Supplementary Figure 1C). In fact, we believe that this is one of the main reasons for the eIF3d/e knockdowns differences.

- The western blots in Figures 4 and 6 show modest changes to target protein levels and would be strengthened by quantification.

We have added the quantifications as requested by this reviewer and the reviewer 3.

- For Figure 4, this figure would be strengthened by experiments showing if the increase in ribosomal protein levels is correlated with actual changes to ribosome biogenesis.

As suggested, we performed polysome profiling in the presence of EDTA to monitor changes in the 60S/40S ratio, indicating a potential imbalance in the biogenesis of individual ribosome subunits. We found that it was not affected (Figure 3G). In addition, we performed the same experiment, normalizing all samples to the same number of cells (cells were carefully counted before lysis). In this way, we confirmed that eIF3dKD and eIF3eKD cells indeed contain a significantly increased number of ribosomes, in agreement with the western blot analysis (Figure 3H).

- In Figure 6, there needs to be a nuclear loading control.

This experiment was repeated with Lamin B1 used as a nuclear loading control – it is now shown as Fig. 5F.

- For Figure 8, these findings would be strengthened using luciferase reporter assays where the various RNA determinants are experimentally tested. Similarly, 5′ TOP RNA reporters would have been appreciated in Figure 4.

This is indeed a logical continuation of our work, which represents the current work in progress of one of the PhD students. We apologize, but we consider this time- and resource-demanding analysis out of scope of this article.

Reviewer #3 (Recommendations For The Authors):

(1) Within the many effects observed, it is mentioned that eIF3d is known to be overexpressed while eIF3e is underexpressed in many cancers, but knockdown of either subunit decreases MDM2 levels, which would be expected to increase P53 activity and decrease tumor cell transformation. In contrast, they also report that 3e/3d knockdown dramatically increases levels of cJUN, presumably due to increased MAPK activity, and is expected to increase protumor gene expression. Additional discussion is needed to clarify the significance of the findings, which are a bit confusing.

This is indeed true. However, considering the complexity of eIF3, the largest initiation factor among all, as well as the broad portfolio of its functions, it is perhaps not so surprising that the observed effects are complex and may seem even contradictory in respect to cancer. To acknowledge that, we expanded the corresponding part of discussion as follows: “Here, we demonstrate that alterations in the eIF3 subunit stoichiometry and/or eIF3 subcomplexes have distinct effects on the translatome; for example, they affect factors that play a prominent (either positive or negative) role in cancer biology (e.g., MDM2 and cJUN), but the resulting impact is unclear so far. Considering the complex interactions between these factors as well as the complexity of the eIF3 complex per se, future studies are required to delineate the specific oncogenic and tumor suppressive pathways that play a predominant role in mediating the effects of perturbations in the eIF3 complex in the context of neoplasia.”

(2) There are places in the text where the authors refer to changes in transcriptional control when RNA levels differ, but transcription versus RNA turnover wasn't tested, e.g. page 16 and Figure S10, qPCR does not confirm "transcriptional upregulation in all three knockdowns" and page 19 "despite apparent compensatory mechanisms that increase their transcription."

This is indeed true, the sentences in question were corrected. The term “increased mRNA levels” was used instead of transcriptional upregulation (increased mRNA stabilization is also possible).

(3) Similarly, the authors suggest that steady-state LARP1 protein levels are unaffected based on ribosome footprint counts (page 21). It is incorrect to assume this, because ribosome footprints can be elevated due to stalling on RNA that isn't being translated and doesn't yield more protein, and because levels of translated RNA/synthesized proteins do not always reflect steady-state protein levels, especially in mutants that could affect lysosome levels and protein turnover. Also page 12, 1st paragraph suggests protein production is down when ribosome footprints are changed.

Yes, we are well-aware of this known limitation of Ribo-seq analysis. Therefore, the steadystate protein levels of our key hits were verified by western blotting. In addition, we have removed the sentence about LARP1 because it was based on Ribo-Seq data only without experimental evaluation of the steady-state LARP1 protein levels.

(4) The translation buffering effect is not clear in some Figures, e.g. S6, S8, 8A, and B. The authors show a scheme for translationally buffered RNAs being clustered in the upper right and lower left quadrants in S4H (translation up with transcript level down and v.v.), but in the FP versus RNA plots, the non-TOP RNAs and 4E-P-regulated RNAs don't show this behavior, and appear to show a similar distribution to the global changes. Some of the right panels in these figures show modest shifts, but it's not clear how these were determined to be significant. More information is needed to clarify, or a different presentation, such as displaying the RNA subsets in the left panels with heat map coloring to reveal whether RNAs show the buffered translation pattern defined in purple in Figure S4H, or by reporting a statistical parameter or number of RNAs that show behavior out of total for significance. Currently the conclusion that these RNAs are translationally buffered seems subjective since there are clearly many RNAs that don't show changes, or show translation-only or RNA-only changes.

We would like to clarify that S4H does not indicate a necessity for changes in FPs in the buffered subsets. Although opposing changes in total mRNA and FPs are classified as buffering, often we also consider the scenario where there are changes to the total mRNA levels not accompanied by changes in ribosome association.

In figure S6, the scatterplots indicate a high density of genes shifted towards negative fold changes on the x-axis (total mRNA). This is also reflected in the empirical cumulative distribution functions (ecdfs) for the log2 fold changes in total mRNA in the far right panels of A and B, and the lack of changes in log2 fold change for FPs (middle panels). Similarly, in figure S8, the scatterplots indicate a density of genes shifted towards positive fold changes on the x-axis for total mRNA. The ecdfs also demonstrate that there is a significant directional shift in log2 fold changes in the total mRNA that is not present to a similar degree in the FPs, consistent with translational offsetting. It is rightly pointed out that not all genes in these sets follow the same pattern of regulation. We have revised the title of Supplementary Figure S6 (now S7) to reflect this. However, we would like to emphasize that these figures are not intended to communicate that all genes within these sets of interest are regulated in the same manner, but rather that when considered as a whole, the predominant effect seen is that of translational offsetting (directional shifts in the log2 fold change distribution of total mRNA that are not accompanied by similar shifts in FP mRNA log2 fold changes).

The significance of these differences was determined by comparing the ecdfs of the log2 fold changes for the genes belonging to a particular set (e.g. non-TOP mTOR-sensitive, p-eIF4E-sensitive) against all other expressed genes (background) using a Wilcoxan rank sum test. This allows identification of significant shifts in the distributions that have a clear directionality (if there is an overall increase, or decrease in fold changes of FPs or total mRNA compared to background). If log2 fold changes are different from background, but without a clear directionality (equally likely to be increased or decreased), the test will not yield a significant result. This approach allows assessment of the overall behavior of gene signatures within a given dataset in a manner that is completely threshold-independent, such that it does not rely on classification of genes into different regulatory categories (translation only, buffering, etc.) based on significance or fold-change cut-offs (as in S4H). Therefore, we believe that this unbiased approach is well-suited for identifying cases when there are many genes that follow similar patterns of regulation within a given dataset.

(5) Page 10-"These results suggest that eIF3h depletion impacts the translatome differentially than depletion of eIF3e or eIF3d" ...These results suggest that eIF3h has less impact on the translatome, not that it does so differently. If it were changing translation by a different mechanism, I would not expect it to cluster with control.

This sentence was rewritten as follows: “The PCA plot and hierarchical clustering (Figure 2A and Supplementary Figure 4A) showed clustering of the samples into two main groups: RiboSeq and RNA-seq, and also into two subgroups; NT and eIF3hKD samples clustered on one side and eIF3eKD and eIF3dKD samples on the other. These results suggest that the eIF3h depletion has a much milder impact on the translatome than depletion of eIF3e or eIF3d, which agrees with the growth phenotype and polysome profile analyses (Supplementary Figure 1A and 1D).”

Other minor issues:

(1) There are some typos: Figure 2 leves, Figure 4 variou,

Corrected.

(2) Figure 3, font for genes on volcano plot too small

Yes, maybe, however the resolution of this image is high enough to enlarge a certain part of it at will. In our opinion, a larger font would take up too much space, which would reduce the informativeness of this graph.

(3) Figure S5, highlighting isn't defined.

The figure legend for S5A (now S6A) states: “Less significant terms ranking 11 and below are in grey. Terms specifically discussed in the main text are highlighted in green.” Perhaps it was overlooked by this reviewer.

(4) At several points the authors refer to "the MAPK signaling pathway", suggesting there is a single MAPK that is affected, e.g in the title, page 3, and other places when it seems they mean "MAPK signaling pathways" since several MAPK pathways appear to be affected.

We apologize for any terminological inaccuracies. There are indeed several MAPK pathways operating in cells. In our study, we focused mainly on the MAPK/ERK pathway. The confusion probably stems from the fact that the corresponding term in the KEGG pathway database is labeled "MAPK signaling pathway" and this term, although singular, includes all MAPK pathways. We have carefully reviewed the entire article and have corrected the term used accordingly to either: 1) MAPK pathways in general, 2) the MAPK/ERK pathway for this particular pathway, or 3) "MAPK signaling pathway", where the KEGG term is meant.

(5) Some eIF3 subunit RNAs have TOP motifs. One might expect 3e and 3h levels to change as a function of 3d knockdown due to TOP motifs but this is not observed. Can the authors speculate why the eIF3 subunit levels don't change but other TOP RNAs show TE changes? Is this true for other translation factors, or just for eIF3, or just for these subunits? Could the Western blot be out of linear range for the antibody or is there feedback affecting eIF3 levels differently than the other TOP RNAs, or a protein turnover mechanism to maintain eIF3 levels?

This is indeed a very interesting question. In addition to the mRNAs encoding ribosomal proteins, we examined all TOP mRNAs and added an additional sheet to the S2 supplemental spreadsheet with all TOP RNAs listed in (Philippe et al., 2020, PMID: 32094190). According to our Ribo-Seq data, we could expect to see increased protein levels of eIF3a and eIF3f in eIF3dKD and eIF3eKD, but this is not the case, as judged from extensive western blot analysis performed in (Wagner et. al 2016, PMID: 27924037). Indeed, we cannot rule out the involvement of a compensatory mechanism monitoring and maintaining the levels of eIF3 subunits at steady-state – increasing or decreasing them if necessary, which could depend on the TOP motif-mediated regulation. However, we think that in our KDs, all non-targeted subunits that lose their direct binding partner in eIF3 due to siRNA treatment become rapidly degraded. For example, co-downregulation of subunits d, k and l in eIF3eKD is very likely caused by protein degradation as a result of a loss of their direct binding partner – eIF3e. Since we showed that the yeast eIF3 complex assembles co-translationally (Wagner et. al 2020, PMID: 32589964), and there is no reason to think that mammalian eIF3 differs in this regard, our working hypothesis is that free subunits that are not promptly incorporated into the eIF3 complex are rapidly degraded, and the presence or absence of the TOP motif in the 5’ UTR of their mRNAs has no effect. As for the other TOP mRNAs, translation factors eEF1B2, eEF1D, eEF1G, eEF2 have significantly increased FPs in both eIF3dKD and eIF3eKD, but we did not check their protein levels by western blotting to conclude anything specific.

Author response:

The following is the authors’ response to the original reviews.

The detailed, thorough critique provided by the three reviewers is very much appreciated. We believe the manuscript is greatly improved by the changes we have made based on those reviews. The major changes are described below, followed by a point by point response.

Major Changes:

(1) We revised our model (old Fig. 10; new Fig. 9) to keep the explanation focused on the data shown in the current study. Specifically, references to GTP/GDP states of Rab3A and changes in the presynaptic quantum have been removed and the mechanisms depicted are confined to pre- or post-synaptic Rab3A participating in either controlling release of a trophic factor that regulates surface GluA2 receptors (pre- or postsynaptic) or directly affecting fusion of GluA2-receptor containing vesicles (postsynaptic).

(2) We replaced all cumulative density function plots and ratio plots, based on multiple quantile samples per cell, with box plots of cell means. This affects new Figures 1, 2, 3, 5, 6, 7 and 8. All references to “scaling,” “divergent scaling,” or “uniform scaling,” have been removed. New p values for comparison of means are provided above every box plot in Figures 1, 2, 3, 5, 6, 7 and 8. The number of cultures is provided in the figure legends.

(3) We have added frequency to Figures 1, 2 and 8. Frequency values overall are more variable, and the effect of activity blockade less robust, than for mEPSC amplitudes. We have added text indicating that the increase in frequency after activity blockade was significant in neurons from cultures prepared from WT in the Rab3A+/- colony but not cultures prepared from KO mice (Results, lines 143 to 147, new Fig. 1G. H). The TTX-induced increase in frequency was significant in the NASPM experiments before NASPM, but not after NASPM (Results, lines 231 to 233, new Fig. 3, also cultures from WT in Rab3A+/- colony). The homeostatic plasticity effect on frequency did not reach significance in WT on WT glia cultures or

WT on KO glia cultures, possibly due to the variability of frequency, combined with smaller sample sizes (Results, lines 400 to 403, new Fig. 8). In the cultures prepared from WT mice in the Rab3A+/Ebd colony, there was a trend towards higher frequency after TTX that did not reach statistical significance, and in cultures prepared from mutant mice, the p value was large, suggesting disruption of the effect, which appears to be due to an increase in frequency in untreated cultures, similar to the behavior of mEPSC amplitudes in neurons from mutant mice (Results, lines 161-167). In sum, the effect of activity on frequency requires Rab3A and Ca2+-permeable receptors, and is mimicked by the presence of the Rab3A Earlybird mutant. We have also added a discussion of these results (Discussion, lines 427-435). 

(4) In the revised manuscript we have added analysis of VGLUT1 levels for the same synaptic sites that we previously analyzed GluA2 levels, and these data are described in Results, lines 344 to 371, and appear in new Table 2. In contrast to previous studies, we did not find any evidence for an increase in VGLUT1 levels after activity blockade. We reviewed those studies to determine whether there might be differences in the experimental details that could explain the lack of effect we observed. In (De Gois et al., 2005), the authors measured mRNA and performed western blots to show increases in VGLUT1 after TTX treatment in older rat cortical cultures (DIV 19). The study performs immunofluorescence imaging of VGLUT1 but only after bicuculline treatment (it decreases), not after TTX treatment. In (Wilson et al.,

2005), the hippocampal cultures are treated with AP5, not TTX, and the VGLUT1 levels in immunofluorescence images are reported relative to synapsin I. That the type of activity blockade matters is illustrated by the failure of Wilson and colleagues to observe a consistent increase in VGLUT1/Synapsin ratio in cultures treated with AMPA receptor blockade (NBQX; supplementary information). These points have been added to the Discussion, lines 436 to 447.)

Reviewer #1:

(1) (model…is not supported by the data), (2) (The analysis of mEPSC data using quantile sampling…), (3) (…statistical analysis of CDFs suffers from n-inflation…), (4) (How does recording noise and the mEPSC amplitude threshold affect “divergent scaling?”) (5) (…justification for the line fits of the ratio data…), (7) (A comparison of p-values between conditions….) and (10) (Was VGLUT intensity altered in the stainings presented in the manuscript?)

The major changes we made, described above, address Reviewer #1’s points. The remaining points are addressed below.

(6) TTX application induces a significant increase in mEPSC amplitude in Rab3A-/- mice in two out of three data sets (Figs. 1 and 9). Hence, the major conclusion that Rab3A is required for homeostatic scaling is only partially supported by the data. 

The p values based on CDF comparisons were problematic, but the point we were making is that they were much larger for amplitudes measured in cultures prepared from Rab3A-/- mice (Fig. 1, p = 0.04) compared to those from cultures prepared from Rab3A+/+ mice (Fig. 1, p = 4.6 * 10-4). Now that we are comparing means, there are no significant TTX-induced effects on mEPSC amplitudes for Rab3A-/- data. However, acknowledging that some increase after activity blockade remains, we describe homeostatic plasticity as being impaired or not significant, rather than abolished, by loss of Rab3A, (Abstract, lines 37 to 39; Results, lines 141 to 143; Discussion, lines 415 to 418).

(8) There is a significant increase in baseline mEPSC amplitude in Rab3AEbd/Ebd (15 pA) vs. Rab3AEbd/+ (11 pA) cultures, but not in Rab3A-/- (13.6 pA) vs. Rab3A+/- (13.9 pA). Although the nature of scaling was different between Rab3AEbd/Ebd vs. Rab3AEbd/+ and Rab3AEbd/Ebd with vs. without TTX, the question arises whether the increase in mEPSC amplitude in Rab3AEbd/Ebd is Rab3A dependent. Could a Rab3A independent mechanism occlude scaling?

The Reviewer is concerned that the increase in mEPSC amplitude in the presence of the Rab3A point mutant may be through a ‘non-Rab3A’ mechanism (a concern raised by the lack of such effect in cultures from the Rab3A-/- mice), and secondly, that the already large mEPSC cannot be further increased by the homeostatic plasticity mechanism. It must always be considered that a mutant with an altered genetic sequence may bind to novel partners, causing activities that would not be either facilitated or inhibited by the original molecule. We have added this caveat to Results, lines 180 to 186 We added that a number of other manipulations, implicating individual molecules in the homeostatic mechanism, have caused an increase in mEPSC amplitude at baseline, potentially nonspecifically occluding the ability of activity blockade to induce a further increase (Results lines 186 to 189). Still, it is a strong coincidence that the novel activity of the mutant Rab3A would affect mEPSC amplitude, the same characteristic that is affected by activity blockade in a Rab3A dependent manner, a point which we added to Results, lines 189 to 191.

(9) Figure 4: NASPM appears to have a stronger effect on mEPSC frequency in the TTX condition vs. control (-40% vs -15%). A larger sample size might be necessary to draw definitive conclusions on the contribution of Ca2+-permeable AMPARs.

Our results, even with the modest sample size of 11 cells, are clear: NASPM does not disrupt the effect of TTX treatment on mEPSC amplitude (new Fig. 3A). It also looks like there is a greater magnitude effect of NAPSM on frequency in TTX-treated cells; we note this, but point out that nevertheless, these mEPSCs are not contributing to the increase in mEPSC amplitude (Results, lines 238-241). 

(11) The change in GluA2 area or fluorescence intensity upon TTX treatment in controls is modest. How does the GluA2 integral change?

We had reported that GluA2 area showed the most prominent increase following activity blockade, with intensity changing very little. When we examined the integral, it closely matched the change in area. We have added the values for integral to new Fig. 5 D, H; new Fig. 6 A-C; new Fig. 7 A-C and new Table 1 (for GluA2) and new Table 2 (for VGLUT1). These results are described in the text in the following places: Results, lines 289-292; 298-299; 311-319; 328-324). For VGLUT1, both area and intensity changed modestly, and the integral appeared to be a combination of the two, being higher in magnitude and resulting in smaller p values than either area or intensity (Results, lines 344-348; 353-359; new Table 2).

(12) The quantitative comparison between physiology and microscopy data is problematic. The authors report a mismatch in ratio values between the smallest mEPSC amplitudes and the smallest GluA2 receptor cluster sizes (l. 464; Figure 8). Is this comparison affected by the fluorescence intensity threshold? What was the rationale for a threshold of 400 a.u. or 450 a.u.? How does this threshold compare to the mEPSC threshold of 3 pA.

This concern is partially addressed by no longer comparing the rank ordered mEPSC amplitudes with the rank ordered GluA2 receptor characteristics. We had used multiple thresholds in the event that an experiment was not analyzable with the chosen threshold (this in fact happened for VGLUT1, see end of this paragraph). We created box plots of the mean GluA2 receptor cluster size, intensity and integral, for experiments in which we used all three thresholds, to determine if the effect of activity blockade was different depending on which threshold was applied, and found that there was no obvious difference in the results (Author response image 1). Nevertheless, since there is no need to use a different threshold for any of the 6 experiments (3 WT and 3KO), for new Figures 5, 6 and 7 we used the same threshold for all data, 450; described in Methods, lines 746 to 749. For VGLUT1 levels, it was necessary to use a different threshold for Rab3A+/+ Culture #1 (400), but a threshold of 200 for the other five experiments (Methods, lines 751-757). The VGLUT1 immunofluorescent sites in Culture #1 had higher levels overall, and the low threshold caused the entire AOI to be counted as the synapse, which clearly included background levels outside of the synaptic site. Conversely, to use a threshold of 400 on the other experiments meant that the synaptic site found by the automated measurement tool was much smaller that what was visible by eye. In our judgement it would have been meaningless to adhere to a single threshold for VGLUT1 data.

Author response image 1.

Using different thresholds does not substantially alter GluA2 receptor cluster size data. A) Rab3A+/+ Culture #1, size data for three different thresholds, depicted above each graph. B) Rab3A+/+ Culture #2, size data for three different thresholds, depicted above each graph. Note scale bar in A is different from B, to highlight differences for different thresholds. (Culture #3 was only analyzed with 450 threshold).

The conclusion that an increase in AMPAR levels is not fully responsible for the observed mEPSC increase is mainly based on the rank-order analysis of GluA2 intensity, yielding a slope of ~0.9. There are several points to consider here: (i) GluA2 fluorescence intensity did increase on average, as did GluA2 cluster size.

(ii) The increase in GluA2 cluster size is very similar to the increase in mEPSC amplitude (each approx. 1820%). (iii) Are there any reports that fluorescence intensity values are linearly reporting mEPSC amplitudes (in this system)? Antibody labelling efficiency, and false negatives of mEPSC recordings may influence the results. The latter was already noted by the authors.

Our comparison between mEPSC amplitude and GluA2 receptor cluster characteristics has been reexamined in the revised version using means rather than rank-ordered data in rank-order plots or ratio plots. Importantly, all of these methods revealed that in one out of three WT cultures (Culture #3) GluA2 receptor cluster size (old Fig. 8, old Table 1; new Fig. 6, new Table 1), intensity and integral (new Fig. 6, new Table 1) values decreased following activity blockade while in the same culture, mEPSC amplitudes increased. It is based on this lack of correspondence that we conclude that increases in mEPSC amplitude are not fully explained by increases in GluA2 receptors, and suggest there may be other contributors. These points are made in the Abstract (lines 108-110); Results (lines 319 to 326; 330337; 341-343) and the Discussion (lines 472 to 474). To our knowledge, there are not any reports that quantitatively compare receptor levels (area, intensity or integrals) to mEPSC amplitudes in the same cultures. We examined the comparisons very closely for 5 studies that used TTX to block activity and examined receptor levels using confocal imaging at identified synapses (Hou et al., 2008; Ibata et al., 2008; Jakawich et al., 2010a; Xu and Pozzo-Miller, 2017; Dubes et al., 2022). We were specifically looking for whether the receptor data were more variable than the mEPSC amplitude data, as we found. However, for 4 of the studies, sample sizes were very different so that we cannot simply compare the p values. Below is a table of the comparisons.

Author response table 1.

In Xu 2017 the sample sizes are close enough that we feel comfortable concluding that the receptor data were slightly more variable (p < 0.05) than mEPSC data (p<0.01) but recognize that it is speculative to say our finding has been confirmed. A discussion of these articles is in Discussion, lines 456-474.

(iv) It is not entirely clear if their imaging experiments will sample from all synapses. Other AMPAR subtypes than GluA2 could contribute, as could kainite or NMDA receptors.

While our imaging data only examined GluA2, we used the application of NASPM to demonstrate Ca2+permeable receptors did not contribute quantitatively to the increase in mEPSC amplitude following TTX treatment. Since GluA3 and GluA4 are also Ca2+-permeable, the findings in new Figure 3 (old Fig. 4) likely rule out these receptors as well.  There are also reports that Kainate receptors are Ca2+-permeable and blocked by NASPM (Koike et al., 1997; Sun et al., 2009), suggesting the NASPM experiment also rules out the contribution of Kainate receptors. Finally, given our recording conditions, which included normal magnesium levels in the extracellular solution as well as TTX to block action-potential evoked synaptic transmission, NMDA receptors would not be available to contribute currents to our recordings due to block by magnesium ions at resting Vm. These points have been added to the Methods section, lines 617 to 677 (NMDA); 687-694 (Ca2+-permeable AMPA receptors and Kainate receptors).

Furthermore, the statement “complete lack of correspondence of TTX/CON ratios” is not supported by the data presented (l. 515ff). First, under the assumption that no scaling occurs in Rab3A-/-, the TTX/CON ratios show a 20-30% change, which indicates the variation of this readout. Second, the two examples shown in Figure 8 for Rab3A+/+ are actually quite similar (culture #1 and #2, particularly when ignoring the leftmost section of the data, which is heavily affected by the raw values approaching zero.

We are no longer presenting ratio plots in the revised manuscript, so we do not base our conclusion that mEPSC amplitude data is not always corresponding to GluA2 receptor data on the difference in behavior of TTX/CON ratio values, but only on the difference in direction of the TTX effect in one out of three cultures. We agree with the reviewer that the ratio plots are much more sensitive to differences between control and treated values than the rank order plot, and we feel these differences are important, for example, there is still a homeostatic increase in the Rab3A-/- cultures, and the effect is still divergent rather than uniform. But the comparison of ratio data will be presented elsewhere.

(13) Figure 7A: TTX CDF was shifted to smaller mEPSC amplitude values in Rab3A-/- cultures. How can this be explained?

While this result is most obvious in CDF plots, we still observe a trend towards smaller mEPSC amplitudes after TTX treatment in two of three individual cultures prepared from Rab3A-/- mice when comparing means (new Fig. 7, Table 1) which did not reach statistical significance for the pooled data (new Fig. 5, new Table 1). There was not any evidence of this decrease in the larger data set (new Fig. 1) nor for Rab3A-/- neurons on Rab3A+/+ glia (new Fig. 8). Given that this effect is not consistent, we did not comment on it in the revised manuscript. It may be that there is a non-Rab3A-dependent mechanism that results in a decrease in mEPSC amplitude after activity blockade, which normally pulls down the magnitude of the activity-dependent increase typically observed. But studying this second component would be difficult given its magnitude and inconsistent presentation.

Reviewer #1 (Recommendations For the Authors):

(1) Abstract, last sentence: The conclusion of the present manuscript should be primarily based on the results presented. At present, it is mainly based on a previous publication by the authors.

We have revised the last sentence to reflect actual findings of the current study (Abstract, lines 47 to 49).

(2) Line 55: “neurodevelopmental”

This phrase has been removed.

(3) Line 56: “AMPAergic” should be replaced by AMPAR-mediated

This sentence was removed when all references to “scaling” were removed; no other instances of “AMPAergic” are present.

(4) Figure 9: The use of BioRender should be disclosed in the Figure Legend.

We used BioRender in new Figures 3, 7 and 8, and now acknowledge BioRender in those figure legends.

(5) Figure legends and results: The number of cultures should be indicated for each comparison.

Number of cultures has been added to the figure legends.

(6) Line 289: A comparison of p-values between conditions does not allow any meaningful conclusions.

Agreed, therefore we have removed CDFs and the KS test comparison p values. All comparisons in the revised manuscript are for cell means.

(7) Line 623ff: The argument referring to NMJ data is weak, given that different types of receptors are involved.

We still think it is valid to point out that Rab3A is required for the increase in mEPC at the NMJ but that ACh receptors do not increase (Discussion, lines 522 to 525). We are not saying that postsynaptic receptors do not contribute in cortical cultures, only that there could be another Rab3A-dependent mechanism that also affects mEPSC amplitude.

(8) Plotting data points outside of the ranges should be avoided (e.g., Fig. 2Giii, 7F).

These two figures are no longer present in the revised manuscript. In revising figures, we made sure no other plots have data points outside of the ranges.

(9) The rationale for investigating Rab3AEbd/Ebd remains elusive and should be described.

A rationale for investigating Rab3AEbd/Ebd is that if the results are similar to the KO, it strengthens the evidence for Rab3A being involved in homeostatic synaptic plasticity. In addition, since its phenotype of early awakening was stronger than that demonstrated in Rab3A KO mice (Kapfhamer et al., 2002), it was possible we would see a more robust effect. These points have been added to the Results, lines 118 to 126.

(10) Figures 3 and 4, as well as Figure 5 and 6 could be merged.

In the revised version, Figure 3 has been eliminated since its main point was a difference in scaling behavior. Figure 4 has been expanded to include a model of how NASPM could reduce frequency (new Fig. 3.) Images of the pyramidal cell body have been added to Figure 5 (new Fig. 4), and Figure 6 has been completely revised and now includes pooled data for both Rab3A+/+ and Rab3A-/- cultures, for mEPSC amplitude, GluA2 receptor cluster size, intensity and integral.

(11) Figure 5: The legend refers to MAP2, but this is not indicated in the figure.

MAP2 has now been added to the labels for each image and described in the figure legend (new Fig. 4).

Reviewer #2:

Technical concerns:

(1) The culture condition is questionable. The authors saw no NMDAR current present during spontaneous recordings, which is worrisome since NMDARs should be active in cultures with normal network activity (Watt et al., 2000; Sutton et al., 2006). It is important to ensure there is enough spiking activity before doing any activity manipulation. Similarly it is also unknown whether spiking activity is normal in Rab3AKO/Ebd neurons.

In the studies cited by the reviewer, NMDA currents were detected under experimental conditions in which magnesium was removed. In our recordings, we have normal magnesium (1.3 mM) and also TTX, which prevents the necessary depolarization to allow inward current through NMDA receptors. This point has been added to our Methods, lines 674 to 677. We acknowledge we do not know the level of spiking in cultures prepared from Rab3A+/+, Rab3A-/- or Rab3A_Ebd/Ebd_ mice. Given the similar mEPSC amplitude for untreated cultures from WT and KO studies, we think it unlikely that activity was low in the latter, but it remains a possibility for untreated cultures from Rab3A_Ebd/Ebd_ mice, where mEPSC amplitude was increased. These points are added to the Methods, lines 615 to 622.

(2) Selection of mEPSC events is not conducted in an unbiased manner. Manually selecting events is insufficient for cumulative distribution analysis, where small biases could skew the entire distribution. Since the authors claim their ratio plot is a better method to detect the uniformity of scaling than the well-established rank-order plot, it is important to use an unbiased population to substantiate this claim.

We no longer include any cumulative distributions or ratio plot analysis in the revised version. We have added the following text to Methods, lines 703 to 720:

“MiniAnalysis selects many false positives with the automated feature when a small threshold amplitude value is employed, due to random fluctuations in noise, so manual re-evaluation of the automated process is necessary to eliminate false positives. If the threshold value is set high, there are few false positives but small amplitude events that visually are clearly mEPSCs are missed, and manual re-evaluation is necessary to add back false negatives or the population ends up biased towards large mEPSC amplitudes. As soon as there is a manual step, bias is introduced. Interestingly, a manual reevaluation step was applied in a recent study that describes their process as ‘unbiased (Wu et al., 2020). In sum, we do not believe it is currently possible to perform a completely unbiased detection process. A fully manual detection process means that the same criterion (“does this look like an mEPSC?”) is applied to all events, not just the false positives, or the false negatives, which prevents the bias from being primarily at one end or the other of the range of mEPSC amplitudes. It is important to note that when performing the MiniAnalysis process, the researcher did not know whether a record was from an untreated cell or a TTX-treated cell.”

(3) Immunohistochemistry data analysis is problematic. The authors only labeled dendrites without doing cell-fills to look at morphology, so it is questionable how they differentiate branches from pyramidal neurons and interneurons. Since glutamatergic synapse on these two types of neuron scale in the opposite directions, it is crucial to show that only pyramidal neurons are included for analysis.

We identified neurons with a pyramidal shape and a prominent primary dendrite at 60x magnification without the zoom feature. This should have been made clear in the description of imaging. We have added an image of the two selected cells to our figure of dendrites (old Fig. 5, new Fig. 4), and described this process in the Methods, lines 736 to 739, and Results, lines 246 to 253. Given the morphology of the neurons selected it is highly unlikely that the dendrites we analyzed came from interneurons.

Conceptual Concerns

The only novel finding here is the implicated role for Rab3A in synaptic scaling, but insights into mechanisms behind this observation are lacking. The authors claim that Rab3A likely regulates scaling from the presynaptic side, yet there is no direct evidence from data presented. In its current form, this study’s contribution to the field is very limited.

We have demonstrated that loss of Rab3A and expression of a Rab3A point mutant disrupt homeostatic plasticity of mEPSC amplitudes, and that in the absence of Rab3A, the increase in GluA2 receptors at synaptic sites is abolished. Further, we show that this effect cannot be through release of a factor, like TNFα, from astrocytes. In the new version, we add the finding that VGLUT1 is not increased after activity blockade, ruling out this presynaptic factor as a contributor to homeostatic increases in mEPSC amplitude. We show for the first time by examining mEPSC amplitudes and GluA2 receptors in the same cultures that the increases in GluA2 receptors are not as consistent as the increases in mEPSC amplitude, suggesting the possibility of another contributor to homeostatic increases in mEPSC amplitude. We first proposed this idea in our previous study of Rab3A-dependent homeostatic increases in mEPC amplitudes at the mouse neuromuscular junction. In sum, we dispute that there is only one novel finding and that we have no insights into mechanism. We acknowledge that we have no direct evidence for regulation from the presynaptic side, and have removed this claim from the revised manuscript. We have retained the Discussion of potential mechanisms affecting the presynaptic quantum and evidence that Rab3A is implicated in these mechanisms (vesicle size, fusion pore kinetics; Discussion, lines 537 to 563). One way to directly show that the amount of transmitter released for an mEPSC has been modified after activity blockade is to demonstrate that a fast off-rate antagonist has become less effective at inhibiting mEPSCs (because the increased glutamate released out competes it; see (Liu et al., 1999) and (Wilson et al., 2005) for example experiments). This set of experiments is underway but will take more time than originally expected, because we are finding surprisingly large decreases in frequency, possibly the result of mEPSCs with very low glutamate concentration that are completely inhibited by the dose used. Once mEPSCs are lost, it is difficult to compare the mEPSC amplitude before and after application of the antagonist. Therefore we intend to include this experiment in a future report, once we determine the reason for the frequency reduction, or, can find a dose where this does not occur.

(1) Their major argument for this is that homeostatic effects on mEPSC amplitudes and GluA2 cluster sizes do not match. This is inconsistent with reports from multiple labs showing that upscaling of mEPSC amplitude and GluA2 accumulation occur side by side during scaling (Ibata et al., 2008; Pozo et al., 2012; Tan et al., 2015; Silva et al., 2019). Further, because the acquisition and quantification methods for mEPSC recordings and immunohistochemistry imaging are entirely different (each with its own limitations in signal detection), it is not convincing that the lack of proportional changes must signify a presynaptic component.

Within the analyses in the revised manuscript, which are now based only on comparison of cell/dendrite means, we find a very good match in the magnitude of increase for the pooled data of mEPSC amplitudes and GluA2 receptor cluster sizes (+19.7% and +20.0% respectively; new Table 1). However, when looking at individual cultures, we had one of three WT cultures in which mEPSC amplitude increased 17.2% but GluA2 cluster size decreased 9.5%. This result suggests that while activity blockade does lead to an increase in GluA2 receptors after activity blockade, the effect is more variable than that for mEPSC amplitude. We went back to published studies to see if this has been previously observed, but found that it was difficult to compare because the sample sizes were different for the two characteristics (see Author response table 1). We included these particular 5 studies because they use the same treatment (TTX), examine receptors using imaging of identified synaptic sites, and record mEPSCs in their cultures (although the authors do not indicate that imaging and recordings are done simultaneously on the same cultures.) Only one of the studies listed by the Reviewer is in our group (Ibata et al., 2008). The study by (Tan et al., 2015) uses western blots to measure receptors; the study by (Silva et al., 2019) blocks activity using a combination of AMPA and NMDA receptor blockers; the study by (Pozo et al., 2012) correlates mEPSC amplitude changes with imaging but not in response to activity blockade, instead for changing the expression of GluA2. While it may seem like splitting hairs to reject studies that use other treatment protocols, there is ample evidence that the mechanisms of homeostatic plasticity depend on how activity was altered, see the following studies for several examples of this (Sutton et al., 2006; Soden and Chen, 2010; Fong et al., 2015). A discussion of the 5 articles we selected is in the revised manuscript, Discussion, lines 456 to 474. In sum, we provide evidence that activity blockade is associated with an overall increase in GluA2 receptors; what we propose is that this increase, being more variable, does not fully explain the increase in mEPSC amplitude. However, we acknowledge that the disparity could be explained by the differences in limitations of the two methods (Discussion, lines 469-472).

(2) The authors also speculate in the discussion that presynaptic Rab3A could be interacting with retrograde BDNF signaling to regulate postsynaptic AMPARs. Without data showing Rab3A-dependent presynaptic changes after TTX treatment, this argument is not compelling. In this retrograde pathway, BDNF is synthesized in and released from dendrites (Jakawich et al., 2010b; Thapliyal et al., 2022), and it is entirely possible for postsynaptic Rab3A to interfere with this process cell-autonomously.

We have added the information that Rab3A could control BDNF from the postsynaptic cell and included the two references provided by the reviewer, Discussion, lines 517 to 518. We have added new evidence, recently published, that the Rab3 family has been shown to regulate targeting of EGF receptors to rafts (among other plasma membrane molecules), with Rab3A itself clearly present in nonneuronal cells (Diaz-Rohrer et al., 2023) (added to Discussion, lines 509 to 515).

(3) The authors propose that a change in AMPAR subunit composition from GluA2-containing ones to GluA1 homomers may account for the distinct changes in mEPSC amplitudes and GluA2 clusters. However, their data from the NASPM wash-in experiments clearly show that the GluA1 homomer contributions have not changed before and after TTX treatment.

We have revised this section in the Discussion, lines 534 to 536, to clarify that any change due to GluA1 homomers should have been detectable by a greater ability of NASPM to reverse the TTX-induced increase.

Reviewer #2 (Recommendations for the Authors):

For authors to have more convincing arguments in general, they will need to clarify/improve certain details in their data collection by addressing the above technical concerns. Additionally, the authors should design experiments to test whether Rab3A regulates scaling from pre- or post-synaptic site. For example, they could sparsely knock out Rab3A in WT neurons to test the postsynaptic possibility. On the other hand, their argument for a presynaptic role would be much more compelling if they could show whether there are clear functional changes such as in vesicle sizes and release probability in the presynaptic terminal of Rab3AKO neurons.

An important next step is to identify whether Rab3A is acting pre- or post-synaptically (Discussion, lines 572 to 573), but these experiments will be undertaken in the future. It would not add much to simply show vesicle size is altered in the KO (and we do not necessarily expect this since mEPSC amplitude is normal in the KO). It will be very difficult to establish that vesicle size is changing with activity blockade and that this change is prevented in the Rab3A KO, because we are looking for a ~25% increase in vesicle volume, which would correspond to a ~7.5% increase in diameter. Finally, we do not believe demonstrating changes in release probability tell us anything about a presynaptic role for Rab3A in regulating the size of the presynaptic quantum.

Reviewer #3 (Public Review)

Weaknesses: However, the rather strong conclusions on the dissociation of AMPAR trafficking and synaptic response are made from somewhat weaker data. The key issue is the GluA2 immunostaining in comparison with the mEPSC recordings. Their imaging method involves only assessing puncta clearly associated with a MAP2 labeled dendrite. This is a small subset of synapses, judging from the sample micrographs (Fig. 5). To my knowledge, this is a new and unvalidated approach that could represent a particular subset of synapses not representative of the synapses contributing to the mEPSC change (they are also sampling different neurons for the two measurements; an additional unknown detail is how far from the cell body were the analyzed dendrites for immunostaining.) While the authors acknowledge that a sampling issue could explain the data, they still use this data to draw strong conclusions about the lack of AMPAR trafficking contribution to the mEPSC amplitude change. This apparent difference may be a methodological issue rather than a biological one, and at this point it is impossible to differentiate these. It will unfortunately be difficult to validate their approach. Perhaps if they were to drive NMDAdependent LTD or chemLTP, and show alignment of the imaging and ephys, that would help. More helpful would be recordings and imaging from the same neurons but this is challenging. Sampling from identified synapses would of course be ideal, perhaps from 2P uncaging combined with SEP-labeled AMPARs, but this is more challenging still. But without data to validate the method, it seems unwarranted to make such strong conclusions such as that AMPAR trafficking does not underlie the increase in mEPSC amplitude, given the previous data supporting such a model.

In the new version, we soften our conclusion regarding the mismatch between GluA2 receptor levels and mEPSC amplitudes, now only stating that receptors may not be the sole contributor to the TTX effect on mEPSC amplitude (Discussion, lines 472 to 474). With our analysis in the new version focusing on comparisons of cell means, the GluA2 receptor cluster size and the mEPSC amplitude data match well in magnitude for the data pooled across the 3 matched cultures (20.0% and 19.7%, respectively, see new Table 1). However, in one of the three cultures the direction of change for GluA2 receptors is opposite that of mEPSC amplitudes (Table 1, Culture #3, -9.5% vs +17.2%, respectively).

It is unlikely that the lack of matching of homeostatic plasticity in one culture, but very good matching in two other cultures, can be explained by an unvalidated focus on puncta associated with MAP2 positive dendrites. We chose to restrict analysis of synaptic GluA2 receptors to the primary dendrite in order to reduce variability, reasoning that we are always measuring synapses for an excitatory pyramidal neuron, synapses that are relatively close to the cell body, on the consistently identifiable primary dendrite. We measured how far this was for the two cells depicted in old Figure 5 (new Fig. 4). Because we always used the 5X zoom window which is a set length, and positioned it within ~10 microns of the cell body, these cells give a ball park estimate for the usual distances. For the untreated cell, the average distance from the cell body was 38.5 ± 2.8 µm; for the TTX-treated cell, it was 42.4 ± 3.2 µm (p = 0.35, KruskalWallis test). We have added these values to the Results, lines 270 to 274.

We did not mean to propose that AMPA receptor levels do not contribute at all to mEPSC amplitude, and we acknowledge there are clear cases where the two characteristics change in parallel (for example, in the study cited by Reviewer #2, (Pozo et al., 2012), increases in GluA2 receptors due to exogenous expression are closely matched by increases in mEPSC amplitudes.) What our matched culture experiments demonstrate is that in the case of TTX treatment, both GluA2 receptors and mEPSC amplitudes increase on average, but sometimes mEPSC amplitudes can increase in the absence of an increase in GluA2 receptors (Culture #3, Rab3A+/+ cultures), and sometimes mEPSC amplitudes do not increase even though GluA2 receptor levels do increase (Culture #3, Rab3A-/- cultures). Therefore, it would not add anything to our argument to examine receptors and mEPSCs in NMDA-dependent LTP, a different plasticity paradigm in which changes in receptors and mEPSCs may more closely align. It has been demonstrated that mEPSCs of widely varying amplitude can be recorded from a single synaptic site (Liu and Tsien, 1995), so we would need to measure a large sample of individual synapse recordings to detect a modest shift in average values due to activity blockade. In addition, it would be essential to express fluorescent AMPA receptors in order to correlate receptor levels in the same cells we record from (or at the same synapses). And yet, even after these heroics, one is still left with the issue that the two methods, electrophysiology and fluorescent imaging, have distinct limitations and sources of variability that may obscure any true quantitative correlation.

Other questions arise from the NASPM experiments, used to justify looking at GluA2 (and not GluA1) in the immunostaining. First, there is a frequency effect that is quite unclear in origin. One would expect NASPM to merely block some fraction of the post-synaptic current, and not affect pre-synaptic release or block whole synapses. It is also unclear why the authors argue this proves that NASPM was at an effective concentration (lines 399-400). Further, the amplitude data show a strong trend towards smaller amplitude. The p value for both control and TTX neurons was 0.08 – it is very difficult to argue that there is no effect. And the decrease is larger in the TTX neurons. Considering the strong claims for a presynaptic locus and the use of this data to justify only looking at GluA2 by immunostaining, these data do not offer much support of the conclusions. Between the sampling issues and perhaps looking at the wrong GluA subunit, it seems premature to argue that trafficking is not a contributor to the mEPSC amplitude change, especially given the substantial support for that hypothesis. Further, even if trafficking is not the major contributor, there could be shifts in conductance (perhaps due to regulation of auxiliary subunits) that does not necessitate a pre-synaptic locus. While the authors are free to hypothesize such a mechanism, it would be prudent to acknowledge other options and explanations.

We have created a model cartoon to explain how NASPM could reduce mEPSC frequency (new Fig. 3D). mEPSCs that arise from a synaptic site that has only Ca2+-permeable AMPA receptors will be completely blocked by NASPM, if the NASPM concentration is maximal. The reason we conclude that we have sufficient NASPM reaching the cells is that the frequency is decreased, as expected if there are synaptic sites with only Ca2+-permeable AMPA receptors. We previously were not clear that there is an effect of NASPM on mEPSC amplitude, although it did not reach statistical significance (new Fig. 3B). Where there is no effect is on the TTX-induced increase in mEPSC amplitude, which remains after the acute NASPM application (new Fig. 3A). We have revised the description of these findings in Results, lines 220 to 241. In reviewing the literature further, we could find no previous studies demonstrating an increase in conductance in GluA2 or Ca2+-impermeable receptors, only in GluA1 homomers. In other words, any conductance change would have been due to a change in GluA1 homomers, and should have been visible as a disruption of the homeostatic plasticity by NASPM application. We have added text to Results, lines 211 to 217; 236-241; Discussion, lines 420 to 422; 526-536 and Methods, lines 685 to 695 regarding this point.

The frequency data are missing from the paper, with the exception of the NASPM dataset. The mEPSC frequencies should be reported for all experiments, particularly given that Rab3A is generally viewed as a pre-synaptic protein regulating release. Also, in the NASPM experiments, the average frequency is much higher in the TTX treated cultures. Is this statistically above control values?

This comment is addressed by the major change #3, above.

Unaddressed issues that would greatly increase the impact of the paper:

(1) Is Rab3A activity pre-synaptically, post-synaptically or both. The authors provide good evidence that Rab3A is acting within neurons and not astrocytes. But where is it acting (pre or post) would aid substantially in understanding its role (and particularly the hypothesized and somewhat novel idea that the amount of glutamate released per vesicle is altered in HSP). They could use sparse knockdown of Rab3A, or simply mix cultures from KO and WT mice (with appropriate tags/labels). The general view in the field has been that HSP is regulated post-synaptically via regulation of AMPAR trafficking, and considerable evidence supports this view. The more support for their suggestion of a pre-synaptic site of control, the better.

This is similar to the request of Reviewer #2, Recommendations to the Authors. An important next step is to identify whether Rab3A is working pre- or postsynaptically. However, it is possible that it is acting pre-synaptically to anterogradely regulate trafficking of AMPAR, as we have depicted in our model, new Fig. 9. To demonstrate that the presynaptic quantum is being altered, we would need to show that vesicle size is increased, or the amount of transmitter being released during an mEPSC is increased after activity blockade. To that end, we are currently performing experiments using a fast off-rate antagonist. As described above in response to Reviewer #2’s Conceptual Concerns, we find dramatic decreases in frequency not explained by the 30-60% inhibition observed for the largest amplitude mEPSCs, which suggests the possibility that small mEPSCs are more sensitive than large mEPSCs and therefore may have less transmitter. Due to these complexities and the delay while we test other antagonists to see if the effect is specific to fast-off rate antagonists, we are not including these results here.

(2) Rab3A is also found at inhibitory synapses. It would be very informative to know if HSP at inhibitory synapses is similarly affected. This is particularly relevant as at inhibitory synapses, one expects a removal of GABARs and/or a decrease of GABA-packaging in vesicles (ie the opposite of whatever is happening at excitatory synapses.). If both processes are regulated by Rab3A, this might suggest a role for this protein more upstream in the signaling, an effect only at excitatory synapses would argue for a more specific role just at these synapses.

It will be important to determine if homeostatic synaptic plasticity at inhibitory synapses on excitatory neurons is sensitive to Rab3A deletion, especially in light of the fact that unlike many of the other molecules implicated in homeostatic increases in mEPSCS, Rab3A is not a molecule known to be selective for glutamate receptor trafficking (in contrast to Arc/Arg3.1 or GRIP1, for example). Such a study would warrant its own publication.

Reviewer #3 (Recommendations for the Authors):

There are a number of minor points or suggestions for the authors:

Is RIM1 part of this pathway (or expected to be)? Some discussion of this would be nice.

RIM, Rab3-interacting molecule, has been implicated at the drosophila neuromuscular junction in a presynaptic form of homeostatic synaptic plasticity in which evoked release is increased after block of postsynaptic receptors (Muller et al., 2012), a plasticity that also requires Rab3-GAP (Muller et al., 2011). To our knowledge there is no evidence that RIM is involved in the homeostatic plasticity of mEPSC amplitude after activity blockade by TTX. The Rim1a KO does not have a change in mEPSC amplitude relative to WT (Calakos et al., 2004), but that is not unexpected given the normal mEPSC amplitude in neurons from cultures prepared from Rab3A-/- mice in the current study. It would be interesting to look at homeostatic plasticity in cortical cultures prepared from Rim1a or other RIM deletion mice, but we have not added these points to the revised manuscript since there are a number of directions one could go in attempting to define the molecular pathway and we feel it is more important to discuss the potential location of action and physiological mechanisms.

Is the Earlybird mutation a GOF? More information about this mutation would help.

We have added a description of how the Earlybird mutation was identified, in a screen for rest:activity mutants (Results, lines 118 to 123). Rab3A Earlybird mice have a shortened circadian period, shifting their wake cycle earlier and earlier. When Rab3A deletion mice were tested in the same activity raster plot measurements, the shift was smaller than that for the Earlybird mutant, suggesting the possibility that it is a dominant negative mutation.

The high K used in the NASPM experiments seems a bit unusual. Have the authors done high K/no drug controls to see if this affects the synapses in any way?

We used the high K based on previous studies that indicated the blocking effect of the Ca2+-permeable receptor blockers was use dependent (Herlitze et al., 1993; Iino et al., 1996; Koike et al., 1997). We reasoned that a modest depolarization would increase the frequency of AMPA receptor mEPSCs and allow access of the NASPM.  We have added this point to the Methods, lines 695 to 708. 

The NASPM experiments do not show that GluA1 does not contribute (line 401), only that GluA1 homomers are not contributing (much – see above). GluA1/A2 heteromers are quite likely involved. Also, the SEM is missing from the WT pre/post NASPM data.

Imaging of GluA2-positive sites will not distinguish between GluA2 homomers and GluA2-GluA1 heteromers, so we have added this clarification to Results, lines 242 to 246. We have remade the NASPM pre-post line plots so that the mean values and error bars are more visible (new Fig. 3B, C).

It seems odd to speculate based on non-significant findings (line 650-1), with lower significance (p = 0.11) than findings being dismissed in the paper (NASPM on mEPSC amplitude; p = 0.08).

We did not mean to dismiss the effect of NASPM on mEPSC amplitude (new Fig. 3B), rather, we dismiss the effect of NASPM on the homeostatic increase in mEPSC amplitude caused by TTX treatment (new Fig. 3A). We have emphasized this distinction in Results, lines 223 to 225, and Discussion, lines 420 to 422, as well as adding that the stronger effect of NASPM on frequency after TTX treatment suggests an activity-dependent increase in the number of synapses expressing only Ca2+ permeable homomers (Results, lines 236 to 241; Discussion, lines 431 to 435).

Fig. 4 could be labeled better (to make it clear that B is amplitude and C is freq from the same cells).

Fig. 4 has been revised—now the amplitude and frequency plots from the same condition (new Fig. 3, B, C; CON or TTX) are in a vertical line and the figure legend states that the frequency data are from the same cells as in Fig. 3A.

The raw amplitude data seems a bit hidden in the inset panels – I would suggest these data are at least as important as the cumulative distributions in the main panel. Maybe re-organizing the figures would help.

We have removed all cumulative distributions, rank order plots, and ratio plots. The box plots are now full size in new Figures 1, 2, 5, 6, 7 and 8.

I’m not sure I would argue in the paper that 12 cells a day is a limiting issue for experiments. It doesn’t add anything and doesn’t seem like that high a barrier. It is fine to just say it is difficult and therefore there is a limited amount of data meeting the criteria.

We have removed the comment regarding difficulty.

Calakos N, Schoch S, Sudhof TC, Malenka RC (2004) Multiple roles for the active zone protein RIM1alpha in late stages of neurotransmitter release. Neuron 42:889-896.

De Gois S, Schafer MK, Defamie N, Chen C, Ricci A, Weihe E, Varoqui H, Erickson JD (2005) Homeostatic scaling of vesicular glutamate and GABA transporter expression in rat neocortical circuits. J Neurosci 25:7121-7133.

Diaz-Rohrer B, Castello-Serrano I, Chan SH, Wang HY, Shurer CR, Levental KR, Levental I (2023) Rab3 mediates a pathway for endocytic sorting and plasma membrane recycling of ordered microdomains. Proc Natl Acad Sci U S A 120:e2207461120.

Dubes S, Soula A, Benquet S, Tessier B, Poujol C, Favereaux A, Thoumine O, Letellier M (2022) miR-124dependent tagging of synapses by synaptopodin enables input-specific homeostatic plasticity. EMBO J 41:e109012.

Fong MF, Newman JP, Potter SM, Wenner P (2015) Upward synaptic scaling is dependent on neurotransmission rather than spiking. Nat Commun 6:6339.

Herlitze S, Raditsch M, Ruppersberg JP, Jahn W, Monyer H, Schoepfer R, Witzemann V (1993) Argiotoxin detects molecular differences in AMPA receptor channels. Neuron 10:1131-1140.

Hou Q, Zhang D, Jarzylo L, Huganir RL, Man HY (2008) Homeostatic regulation of AMPA receptor expression at single hippocampal synapses. Proc Natl Acad Sci U S A 105:775-780.

Ibata K, Sun Q, Turrigiano GG (2008) Rapid synaptic scaling induced by changes in postsynaptic firing. Neuron 57:819-826.

Iino M, Koike M, Isa T, Ozawa S (1996) Voltage-dependent blockage of Ca(2+)-permeable AMPA receptors by joro spider toxin in cultured rat hippocampal neurones. J Physiol 496 ( Pt 2):431437.

Jakawich SK, Neely RM, Djakovic SN, Patrick GN, Sutton MA (2010a) An essential postsynaptic role for the ubiquitin proteasome system in slow homeostatic synaptic plasticity in cultured hippocampal neurons. Neuroscience 171:1016-1031.

Jakawich SK, Nasser HB, Strong MJ, McCartney AJ, Perez AS, Rakesh N, Carruthers CJ, Sutton MA (2010b) Local presynaptic activity gates homeostatic changes in presynaptic function driven by dendritic BDNF synthesis. Neuron 68:1143-1158.

Kapfhamer D, Valladares O, Sun Y, Nolan PM, Rux JJ, Arnold SE, Veasey SC, Bucan M (2002) Mutations in Rab3a alter circadian period and homeostatic response to sleep loss in the mouse. Nat Genet 32:290-295.

Koike M, Iino M, Ozawa S (1997) Blocking effect of 1-naphthyl acetyl spermine on Ca(2+)-permeable AMPA receptors in cultured rat hippocampal neurons. Neurosci Res 29:27-36.

Liu G, Tsien RW (1995) Properties of synaptic transmission at single hippocampal synaptic boutons. Nature 375:404-408.

Liu G, Choi S, Tsien RW (1999) Variability of neurotransmitter concentration and nonsaturation of postsynaptic AMPA receptors at synapses in hippocampal cultures and slices. Neuron 22:395409.

Muller M, Pym EC, Tong A, Davis GW (2011) Rab3-GAP controls the progression of synaptic homeostasis at a late stage of vesicle release. Neuron 69:749-762.

Muller M, Liu KS, Sigrist SJ, Davis GW (2012) RIM controls homeostatic plasticity through modulation of the readily-releasable vesicle pool. J Neurosci 32:16574-16585.

Pozo K, Cingolani LA, Bassani S, Laurent F, Passafaro M, Goda Y (2012) beta3 integrin interacts directly with GluA2 AMPA receptor subunit and regulates AMPA receptor expression in hippocampal neurons. Proc Natl Acad Sci U S A 109:1323-1328.

Silva MM, Rodrigues B, Fernandes J, Santos SD, Carreto L, Santos MAS, Pinheiro P, Carvalho AL (2019) MicroRNA-186-5p controls GluA2 surface expression and synaptic scaling in hippocampal neurons. Proc Natl Acad Sci U S A 116:5727-5736.

Soden ME, Chen L (2010) Fragile X protein FMRP is required for homeostatic plasticity and regulation of synaptic strength by retinoic acid. J Neurosci 30:16910-16921.

Sun HY, Bartley AF, Dobrunz LE (2009) Calcium-permeable presynaptic kainate receptors involved in excitatory short-term facilitation onto somatostatin interneurons during natural stimulus patterns. J Neurophysiol 101:1043-1055.

Sutton MA, Ito HT, Cressy P, Kempf C, Woo JC, Schuman EM (2006) Miniature neurotransmission stabilizes synaptic function via tonic suppression of local dendritic protein synthesis. Cell 125:785-799.

Tan HL, Queenan BN, Huganir RL (2015) GRIP1 is required for homeostatic regulation of AMPAR trafficking. Proc Natl Acad Sci U S A 112:10026-10031.

Thapliyal S, Arendt KL, Lau AG, Chen L (2022) Retinoic acid-gated BDNF synthesis in neuronal dendrites drives presynaptic homeostatic plasticity. Elife 11.

Wilson NR, Kang J, Hueske EV, Leung T, Varoqui H, Murnick JG, Erickson JD, Liu G (2005) Presynaptic regulation of quantal size by the vesicular glutamate transporter VGLUT1. J Neurosci 25:62216234.

Wu YK, Hengen KB, Turrigiano GG, Gjorgjieva J (2020) Homeostatic mechanisms regulate distinct aspects of cortical circuit dynamics. Proc Natl Acad Sci U S A 117:24514-24525.

Xu X, Pozzo-Miller L (2017) EEA1 restores homeostatic synaptic plasticity in hippocampal neurons from Rett syndrome mice. J Physiol 595:5699-5712.

Author Response

The following is the authors’ response to the original reviews.

eLife assessment

This important study enhances our understanding of the effects of landscape context on grassland plant diversity and biomass. Notably, the authors use a well-designed field sampling method to separate the effects of habitat loss and fragmentation per se. Most of the data and analyses provide solid support for the findings that habitat loss weakens the positive relationship between grassland plant richness and biomass.

Response: Thanks very much for organizing the review of the manuscript. We are grateful to you for the recognition. We have carefully analyzed all comments of the editors and reviewers and revised our manuscript to address them. All comments and recommendations are helpfully and constructive for improving our manuscript. We have described in detail our response to each of comment below.

In addition to the reviewers' assessments, we have the following comments on your paper.

(1) Some of the results are not consistent between figures. The relationships between overall species richness and fragmentation per se are not consistent between Figs. 3 and 5. The relationships between aboveground biomass and habitat loss are not consistent between Figs. 4 and 5. How shall we interpret these inconsistent results?

Response: Thanks for your insightful comments. The reason for these inconsistencies is that the linear regression model did not take into account the complex causal relationships (including direct and indirect effects) among the different influencing factors. The results in Figures 3 and 4 just represent the pairwise relationship pattern and relative importance, respectively. The causal effects of habitat loss and fragmentation per se on plant richness and above-ground biomass should be interpreted based on the structural equation model results (Figure 6). We have revised the data analysis to clear these inconsistent results. Line 225-228

In the revised manuscript, we have added the interpretation for these inconsistent results. The inconsistent effects between Figures 3 and 6 suggest that fragmentation per se actually had a positive effect on plant richness after accounting for the effects of habitat loss and environmental factors simultaneously.

The inconsistent effects between Figures 4 and 6 are because the effects of habitat loss and fragmentation per se on above-ground biomass were mainly mediated by plant richness and environmental factors, which had no significant direct effect (Figure 6). Thus, habitat loss and fragmentation per se showed no significant relative effects on above-ground biomass after controlling the effects of plant richness and environmental factors (Figure 4).

(2) One of the fragmentation indices, mean patch area metric, seems to be more appropriate as a measure of habitat loss, because it represents "a decrease in grassland patch area in the landscape".

Response: Thanks for your insightful comments. We apologize for causing this confusion. The mean patch area metric in our study represents the mean size of grassland patches in the landscape for a given grassland amount. Previous studies have often used the mean patch metric as a measure of fragmentation, which can reflect the processes of local extinction in the landscape (Fahrig, 2003; Fletcher et al., 2018). We have revised the definition of the mean patch area metric and added its ecological implication in the revised manuscript to clarify this confusion.

(3) It is important to show both the mean and 95% CI (or standard error) of the slope coefficients regarding to Figs. 3 and 6.

Response: Thanks for your suggestions. We have added the 95% confidence intervals to the Figure 3 and Figure 6 in the revised manuscript.

(4) It would be great to clarify what patch-level and landscape-level studies are in lines 302-306. Note that this study assesses the effects of landscape context on patch-level variables (i.e., plot-based plant richness and plot-based grassland biomass) rather than landscape-level variables (i.e., the average or total amount of biomass in a landscape).

Response: Thanks for your insightful comment. We agree with your point that our study investigated the effect of fragmented landscape context (habitat loss and fragmentation per se) on plot-based plant richness and plot-based above-ground biomass rather than landscape-level variables.

Therefore, we no longer discussed the differences between the patch-level and landscape-level studies here, instead focusing on the different ecological impacts of habitat loss and fragmentation per se in the revised manuscript.

Line 369-374:

“Although habitat loss and fragmentation per se are generally highly associated in natural landscapes, they are distinct ecological processes that determine decisions on effective conservation strategies (Fahrig, 2017; Valente et al., 2023). Our study evaluated the effects of habitat loss and fragmentation per se on grassland plant diversity and above-ground productivity in the context of fragmented landscapes in the agro-pastoral ecotone of northern China, with our results showing the effects of these two facets to not be consistent.”

(5) One possible way to avoid the confusion between "habitat fragmentation" and "fragmentation per se" could be to say "habitat loss and fragmentation per se" when you intend to express "habitat fragmentation".

Response: Thanks for your constructive suggestions. To avoid this confusion, we no longer mention habitat fragmentation in the revised manuscript but instead express it as habitat loss and fragmentation per se.

Reviewer #1 (Public Review):

This is a well-designed study that explores the BEF relationships in fragmented landscapes. Although there are massive studies on BEF relationships, most of them were conducted at local scales, few considered the impacts of landscape variables. This study used a large dataset to specifically address this question and found that habitat loss weakened the BEF relationships. Overall, this manuscript is clearly written and has important implications for BEF studies as well as for ecosystem restoration.

Response: We are grateful to you for the recognition and constructive comments. All the comments and suggestions are very constructive for improving this manuscript. We have carefully revised the manuscript following your suggestions. All changes are marked in red font in the revised manuscript.

My only concern is that the authors should clearly define habitat loss and fragmentation. Habitat loss and fragmentation are often associated, but they are different terms. The authors consider habitat loss a component of habitat fragmentation, which is not reasonable. Please see my specific comments below.

Response: We agree with your point. In the revised manuscript, we no longer consider habitat loss and fragmentation per se as two facets of habitat fragmentation. We have clearly defined habitat loss and fragmentation per se and explicitly evaluated their relative effects on plant richness, above-ground biomass, and the BEF relationship.

Reviewer #1 (Recommendations For The Authors):

Title: It is more proper to say habitat loss, rather than habitat fragmentation.

Response: Thanks for your suggestion. We have revised the title to “Habitat loss weakens the positive relationship between grassland plant richness and above-ground biomass”

Line 22, remove "Anthropogenic", this paper is not specifically discussing habitat fragmentation driven by humans.

Response: Thanks for your suggestion. We have removed the “Anthropogenic” from this sentence.

Line 26, revise to "we investigated the effects of habitat loss and fragmentation per se on plant richness... in grassland communities by using a structural equation model".

Response: Thanks for your suggestion. We have revised this sentence.

Line 25-28:

“Based on 130 landscapes identified by a stratified random sampling in the agro-pastoral ecotone of northern China, we investigated the effects of landscape context (habitat loss and fragmentation per se) on plant richness, above-ground biomass, and the relationship between them in grassland communities using a structural equation model.”

Line 58-60, habitat fragmentation generally involves habitat loss, but habitat loss is independent of habitat fragmentation, it is not a facet of habitat fragmentation.

Response: Thanks for your insightful comment. We have no longer considered habitat loss and fragmentation per se as two facets of habitat fragmentation. In the revised manuscript, we consider habitat loss and fragmentation as two different processes in fragmented landscapes.

Line 65-67, this sentence is not very relevant to this paragraph and can be deleted.

Response: Thanks for your suggestion. We have deleted this sentence from the paragraph.

Line 87-90, these references are mainly based on microorganisms, are there any references based on plants? These references are more relevant to this study. In addition, this is a key mechanism mentioned in this study, this section needs to be strengthened with more evidence and further exploration.

Response: Thanks for your comment and suggestion. Thanks for your comment and suggestion. We have added some references based on plants here to strengthen the evidence and mechanism of habitat specialisation determines the BEF relationship.

Line 89-95:

“In communities, specialists with specialised niches in resource use may contribute complementary roles to ecosystem functioning, whereas generalists with unspecialised in resource use may contribute redundant roles to ecosystem functioning due to overlapping niches (Dehling et al., 2021; Denelle et al., 2020; Gravel et al., 2011; Wilsey et al., 2023). Therefore, communities composed of specialists should have a higher niche complementarity effect in maintaining ecosystem functions and a more significant BEF relationship than communities composed of generalists.”

Denelle, P., Violle, C., DivGrass, C., Munoz, F. 2020. Generalist plants are more competitive and more functionally similar to each other than specialist plants: insights from network analyses. Journal of Biogeography 47: 1922-1933.

Dehling, D.M., Bender, I.M.A., Blendinger, P.G., Böhning-Gaese, K., Muñoz, M.C., Neuschulz, E.L., Quitián, M., Saavedra, F., Santillán, V., Schleuning, M., Stouffer, D.B. 2021. Specialists and generalists fulfil important and complementary functional roles in ecological processes. Functional Ecology 35: 1810-1821.

Wilsey, B., Martin, L., Xu, X., Isbell, F., Polley, H.W. 2023. Biodiversity: Net primary productivity relationships are eliminated by invasive species dominance. Ecology Letters.

Line 129-130, Although you can use habitat loss in the discussion or the introduction, here preferably use habitat amount or habitat area, rather than habitat loss in this case. Habitat loss represents changes in habitat area, but the remaining grasslands could be the case of natural succession or other processes, rather than loss of natural habitat.

Response: Thanks for your insightful comment. We agree with your point. In the revised manuscript, we have explicitly stated that habitat loss was represented by the loss of grassland amount in the landscape.

Since the remaining grassland fragments in this region were mainly caused by grassland loss due to human activities such as cropland expansion (Chen et al., 2019; Yang et al., 2020), we used the percentage of non-grassland cover in the landscape to represent habitat loss in our study.

Line 132-135:

“Habitat loss was represented by the loss of grassland amount in the landscape. As the remaining grassland fragments in this region were mainly caused by grassland loss due to human activities such as cropland expansion (Chen et al., 2019; Yang et al., 2020), the percentage of non-grassland cover in the landscape was used in our study to represent habitat loss.”

Lines 245-246, please also give more details of the statistical results, such as n, r value et al in the text.

Response: Thanks for your suggestion. We have added the details of the statistical results in the revised manuscript.

Line 283-290:

“Habitat loss was significantly negatively correlated with overall species richness (R = -0.21, p < 0.05, Figure 3a) and grassland specialist richness (R = -0.41, p < 0.01, Figure 3a), but positively correlated with weed richness (R = 0.31, p < 0.01, Figure 3a). Fragmentation per se was not significantly correlated with overall species richness and grassland specialist richness, but was significantly positively correlated with weed richness (R = 0.26, p < 0.01, Figure 3b). Habitat loss (R = -0.39, p < 0.01, Figure 3c) and fragmentation per se (R = -0.26, p < 0.01, Figure 3d) were both significantly negatively correlated with above-ground biomass.”

Fig. 5, is there any relationship between habitat amount and fragmentation per se in this study?

Response: Thanks for your insightful comment. We have considered a causal relationship between habitat loss and fragmentation per se in the structural equation model. We have discussed this relationship in the revised manuscript.

Line 290-293, how about the BEF relationships with different fragmentation levels? I may have missed something somewhere, but it was not shown here.

Response: Thanks for your insightful comment. We have added the BEF relationships with different fragmentation per se levels here.

Line 323-340:

“The linear regression models showed that habitat loss had a significant positive modulating effect on the positive relationship between plant richness and above-ground biomass, and fragmentation per se had no significant modulating effect (Figure 5). The positive relationship between plant richness and above-ground biomass weakened with increasing levels of habitat loss, strengthened and then weakened with increasing levels of fragmentation per se.

Author response image 1.

Relationships between grassland plant richness and above-ground biomass at different levels of habitat loss and fragmentation per se from 130 landscapes in the Tabu River Basin, a typical agro-pastoral ecotone of northern China: (a) high habitat loss and low fragmentation per se, (b) high habitat loss and moderate fragmentation per se, (c) high habitat loss and high fragmentation per se, (d) moderate habitat loss and low fragmentation per se, (e) moderate habitat loss and moderate fragmentation per se, (f) moderate habitat loss and high fragmentation per se, (g) low habitat loss and low fragmentation per se, (h) low habitat loss and moderate fragmentation per se. The R2 values in each panel are from linear regression models. The n in each panel is the number of surveying sites used in the linear regression models. The blue solid and dashed trend lines represent the significant and not significant effects, respectively. The shaded area around the trend line represents the 95% confidence interval. * represent significance at the 0.05 level. ** represent significance at the 0.01 level.”

Discussion

The Discussion (Section 4.2) needs to be revised and focused on your key findings, it is habitat loss, not fragmentation per se, that weakens the BEF relationships.

Response: Thanks for your insightful comment and suggestion. In the revised manuscript, we have rephrased the Discussion (Section 4.2) to mainly discuss the inconsistent effects of habitat loss and fragmentation per se on the BEF relationship.

Line 414-416:

“4.2 Habitat loss rather than fragmentation per se weakened the magnitude of the positive relationship between plant diversity and ecosystem function”

The R2 in the results are low (e.g., Fig. 3), please also mention other variables that might influence the observed pattern in the Discussion, such as soil and topography, though I understand it is difficult to collect such data in this study.

Response: Thanks for your insightful comment and suggestion. We agree with you and reviewer 3 that the impact of environmental factors should also be considered.

Therefore, we have considered two environmental factors related to water and temperature (soil water content and land surface temperature) in the analysis and discussed their impacts on plant diversity and above-ground biomass in the revised manuscript.

Lines 344-345, its relative importance was stronger in the intact landscape than that of the fragmented landscape?

Response: We apologize for making this confusion. We have rephrased this sentence.

Line 422-426:

“Our study found grassland plant diversity showed a stronger positive impact on above-ground productivity than landscape context and environmental factors. This result is consistent with findings by Duffy et al. (2017) in natural ecosystems, indicating grassland plant diversity has an important role in maintaining grassland ecosystem functions in the fragmented landscapes of the agro-pastoral ecotone of northern China.”

Reviewer #2 (Public Review):

Summary:

In this manuscript, Yan et al. assess the effect of two facets of habitat fragmentation (i.e., habitat loss and habitat fragmentation per se) on biodiversity, ecosystem function, and the biodiversity-ecosystem function (BEF) relationship in grasslands of an agro-pastoral ecotone landscape in northern China. The authors use stratified random sampling to select 130 study sites located within 500m-radius landscapes varying along gradients of habitat loss and habitat fragmentation per se. In these study sites, the authors measure grassland specialist and generalist plant richness via field surveys, as well as above-ground biomass by harvesting and dry-weighting the grass communities in each 3 x 1m2 plots of the 130 study sites. The authors find that habitat loss and fragmentation per se have different effects on biodiversity, ecosystem function and the BEF relationship: whereas habitat loss was associated with a decrease in plant richness, fragmentation per se was not; and whereas fragmentation per se was associated with a decrease in above-ground biomass, habitat loss was not. Finally, habitat loss, but not fragmentation per se was linked to a decrease in the magnitude of the positive biodiversity-ecosystem functioning relationship, by reducing the percentage of grassland specialists in the community.

Strengths:

This study by Yan et al. is an exceptionally well-designed, well-written, clear and concise study shedding light on a longstanding, important question in landscape ecology and biodiversity-ecosystem functioning research. Via a stratified random sampling approach (cf. also "quasi-experimental design" Butsic et al. 2017), Yan et al. create an ideal set of study sites, where habitat loss and habitat fragmentation per se (usually highly correlated) are decorrelated and hence, separate effects of each of these facets on biodiversity and ecosystem function can be assessed statistically in "real-world" (and not experimental, cf. Duffy et al. 2017) communities. The authors use adequate and well-described methods to investigate their questions. The findings of this study add important empirical evidence from real-world grassland ecosystems that help to advance our theoretical understanding of landscape-moderation of biodiversity effects and provide important guidelines for conservation management.

Weaknesses:

I found only a few minor issues, mostly unclear descriptions in the study that could be revised for more clarity.

Response: Thanks very much for your review of the manuscript. We are grateful to you for the recognition. All the comments and suggestions are very insightful and constructive for improving this manuscript. We have carefully studied the literature you recommend and revised the manuscript carefully following your suggestions. All changes are marked in red font in the revised manuscript.

Reviewer #2 (Recommendations For The Authors):

Specific comments

(1) Some aspects of the Methods section were not entirely clear to me, could you revise them for more clarity?

(a) Whereas you describe 4 main facets of fragmentation per se that are used to create the PC1 as a measure of overall fragmentation per se, it looks as if this PC1 is mainly driven by 3 facets only (ED, PD and AREA_MN), and patch isolation (nearest neighbour distance, ENN) having a relatively low loading on PC1 (Figure A1). I think it would be good to discuss this fact and the consequences of it, that your definition of fragmentation is focused more on edge density, patch density and mean patch area, and less on patch isolation in your Discussion section?

Response: Thanks for your insightful comment and suggestion. We agree with your point. We have discussed this fact and its implications for understanding the effects of fragmentation per se in our study.

Line 384-389:

“However, it is important to stress that the observed positive effect of fragmentation per se does not imply that increasing the isolation of grassland patches would promote biodiversity, as the metric of fragmentation per se used in our study was more related to patch density, edge density and mean patch area while relatively less related to patch isolation (Appendix Table A1). The potential threats from isolation still need to be carefully considered in the conservation of biodiversity in fragmented landscapes (Haddad et al., 2015).”

(b) Also, from your PCA in Figure A1, it seems that positive values of PC1 mean "low fragmentation", whereas high values of PC1 mean "high fragmentation", however, in Figure A2, the inverse is shown (low values of PC1 = low fragmentation, high values of PC1 = high fragmentation). Could you clarify in the Methods section, if you scaled or normalized the PC1 to match this directionality?

Response: We apologize for making this confusion. In order to be consistent with the direction of change in fragmentation per se, we took the inverse of the PC1 as a single fragmentation per se index, which was positively correlated with patch density, edge density, mean nearest-neighbor distance metric, and negatively with mean patch area (Appendix Figure A1 and Table A1). We have clarified this point in the Method section.

Line 160-163:

“We took the inverse of the PC1 as a single fragmentation per se index, which was positively correlated with patch density, edge density, mean nearest-neighbor distance metric, and negatively with mean patch area (Appendix Figure A1 and Table A1).”

(c) On line 155 you describe that you selected at least 20 landscapes using stratified sampling from each of the eight groups of habitat amount and fragmentation combination. Could you clarify: 1) did you randomly sample within these groups with a minimum distance condition or was it a non-random selection according to other criteria? (I think you could move the "To prevent overlapping landscapes..." sentence up here to the description of the landscape selection process) 2) Why did you write "at least 20 landscapes" - were there in some cases more or less landscapes selected? 130 study landscapes divided by 8 groups only gives you 16.25, hence, at least for some groups there were less than 20 landscapes? Could you describe your final dataset in more detail, i.e. the number of landscapes per group and potential repercussions for your analysis?

Response: Thanks for your insightful comments. In the revised manuscript, we have rephrased the method to provide more detail for the sampling landscape selection.

(1) Line 169-172

We randomly selected at least 20 grassland landscapes with a minimum distance condition using stratified sampling from each of the remaining eight grassland types as alternative sites for field surveys. The minimum distance between each landscape was at least 1000 m to prevent overlapping landscapes and potential spatial autocorrelation.

(2) Line 184-191

The reason for selecting at least 20 grassland landscapes of each type in this study was to ensure enough alternative sites for the field survey. This is because the habitat type of some selected sites was not the natural grasslands, such as abandoned agricultural land. Some of the selected sites may not be permitted for field surveys.

Thus, we finally established 130 sites in the field survey. The types of the 130 sites were: 19 high-moderate, 14 high-low, 19 moderate-high, 16 moderate-moderate, 18 moderate-low, 16 low-high, 17 low-moderate, 11 low-low habitat amount and fragmentation per se.

(d) On line 166, you describe that you established 130 sites of 30 m by 30 m - I assume they were located (more or less) exactly in the centre of the selected 500 m - radius landscapes? Were they established so that they were fully covered with grassland? And more importantly, how did you establish the 10 m by 10 m areas and the 1 m2 plots within the 30 m by 30 m sites? Did you divide the 30 m by 30 m areas into three rectangles of 10 m by 10 m and then randomly established 1 m2 plots? Were the 1 m2 plots always fully covered with grassland/was there a minimum distance to edge criterion? Please describe with more detail how you established the 1 m2 study sites, and how many there were per landscape.

Response: Thanks for your insightful comments. In the revised manuscript, we have provided more detailed information on how to set up 130 sites of 30 m by 30 m and three plots of 1 m by 1 m.

(1) As these 130 sites were selected based on the calculation of the moving window, they were located (more or less) exactly in the centre of the 500-m radius buffer.

(2) These sites were fully covered with grassland because their size (30 m by 30 m) was the same as the size of the grassland cell (30 m by 30 m) used in the calculation of the moving window.

(3) We randomly set up three 1 m * 1 m plots in a flat topographic area at the 10 m * 10 m centre of each site. Thus, there was a minimum distance of 10 m to the edge for each 1 m * 1 m plot.

(4) There are three 1 m * 1 m plots per landscape.

Line 182-191:

“Based on the alternative sites selected above, we established 130 sites (30 m * 30 m) between late July to mid-August 2020 in the Tabu River Basin in Siziwang Banner, Inner Mongolia Autonomous Region (Figure 1). The types of the 130 sites were: 19 high-moderate, 14 high-low, 19 moderate-high, 16 moderate-moderate, 18 moderate-low, 16 low-high, 17 low-moderate, 11 low-low habitat amount and fragmentation per se. In order to exclude the impact of historical agricultural activities, the habitat type of the established sites was natural grasslands with regional vegetation characteristics. Each site was not abandoned agricultural land, and there was no sign of agricultural reclamation.

At the 10 m * 10 m center of each site, we randomly set up three 1 m * 1 m plots in a flat topographic area to investigate grassland vascular plant diversity and above-ground productivity.”

(e) Line 171: could you explain what you mean by reclaimed?

Response: Thanks for your comment. The “reclaimed” means that historical agricultural activities. We have rephrased this sentence to make it more explicit.

Line 186-189:

“In order to exclude the impact of historical agricultural activities, the habitat type of the established sites was natural grasslands with regional vegetation characteristics. Each site was not abandoned agricultural land, and there was no sign of agricultural reclamation.”

(f) Line 188 ff.: Hence your measure of productivity is average-above ground biomass per 1 m2. I think it would add clarity if you highlighted this more explicitly.

Response: Thanks for your suggestion. We have highlighted that the productivity in our study was the average above-ground biomass per 1 m * 1 m plots in each site.

Line 215-217:

“For each site, we calculated the mean vascular plant richness of the three 1 m * 1 m plots, representing the vascular plant diversity, and mean above-ground biomass of the three 1 m * 1 m plots, representing the above-ground productivity.”

(2) All figures are clear and well-designed!

(a) Just as a suggestion: in Figures 3 and 6, you could maybe add the standard errors of the mean as well?

Response: Thanks for your suggestion. In the revised manuscript, we have added the standard errors of the mean in Figures 3 and 6.

(b) Figure 4: Could you please clarify: Which models were the optimal models on which these model-averaged standardized parameter estimates were based on? And hence, the optimal models contained all 4 predictors (otherwise, no standardized parameter estimate could be calculated)? Or do these model-averaged parameters take into account all possible models (and not only the optimal ones)?

Response: Thanks for your suggestion. We selected the four optimal models based on the AICc value to calculate the model-averaged standardized parameter estimates. The four optimal models contained all predictors in Figure 4. We have added the four optimal models in Appendix Table A3.

Appendix:

Author response table 1.

Four optimal models of landscape context, environment factors, and plant diversity affecting above-ground biomass.

Note: AGB: above-ground biomass; HL: habitat loss; FPS: fragmentation per se; SWT: soil water content; LST: land surface temperature; GSR: grassland specialist richness; WR: weed richness; **: significance at the 0.01 level.”

(c) Please add in all Figures (i.e., Figures 4, 5 and 6, Figure 6 per "high, moderate and low-class") the number of study units the analyses were based on.

Response: Thanks for your suggestion. In the revised manuscript, we have added the number of study units the analyses were based on in all Figures.

(d) Figure 6: I think it would be more consistent to add a second plot where the BEF-relationship is shown for low, moderate and high levels of habitat fragmentation per se. Could you also add a clearer description in the Methods and/or Results section of how you assessed if habitat amount or fragmentation per se affected the BEF-relationship? I.e. based on the significance of the interaction term (habitat amount x species richness) in a linear model?

Response: Thanks for your insightful comment and suggestion. We have added a second plot in Figure 5 to show the BEF relationship at low, moderate and high levels of fragmentation per se.

Line 328-340:

Author response image 2.

Relationships between grassland plant richness and above-ground biomass at different levels of habitat loss and fragmentation per se from 130 landscapes in the Tabu River Basin, a typical agro-pastoral ecotone of northern China: (a) high habitat loss and low fragmentation per se, (b) high habitat loss and moderate fragmentation per se, (c) high habitat loss and high fragmentation per se, (d) moderate habitat loss and low fragmentation per se, (e) moderate habitat loss and moderate fragmentation per se, (f) moderate habitat loss and high fragmentation per se, (g) low habitat loss and low fragmentation per se, (h) low habitat loss and moderate fragmentation per se. The R2 values in each panel are from linear regression models. The n in each panel is the number of surveying sites used in the linear regression models. The blue solid and dashed trend lines represent the significant and not significant effects, respectively. The shaded area around the trend line represents the 95% confidence interval. * represent significance at the 0.05 level. ** represent significance at the 0.01 level.”

We determined whether habitat loss and fragmentation per se moderated the BEF relationship by testing the significance of their interaction term with plant richness. We have added a clearer description in the Methods section of the revised manuscript.

Line 245-250:

“We then assessed the significance of interaction terms between habitat loss and fragmentation per se and plant richness in the linear regression models to evaluate whether they modulate the relationship between plant richness and above-ground biomass. Further, we used a piecewise structural equation model to investigate the specific pathways in which habitat loss and fragmentation per se modulate the relationship between plant richness and above-ground biomass.”

(3) While reading your manuscript, I missed a discussion on the potential non-linear effects of habitat amount and fragmentation per se. In your study, it seems that the effects of habitat amount and fragmentation per se on biodiversity and ecosystem function are quite linear, which contrasts previous research highlighting that intermediate levels of fragmentation/heterogeneity could maximise spatial asynchrony, biodiversity and ecosystem function (e.g. Redon et al. 2014, Thompson & Gonzalez 2016, Tscharntke et al. 2012, Wilcox et al. 2017). I think it would add depth to your study if you discussed your finding of linear effects of habitat amount and fragmentation on biodiversity, ecosystem functioning and BEF. For example:

Response: Thanks for your constructive suggestions. We have carefully studied the literature (e.g. Redon et al. 2014, Thompson & Gonzalez 2016, Tscharntke et al. 2012, Wilcox et al. 2017), which highlights that intermediate levels of fragmentation/heterogeneity could maximise spatial asynchrony, biodiversity and ecosystem function.

In the revised manuscript, we have added the discussion about the linear positive effects of fragmentation on plant diversity and above-ground productivity and discussed possible reasons for this linear effect.

Line 402-413:

“In our study, a possible mechanism for the positive impacts of fragmentation per se on plant diversity and above-ground productivity (indirect positive impact via plant diversity) is that fragmentation per se increases the habitat heterogeneity in the landscape, which can promote biodiversity through spatial asynchrony and spatial insurance effects (Tscharntke et al., 2012). Previous studies indicated that heterogeneity typically has nonlinear effects on biodiversity and ecosystem function, as moderate heterogeneity can maximise spatial asynchrony (Redon et al., 2014; Wilcox et al., 2017). However, our study did not observe nonlinear patterns between fragmentation per se and plant diversity and above-ground productivity. This may be due to the low spatial heterogeneity of this area as a result of agricultural intensification (Benton et al., 2003; Chen et al., 2019). The gradient of fragmentation per se in our study may not cover the optimal heterogeneity levels for maximising plant diversity and above-ground productivity (Thompson and Gonzalez, 2016).”

Meanwhile, we also discussed the nonlinear pattern of the BEF relationship with increasing levels of fragmentation per se to add depth to the discussion.

Line 442-451:

“In addition, our study found that the BEF relationship showed a nonlinear pattern with increasing levels of fragmentation per se. For a given level of habitat loss, the positive BEF relationship was strongest at moderate fragmentation per se level and became neutral at high fragmentation per se level. This can be explained by the increased spatial asynchrony at moderate fragmentation per se level, which can promote niche complementary among species in the community and thus strengthen the BEF relationship (Gonzalez et al., 2020; Thompson and Gonzalez, 2016; Tscharntke et al., 2012). The neutral BEF relationship at high fragmentation per se level may be due to edge effects enhancing environmental filtering, thereby leading to functional redundancy among species and decoupling the BEF relationship (Fetzer et al., 2015; Hu et al., 2016; Zambrano et al., 2019).”

(a) Line 74-75: I was wondering if you also thought of spatial insurance effects or spatial asynchrony effects that can emerge with habitat fragmentation, which could lead to increased ecosystem functioning as well? (refs. above).

Response: Thanks for your constructive suggestions. In the revised manuscript, we have explicitly considered the spatial insurance effect or spatial asynchrony as the important mechanism for fragmentation per se to increase plant diversity, ecosystem function, and the BEF relationship.

Line 74-77:

“In theory, habitat loss and fragmentation per se can regulate ecosystem function and the BEF relationship by altering species composition, interactions, and spatial asynchrony regardless of changes in species richness (Liu et al., 2018; Thompson and Gonzalez, 2016; Tscharntke et al., 2012).”

Line 402-408:

“In our study, a possible mechanism for the positive impacts of fragmentation per se on plant diversity and above-ground productivity (indirect positive impact via plant diversity) is that fragmentation per se increases the habitat heterogeneity in the landscape, which can promote biodiversity through spatial asynchrony and spatial insurance effects (Tscharntke et al., 2012). Previous studies indicated that heterogeneity typically has nonlinear effects on biodiversity and ecosystem function, as moderate heterogeneity can maximise spatial asynchrony (Redon et al., 2014; Wilcox et al., 2017).”

Line 442-451:

“In addition, our study found that the BEF relationship showed a nonlinear pattern with increasing levels of fragmentation per se. For a given level of habitat loss, the positive BEF relationship was strongest at moderate fragmentation per se level and became neutral at high fragmentation per se level. This can be explained by the increased spatial asynchrony at moderate fragmentation per se level, which can promote niche complementary among species in the community and thus strengthen the BEF relationship (Gonzalez et al., 2020; Thompson and Gonzalez, 2016; Tscharntke et al., 2012). The neutral BEF relationship at high fragmentation per se level may be due to edge effects enhancing environmental filtering, thereby leading to functional redundancy among species and decoupling the BEF relationship (Fetzer et al., 2015; Hu et al., 2016; Zambrano et al., 2019).”

(b) I was wondering, if this result of linear effects could also be the result of a fragmentation gradient that does not cover the whole range of potential values? Maybe it would be good to compare the gradient in habitat fragmentation in your study with a theoretical minimum maximum/considering that there might be an optimal medium degree of fragmentation.

Response: Thanks for your insightful comment. We agree with your point that the linear effect of fragmentation per se in our study may be due to the fact that the gradient of fragmentation per se in this region may not cover the optimal heterogeneity levels for maximising spatial asynchrony. This is mainly because the agricultural intensification in the agro-pastoral ecotone of northern China could lead to lower spatial heterogeneity in this region. We have explicitly discussed this point in the revised manuscript.

Line 406-413:

“Previous studies indicated that heterogeneity typically has nonlinear effects on biodiversity and ecosystem function, as moderate heterogeneity can maximise spatial asynchrony (Redon et al., 2014; Wilcox et al., 2017). However, our study did not observe nonlinear patterns between fragmentation per se and plant diversity and above-ground productivity. This may be due to the low spatial heterogeneity of this area as a result of agricultural intensification (Benton et al., 2003; Chen et al., 2019). The gradient of fragmentation per se in our study may not cover the optimal heterogeneity levels for maximising plant diversity and above-ground productivity (Thompson and Gonzalez, 2016).”

(4) Some additional suggestions:

(a) Line 3: Maybe add "via reducing the percentage of grassland specialists in the community"?

Response: Thanks for your suggestion. We have revised this sentence.

Line 19:

“Habitat loss can weaken the positive BEF relationship via reducing the percentage of grassland specialists in the community”

(b) Lines 46-48: Maybe add "but see: Duffy, J.E., Godwin, C.M. & Cardinale, B.J. (2017). Biodiversity effects in the wild are common and as strong as key drivers of productivity. Nature."

Response: Thanks for your suggestion. We have added this reference here.

Line 47-49:

“When research expands from experiments to natural systems, however, BEF relationships remain unclear in the natural assembled communities, with significant context dependency (Hagan et al., 2021; van der Plas, 2019; but see Duffy et al., 2017).”

(c) Lines 82-87 and lines 90-93: Hence, your study actually is in contrast to these findings, i.e., fragmented landscapes do not necessarily have a lower fraction of grassland specialists? If yes, could you highlight this more explicitly?

Response: Thanks for your insightful comment. We have explicitly highlighted this point in the revised manuscript.

Line 434-439:

“Meanwhile, our study demonstrates that habitat loss, rather than fragmentation per se, can decrease the degree of habitat specialisation by leading to the replacement of specialists by generalists in the community, thus weakening the BEF relationship. This is mainly because fragmentation per se did not decrease the grassland specialist richness in this region, whereas habitat loss decreased the grassland specialist richness and led to the invasion of more weeds from the surrounding farmland into the grassland community (Yan et al., 2022; Yan et al., 2023).”

(d) Line 360: Could you add some examples of these multiple ecosystem functions you refer to?

Response: Thanks for your suggestion. We have added some examples of these multiple ecosystem functions here.

Line 456-457:

“Therefore, future studies are needed to focus on multiple ecosystem functions, such as below-ground productivity, litter decomposition, soil carbon stocks, etc.”

Reviewer #3 (Public Review):

Summary:

The authors aim to solve how landscape context impacts the community BEF relationship. They found habitat loss and fragmentation per se have inconsistent effects on biodiversity and ecosystem function. Habitat loss rather than fragmentation per se can weaken the positive BEF relationship by decreasing the degree of habitat specialization of the community.

Strengths:

The authors provide a good background, and they have a good grasp of habitat fragmentation and BEF literature. A major strength of this study is separating the impacts of habitat loss and fragmentation per se using the convincing design selection of landscapes with different combinations of habitat amount and fragmentation per se. Another strength is considering the role of specialists and generalists in shaping the BEF relationship.

Response: We are grateful to you for the recognition and constructive comments. All the comments and suggestions are very constructive for improving this manuscript. We have carefully revised the manuscript following your suggestions. All changes are marked in red font in the revised manuscript.

Weaknesses:

(1) The authors used five fragmentation metrics in their study. However, the choice of these fragmentation metrics was not well justified. The ecological significance of each fragmentation metric needs to be differentiated clearly. Also, these fragmentation metrics may be highly correlated with each other and redundant. I suggest author test the collinearity of these fragmentation metrics for influencing biodiversity and ecosystem function.

Response: Thanks for your constructive suggestion. The fragmentation metrics used in our study represent the different processes of breaking apart of habitat in the landscape, which are widely used by previous studies (Fahrig, 2003; Fahrig, 2017). In the revised manuscript, we have provided more detailed information about the ecological significance of these fragmentation indices.

Line 142-148:

“The patch density metric reflects the breaking apart of habitat in the landscape, which is a direct reflection of the definition of fragmentation per se (Fahrig et al., 2019). The edge density metric reflects the magnitude of the edge effect caused by fragmentation (Fahrig, 2017). The mean patch area metric and the mean nearest-neighbor distance metric are associated with the area and distance effects of island biogeography, respectively, reflecting the processes of local extinction and dispersal of species in the landscape (Fletcher et al., 2018).”

Meanwhile, we have calculated the variance inflation factors (VIF) for each fragmentation metric to assess their collinearity. The VIF of these fragmentation metrics were all less than four, suggesting no significant multicollinearity for influencing biodiversity and ecosystem function.

Author response table 2.

Variance inflation factors of habitat loss and fragmentation per se indices for influencing plant richness and above-ground biomass.

(2) I found the local environmental factors were not considered in the study. As the author mentioned in the manuscript, temperature and water also have important impacts on biodiversity and ecosystem function in the natural ecosystem. I suggest authors include the environmental factors in the data analysis to control their potential impact, especially the structural equation model.

Response: Thanks for your constructive suggestion. We agree with you that environmental factors should be considered in our study. In the revised manuscript, we have integrated two environmental factors related to water and temperature (soil water content and land surface temperature) into the data analysis to control their potential impact. The main results and conclusions of the revised manuscript are consistent with those of the previous manuscript.

Reviewer #3 (Recommendations For The Authors):

(1) L60-63. The necessity to distinguish between habitat loss and fragmentation per se is not clearly stated. More information about biodiversity conservation strategies can be given here.

Response: Thanks for your suggestion. In the revised manuscript, we have provided more evidence about the importance of distinguishing between habitat loss and fragmentation per se for biodiversity conservation.

Line 62-67:

“Habitat loss is often considered the major near-term threat to the biodiversity of terrestrial ecosystems (Chase et al., 2020; Haddad et al., 2015), while the impact of fragmentation per se remains debated (Fletcher Jr et al., 2023; Miller-Rushing et al., 2019). Thus, habitat loss and fragmentation per se may have inconsistent ecological consequences and should be considered simultaneously to establish effective conservation strategies in fragmented landscapes (Fahrig et al., 2019; Fletcher et al., 2018; Miller-Rushing et al., 2019).”

(2) L73-77. The two sentences are hard to follow. Please rephrase to improve the logic. And I don't understand the "however" here. There is no twist.

Response: Thanks for your suggestion. We have rephrased the two sentences to improve their logic.

Line 74-79:

“In theory, habitat loss and fragmentation per se can regulate ecosystem function and the BEF relationship by altering species composition, interactions, and spatial asynchrony regardless of changes in species richness (Liu et al., 2018; Thompson and Gonzalez, 2016; Tscharntke et al., 2012). This is because species in communities are not ecologically equivalent and may respond differently to habitat loss and fragmentation per se, and contribute unequally to ecosystem function (Devictor et al., 2008; Wardle and Zackrisson, 2005).”

(3) L97. Are grasslands really the largest terrestrial ecosystem? Isn't it the forest?

Response: We apologize for making this confusion. We have rephrased this sentence here.

Line 101-104:

“Grasslands have received considerably less attention, despite being one of the largest terrestrial ecosystems, and suffering severe fragmentation due to human activities, such as agricultural reclamation and urbanisation (Fardila et al., 2017).”

(4) Fig.1, whether the four sample plots presented in panel b are from panel a. Please add the scale bar in panel b.

Response: Thanks for your comment. The four sample plots presented in panel b are from panel a in Figure 1. We have also added the scale bar in panel b.

(5) L105. This statement is too specific. Please remove and consider merging this paragraph with the next.

Response: Thanks for your suggestion. We have removed this sentence and merged this paragraph with the next.

(6) L157. The accuracy and kappa value of the supervised classification should be given.

Response: Thanks for your suggestion. We have added the accuracy and kappa value of the supervised classification in the revised manuscript.

Line 176-177:

“The overall classification accuracy was 84.3 %, and the kappa coefficient was 0.81.”

(7) I would recommend the authors provide the list of generalists and specialists surveyed in the supplementary. Readers may not be familiar with the plant species composition in this area.

Response: Thanks for your suggestion. We agree with your point. We have provided the list of generalists and specialists surveyed in the Appendix Table A4.

Line 282-283:

“A total of 130 vascular plant species were identified in our study sites, including 91 grassland specialists and 39 weeds (Appendix Table A4).”

(8) Fig.4, it is better to add the results of variation partition to present the relative contribution of habitat fragmentation, environmental factors, and plant diversity.

Response: Thanks for your suggestion. We have integrated the landscape context, environmental factors, and plant diversity into the multi-model averaging analysis and redraw Figure 4 to present their relative importance for above-ground biomass.

Line 313-319:

Author response image 3.

Standardised parameter estimates and 95% confidence intervals for landscape context, plant diversity, and environmental factors affecting above-ground biomass from 130 landscapes in the Tabu River Basin, a typical agro-pastoral ecotone of northern China. Standardised estimates and 95% confidence intervals are calculated by the multi-model averaging method based on the four optimal models affecting above-ground biomass (Appendix Table A3). ** represent significance at the 0.01 level.

(9) Please redraw Fig.2 and Fig.5 to integrate the environmental factors. Add the R-square to Fig 5.

Response: Thanks for your suggestion. We have integrated two environmental factors into the structural equation model and redraw Figure 2 and Figure 5 in the revised manuscript. And we have added the R-square to the Figure 5.

(10) L354. The authors should be careful to claim that habitat loss could reduce the importance of plant diversity to ecosystem function. This pattern observed may depend on the type of ecosystem function studied.

Response: Thanks for your suggestion. We have avoided this claim in the revised manuscript and explicitly discussed the importance of simultaneously focusing on multiple ecosystem functions, such as below-ground productivity, litter decomposition, soil carbon stocks, etc.

Line 454-457:

“This inconsistency can be explained by trade-offs between different ecosystem functions that may differ in their response to fragmentation per se (Banks-Leite et al., 2020). Therefore, future studies are needed to focus on multiple ecosystem functions, such as below-ground productivity, litter decomposition, soil carbon stocks, etc.”

Author response:

The following is the authors’ response to the original reviews.

eLife assessment This valuable paper reports a theoretical framework and methodology for identifying Cancer Driving Nucleotides (CDNs), primarily based on single nucleotide variant (SNV) frequencies. A variety of solid approaches indicate that a mutation recurring three or more times is more likely to reflect selection rather than being the consequence of a mutation hotspot. The method is rigorously quantitative, though the requirement for larger datasets to fully identify all CDNs remains a noted limitation. The work will be of broad interest to cancer geneticists and evolutionary biologists. 

The key criticism “the requirement for larger datasets to fully identify all CDNs remains a noted limitation” that is also found in both reviews. We have clarified the issue in the main text, the relevant parts, from which are copied below. The response below also addresses many comments in the reviews. In addition, Discussion of eLife-RP-RA-2024-99341 has been substantially expanded to answer the questions of Reviewer 2.

We shall answer the boldface comment in three ways. First, it can be answered using GENIE data. Fig. 7 of the main text (eLife-RP-RA-2024-99340) shows that, when n increases from ~ 1000 to ~ 9,000, the numbers of discovered CDNs increase by 3 – 5 fold, most of which come from the two-hit class. Hence, the power of discovering more CDNs with larger datasets is evident. By extrapolation, a sample size of 100,000 should be able to yield 90% of all CDNs, as calculated here. (Fig. 7 also addresses the queries of whether we have used datasets other than TCGA. We indeed have used all public data, including GENIE and COSMIC.) 

Second, the power of discovering more cancer driver genes by our theory is evident even without using larger datasets. Table 3 of the companion study (eLife-RP-RA-2024-99341) shows that, averaged across cancer types, the conventional method would identify 45 CDGs while the CDN method tallies 258 CDGs. The power of the CDN method is demonstrated. This is because the conventional approach has to identify CDGs (cancer driver genes) in order to identify the CDNs they carry. However, many CDNs occur in non-CDGs and are thus missed by the conventional approach. In Supplementary File S2, we have included a full list of CDNs discovered in our study, along with population allele frequency annotations from gnomAD. The distribution patterns of these CDNs across different cancer types show their pan-cancer properties as further explored in the companion paper.

Third, while many, or even most CDNs occur in non-CDGs and are thus missed, the conventional approach also includes non-CDN mutations in CDGs. This is illustrated in Fig. 5 of the companion study (eLife-RP-RA-2024-99341) that shows the adverse effect of misidentifications of CDNs by the conventional approach. In that analysis, the gene-targeting therapy is effective if the patient has the CDN mutations on EGFR, but the effect is reversed if the EGFR mutations are non-CDN mutations.

Reviewer #1 (Public Review):

The authors developed a rigorous methodology for identifying all Cancer Driving Nucleotides (CDNs) by leveraging the concept of massively repeated evolution in cancer. By focusing on mutations that recur frequently in pan-cancer, they aimed to differentiate between true driver mutations and neutral mutations, ultimately enhancing the understanding of the mutational landscape that drives tumorigenesis. Their goal was to call a comprehensive catalogue of CDNs to inform more effective targeted therapies and address issues such as drug resistance.

Strengths

(1) The authors introduced a concept of using massively repeated evolution to identify CDNs. This approach recognizes that advantageous mutations recur frequently (at least 3 times) across cancer patients, providing a lens to identify true cancer drivers.

(2) The theory showed the feasibility of identifying almost all CDNs if the number of sequenced patients increases to 100,000 for each cancer type.

Weaknesses

(1) The methodology remains theoretical and no novel true driver mutations were identified in this study.

We now address the weakness criticism, which is gratefully received.

The second part of the criticism (no novel true driver mutations were identified in this study) has been answered in the long responses to eLife assessment above. The first part “The methodology remains theoretical” is somewhat unclear. It might be the lead to the second part. However, just in case, we interpret the word “theoretical” to mean “the lack of experimental proof” and answer below.

As Reviewer #1 noted, a common limitation of theoretical and statistical analyses of cancer drivers is the need to validate their selective advantage through in vitro or in vivo functional testing. This concern is echoed by both reviewers in the companion paper (eLife-RP-RA-2024-99341), prompting us to consider the methodology for functional testing of potential cancer drivers. An intuitive approach would involve introducing putative driver mutations into normal cells and observing phenotypic transformation in vitro and in vivo. In a recent stepwise-edited human melanoma model, Hodis et al. demonstrated that disease-relevant phenotypes depend on the “correct” combinations of multiple driver mutations (Hodis et al. 2022). Other high-throughput strategies can be broadly categorized into two approaches: (1) introducing candidate driver mutations into pre-malignant model systems that already harbor a canonical mutant driver (Drost and Clevers 2018; Grzeskowiak et al. 2018; Michels et al. 2020) and (2) introducing candidate driver mutations into growth factor-dependent cell models and assessing their impact on resulting fitness (Bailey et al. 2018; Ng et al. 2018). The underlying assumption of these strategies is that the fitness outcomes of candidate driver mutations are influenced by pre-existing driver mutations and the specific pathways or cancer hallmarks being investigated. This confines the functional test of potential cancer driver mutations to conventional cancer pathways. A comprehensive identification of CDNs is therefore crucial to overcome these limitations. In conjunction with other driver signal detection methods, our study aims to provide a more comprehensive profile of driver mutations, thereby enabling the functional testing of drivers involved in non-conventional cancer evolution pathways.

(2) Different cancer types have unique mutational landscapes. The methodology, while robust, might face challenges in uniformly identifying CDNs across various cancers with distinct genetic and epigenetic contexts.

We appreciate the comment. Indeed, different cancer types should have different genetic and epigenetic landscapes. In that case, one may have expected CDNs to be poorly shared among cancer types. However, as reported in Fig. 4 of the companion study, the sharing of CDNs across cancer types is far more common than the sharing of CDGs (Cancer Driving Genes). We suggest that CDNs have a much higher resolution than CDGs, whereby the signals are diluted by non-driver mutations. In other words, despite that the mutational landscape may be cancer-type specific, the pan-cancer selective pressure may be sufficiently high to permit the detection of CDN sharing among cancer types.

Below, we shall respond in greater details. Epigenetic factors, such as chromatin states, methylation/acetylation levels, and replication timing, can provide valuable insights when analyzing mutational landscapes at a regional scale (Stamatoyannopoulos et al. 2009; Lawrence et al. 2013; Makova and Hardison 2015; Baylin and Jones 2016; Alexandrov et al. 2020; Abascal et al. 2021; Sherman et al. 2022). However, at the site-specific level, the effectiveness of these covariates in predicting mutational landscapes depends on their integration into a detailed model. Overemphasizing these covariates could lead to false negatives for known driver mutations (Hess et al. 2019; Elliott and Larsson 2021). In figure 3B of the main text, we illustrate the discrepancy between the mutation rate predictions from Dig and empirical observation. Ideally, no covariates would be needed under extensive sample sizes, where each mutable genomic sites would have sufficient mutations to yield a statistic significance and consequently, synonymous mutations would be sufficient for the characterization of mutational landscape. In this sense, the integration of mutational covariates represents a compromise under current sample size. In our study, the effect of unique mutational landscapes is captured by E(u), the mean mutation rate for each cancer type. We further accounted for the variability of site-level mutability using a gamma distribution. The primary goal of our study is to determine the upper limit of mutation recurrences under mutational mechanisms only. While selection force acts blindly to genomic features, mutational hotspots should exhibit common characteristics determined by their underlying mechanisms. In the main text, we attempted to identify such shared features among CDNs. Until these mutational mechanisms are fully understood, CDNs should be considered as potential driver mutations.

(3) L223, the statement "In other words, the sequences surrounding the high-recurrence sites appear rather random.". Since it was a pan-cancer analysis, the unique patterns of each cancer type could be strongly diluted in the pan-cancer data.

We now state that the analyses of mutation characteristic have been applied to the individual cancer types and did not find any pattern that deviates from randomness. Nevertheless, it may be argued that, with the exception of those with sufficiently large sample sizes such as lung and breast cancers, most datasets do not have the power to reject the null hypothesis. To alleviate this concern, we applied the ResNet and LSTM/GRU methods for the discovery of potential mutation motifs within each cancer type. All methods are more powerful than the one used but the results are the same – no cancer type yields a mutation pattern that can reject the null hypothesis of randomness (see below).

As a positive control, we used these methods for the discovery of splicing sites of human exons. When aligned up with splicing site situated in the center (position 51 in the following plot), the sequence motif would look like:

Author response image 1.

5-prime

Author response image 2.

3-prime

However, To account for the potential influence of distance from the mutant site in motif analysis, we randomly shuffled the splicing sites within a specified window around the alignment center, and their sequence logo now looks like:

Author response image 3.

5-prime shuffled

Author response image 4.

3-prime shuffled

Author response image 5.

random sequences from coding regions

The classification results of the shuffled 5-prime (donner), 3-prime (acceptor) and random sequences from coding regions (Random CDS) are presented in the Author response table 1 (The accuracy for the aligned results, which is approximately 99%, is not shown here).

Author response table 1.

With the positive results from these positive controls (splicing site motifs) validating our methodology, we applied the same model structure to the train and test of potential mutational motifs of CDN sites. All models achieved approximately 50% accuracy in CDN motif analysis, suggesting that the sequence contexts surrounding CDN sites are not significantly different from other coding regions of the genome. This further implies that the recurrence of mutations at CDN sites is more likely driven by selection rather than mutational mechanisms.

Note that this preliminary analysis may be limited by insufficient training data for CDN sites. Future studies will require larger sample sizes and more sophisticated models to address these limitations.

(4) To solidify the findings, the results need to be replicated in an independent dataset.

Figure 7 validates our CDN findings using the GENIE dataset, which primarily consists of targeted sequencing data from various panels. By focusing on the same genomic regions sequenced by GENIE, we observed a 3-5 fold increase in the number of discovered CDNs as sample size increased from approximately 1000 to 9000. Moreover, the majority of CDNs identified in TCGA were confirmed as CDNs in GENIE.

(5) The key scripts and the list of key results (i.e., CDN sites with i{greater than or equal to}3) need to be shared to enable replication, validation, and further research. So far, only CDN sites with i{greater than or equal to}20 have been shared.

We have now updated the “Data Availability” section in the main text, the corresponding scripts for key results are available on Gitlab at: https://gitlab.com/ultramicroevo/cdn_v1.

(6) The versions of data used in this study are not clearly detailed, such as the specific version of gnomAD and the version and date of TCGA data downloaded from the GDC Data Portal.

The versions of data sources have now been updated in the revised manuscript.

Recommendations For The Authors:

(1) L119, states "22.7 million nonsynonymous sites," but Table 1 lists the number as 22,540,623 (22.5 million). This discrepancy needs to be addressed for consistency.<br /> (2) Figure 2B, there is an unexplained drop in the line at i = 6 and 7 (from 83 to 45). Clarification is needed on why this drop occurs.<br /> (3) Figure 3A, for the CNS type, data for recurrence at 8 and 9 are missing. An explanation should be provided for this absence.<br /> (4) L201, the title refers to "100-mers," but L218 mentions "101-mers." This inconsistency needs to be corrected to ensure clarity and accuracy.<br /> (5) Figures 6 and 7 currently lack titles. Titles should be added to these figures to improve readability.

Thanks. All corrections have been incorporated into the revised manuscript.

Reviewer #2 (Public Review):<br /> Summary:<br /> The authors propose that cancer-driver mutations can be identified by Cancer Driving Nucleotides (CDNs). CDNs are defined as SNVs that occur frequently in genes. There are many ways to define cancer driver mutations, and the strengths and weaknesses are the reliance on statistics to define them.<br /> Strengths:<br /> There are many well-known approaches and studies that have already identified many canonical driver mutations. A potential strength is that mutation frequencies may be able to identify as yet unrecognized driver mutations. They use a previously developed method to estimate mutation hotspots across the genome (Dig, Sherman et al 2022). This publication has already used cancer sequence data to infer driver mutations based on higher-than-expected mutation frequencies. The advance here is to further illustrate that recurrent mutations (estimated at 3 or more mutations (CDNs) at the same base) are more likely to be the result of selection for a driver mutation (Figure 3). Further analysis indicates that mutation sequence context (Figure 4) or mutation mechanisms (Figure 5) are unlikely to be major causes for recurrent point mutations. Finally, they calculate (Figure 6) that most driver mutations identifiable by the CDN approach could be identified with about 100,000 to one million tumor coding genomes.<br /> Weaknesses:<br /> The manuscript does provide specific examples where recurrent mutations identify known driver mutations but do not identify "new" candidate driver mutations. Driver mutation validation is difficult and at least clinically, frequency (ie observed in multiple other cancer samples) is indeed commonly used to judge if an SNV has driver potential. The method would miss alternative ways to trigger driver alterations (translocations, indels, epigenetic, CNVs). Nevertheless, the value of the manuscript is its quantitative analysis of why mutation frequencies can identify cancer driver mutations.

Recommendations For The Authors<br /> Whereas the analysis of driver mutations in WES has been extensive, the application of the method to WGS data (ie the noncoding regions) would provide new information.

We appreciate that Reviewer #2 has suggested the potential application of our method to noncoding regions. Currently, the background mutation model is based on the site level mutations in coding regions, which hinders its direct applications in other mutation types such as CNVs, translocations and indels. We acknowledge that the proportion of patients with driver event involving CNV (73%) is comparable to that of coding point mutations (76%) as reported in the PCAWG analysis (Fig. 2A from Campbell et al., 2020). In future studies, we will attempt to establish a CNV-based background mutation rate model to identify positive selection signals driving tumorigenesis.

References

Abascal F, Harvey LMR, Mitchell E, Lawson ARJ, Lensing SV, Ellis P, Russell AJC, Alcantara RE, Baez-Ortega A, Wang Y, et al. 2021. Somatic mutation landscapes at single-molecule resolution. Nature:1–6.

Alexandrov LB, Kim J, Haradhvala NJ, Huang MN, Tian Ng AW, Wu Y, Boot A, Covington KR, Gordenin DA, Bergstrom EN, et al. 2020. The repertoire of mutational signatures in human cancer. Nature 578:94–101.

Bailey MH, Tokheim C, Porta-Pardo E, Sengupta S, Bertrand D, Weerasinghe A, Colaprico A, Wendl MC, Kim J, Reardon B, et al. 2018. Comprehensive Characterization of Cancer Driver Genes and Mutations. Cell 173:371-385.e18.

Baylin SB, Jones PA. 2016. Epigenetic Determinants of Cancer. Cold Spring Harb Perspect Biol 8:a019505.

Campbell PJ, Getz G, Korbel JO, Stuart JM, Jennings JL, Stein LD, Perry MD, Nahal-Bose HK, Ouellette BFF, Li CH, et al. 2020. Pan-cancer analysis of whole genomes. Nature 578:82–93.

Drost J, Clevers H. 2018. Organoids in cancer research. Nat Rev Cancer 18:407–418.

Elliott K, Larsson E. 2021. Non-coding driver mutations in human cancer. Nat Rev Cancer 21:500–509.

Grzeskowiak CL, Kundu ST, Mo X, Ivanov AA, Zagorodna O, Lu H, Chapple RH, Tsang YH, Moreno D, Mosqueda M, et al. 2018. In vivo screening identifies GATAD2B as a metastasis driver in KRAS-driven lung cancer. Nat Commun 9:2732.

Hess JM, Bernards A, Kim J, Miller M, Taylor-Weiner A, Haradhvala NJ, Lawrence MS, Getz G. 2019. Passenger Hotspot Mutations in Cancer. Cancer Cell 36:288-301.e14.

Hodis E, Triglia ET, Kwon JYH, Biancalani T, Zakka LR, Parkar S, Hütter J-C, Buffoni L, Delorey TM, Phillips D, et al. 2022. Stepwise-edited, human melanoma models reveal mutations’ effect on tumor and microenvironment. Science 376:eabi8175.

Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, Carter SL, Stewart C, Mermel CH, Roberts SA, et al. 2013. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature 499:214–218.

Makova KD, Hardison RC. 2015. The effects of chromatin organization on variation in mutation rates in the genome. Nat Rev Genet 16:213–223.

Michels BE, Mosa MH, Streibl BI, Zhan T, Menche C, Abou-El-Ardat K, Darvishi T, Członka E, Wagner S, Winter J, et al. 2020. Pooled In Vitro and In Vivo CRISPR-Cas9 Screening Identifies Tumor Suppressors in Human Colon Organoids. Cell Stem Cell 26:782-792.e7.

Ng PK-S, Li J, Jeong KJ, Shao S, Chen H, Tsang YH, Sengupta S, Wang Z, Bhavana VH, Tran R, et al. 2018. Systematic Functional Annotation of Somatic Mutations in Cancer. Cancer Cell 33:450-462.e10.

Sherman MA, Yaari AU, Priebe O, Dietlein F, Loh P-R, Berger B. 2022. Genome-wide mapping of somatic mutation rates uncovers drivers of cancer. Nat Biotechnol 40:1634–1643.

Stamatoyannopoulos JA, Adzhubei I, Thurman RE, Kryukov GV, Mirkin SM, Sunyaev SR. 2009. Human mutation rate associated with DNA replication timing. Nat Genet 41:393–395.

Author response:

The following is the authors’ response to the original reviews

Public Reviews: 

Reviewer #1 (Public review): 

Summary:  

Wang et al. investigate sexual dimorphic changes in the transcriptome of aged humans. This study relies upon analysis of the Genotype-Tissue Expression dataset that includes 54 tissues from human donors. The authors investigate 17,000 transcriptomes from 35 tissues to investigate the effect of age and sex on transcriptomic variation, including the analysis of alternative splicing. Alternative splicing is becoming more appreciated as an influence in the aging process, but how it is affected by sexual dimorphism is still largely unclear. The authors investigated multiple tissues but ended up distilling brain tissue down to four separate regions: decision, hormone, memory, and movement. Building upon prior work, the authors used an analysis method called principal component-based signal-to-variation ratio (pcSVR) to quantify differences between sex or age by considering data dispersion. This method also considers differentially expressed genes and alternative splicing events. 

Strengths:  

(1) The authors investigate sexual dimorphism on gene expression and alternative splicing events with age in multiple tissues from a large publicly available data set that allows for reanalysis. 

(2) Furthermore, the authors take into account the ethnic background of donors. Identification of agingmodulating genes could be useful for the reanalysis of prior data sets. 

Weaknesses:  

The models built off of the GTEx dataset should be tested in another data set (ex. Alzheimer's disease) where there are functional changes that can be correlated. Gene-length-dependent transcription decline, which occurs with age and disease, should also be investigated in this data set for potential sexual dimorphism. 

We appreciate the reviewer’s constructive feedback and acknowledgment of the strengths of our study. The detailed results are included in the ‘Recommendations for the authors’ from the editorial office. Below we summarize our feedback that address the concerns of this reviewer:

(1) Independent Alzheimer’s disease (AD) datasets:

We acknowledge the importance of validating our models beyond GTEx to assess their generalizability aging to Alzheimer’s disease. While GTEx provides valuable transcriptomic data across multiple tissues, it lacks direct functional assessments linked to disease states. We have already analyzed RNA-seq data from ROSMAP and GEO in Figure 4, focusing on sex-biased gene expression and splicing changes between aging and AD.  The results showed a male-biased association with Alzheimer’s disease at AS resolution, indicating that the AS changes during aging could contribute more to AD in males than females. We added a highlight to this analysis in the manuscript (Pages 6-7).

(2) Sexual dimorphism in Gene-Length-Dependent Transcription Decline (GLTD) 

We appreciate the reviewer’s suggestion to explore gene-length-dependent transcription decline (GLTD), which has been implicated in both aging and disease. As the reviewer suggested, our analysis revealed that GLTD exhibits sex-biased patterns in different tissues, aligning with recent literature on sex-dimorphic transcriptional aging. Our findings also revealed that longer genes with greater transcriptional decline are enriched in AD-related pathways. We have incorporated this new analysis in the ‘Recommendations for the authors’ in Author response image 5-6 and expanded the discussion of the biological relevance. 

Reviewer #2 (Public review): 

Summary: 

In this manuscript, Wang et al analyze ~17,000 transcriptomes from 35 human tissues from the GTEx database and address transcriptomic variations due to age and sex. They identified both gene expression changes as well as alternative splicing events that differ among sexes. Using breakpoint analysis, the authors find sex dimorphic shifts begin with declining sex hormone levels with males being affected more than females. This is an important pan-tissue transcriptomic study exploring age and sex-dependent changes although not the first one. 

Strengths:  

(1) The authors use sophisticated modeling and statistics for differential, correlational, and predictive analysis. 

(2) The authors consider important variables such as genetic background, ethnicity, sampling bias, sample sizes, detected genes, etc. 

(3) This is likely the first study to evaluate alternative splicing changes with age and sex at a pan-tissue scale. 

(4) Sex dimorphism with age is an important topic and is thoroughly analyzed in this study.  Weaknesses:  

(1) The findings have not been independently validated in a separate cohort or through experiments. Only selective splicing factor regulation has been verified in other studies. 

(2) It seems the authors have not considered PMI or manner of death as a variable in their analysis. 

(3) The manuscript is very dense and sometimes difficult to follow due to many different types of analyses and correlations. 

(4) Short-read data can detect and quantify alternative splicing events with only moderate confidence and therefore the generalizability of these findings remains to be experimentally validated. 

We appreciate the thorough review and thoughtful feedback. We have addressed the reviewer’s concerns and added clarification. The detailed results are included in Recommendations for the authors. Here are the summaries.

(1) Challenge of independent validation in separate cohorts

• The GTEx dataset includes the most comprehensive transcriptome resource for studying population-level differences in age and sex across tissues, particularly including large-scale brain samples. This provides a unique opportunity to analyze sex-dimorphic aging and the relevance of age-associated diseases.  Several technical issues, including cell type heterogeneity, postmortem artifacts, as well as sequencing biases, lead to technical challenges in different cohorts.

• As the reviewer mentioned, we analyzed transcriptomic data from Shen et al. (2024) and compared them with GTEx results (Author response image 2). Limited overlap in differentially expressed genes again highlighted the challenges in cross-dataset validation due to the differences in cell composition and data processing (peripheral blood mononuclear cells (PBMCs) vs whole blood). 

• Due to the limited human brain transcriptome data covering different age and sex groups, we found mouse hippocampus datasets from Mass spectrometry (MS), including young and old, as well as female and male groups.  The results validated the expression of splicing factors in brain (Author response image 9). This cross-species consistency supports the robustness of our findings in human brain aging.

(2) Effects of Postmortem Interval, Manner of Death, and Time of Death

• We agree that the sample collections could introduce confounding effects. To address this, we calculated the correlations between the confounding factors with Postmortem Interval (PMI), Manner of Death (DTHMNNR), or Time of Death (DTHTIME and DTHSEASON). We observed strong correlations in some surrogate variables in most tissues, indicating that those factors could be well-regressed during our analysis (Recommendations for the authors, Figure S4 and R8). 

• In addition, we re-evaluated our analyses while incorporating PMI as a covariate in our models. Our results align with our initial findings (Author response image 1), suggesting that age- and sex-dependent transcriptomic changes are not strongly confounded by PMI and confirming that our model has controlled PMI. These results are detailed in ‘Recommendations for the authors’ and included in Figure S4C-E with the description in text, Page 5. 

(3) Readability of manuscript and flow of analyses

• In summary, our study first examined global alternative splicing (AS) and gene expression (GE) across all tissues before focusing on specific regions for deeper insights. To improve clarity, we have made the following revisions:

• Add clearer statements when transitioning between all-tissue and brain-specific analyses (Page 6-7).

• Modify the subtitle of Results to highlight all-tissue vs. brain analyses (Page 6).

• These refinements could enhance the manuscript’s structure, making the flow of analysis and conclusions more intuitive for readers.

(4) Limitations of short-read RNA-seq for splicing analysis

• Short-read RNA-seq provides only moderate confidence in detecting and quantifying full-length isoforms. However, its higher sequencing depth makes it more suitable for quantifying changes in alternative splicing (AS) events.

• Our analysis focused on splicing event-level quantification, applying stringent filters and using our GPU-based tool, which showed strong concordance with RT-PCR and other pipelines. Therefore, we also cited and included the updated Paean manuscript that benchmarks its performance in AS analysis.

Reviewer #3 (Public review): 

Summary:  

In this study, Wang et al utilized the available GTEx data to compile a comprehensive analysis that attempt to reveal aging-related sex-dimorphic gene expression as well as alternative splicing changes in humans. 

The key conclusions based on their analysis are that. 

(1) extensive sex-dimorphisms during aging with distinct patterns of change in gene expression and alternative splicing (AS), and 

(2) the male-biased age-associated AS events have a stronger association with Alzheimer's disease, and  (3) the female-biased events are often regulated by several sex-biased splicing factors that may be controlled by estrogen receptors. They further performed break-point analysis and revealed that in males there are two main breakpoints around ages 35 and 50, while in females, there is only one breakpoint at 45. 

Strengths:  

This study sets an ambitious goal, leveraging the extensive GTEx dataset to investigate aging-related, sexdimorphic gene expression and alternative splicing changes in humans. The research addresses a significant question, as our understanding of sex-dimorphic gene expression in the context of human aging is still in its early stages. Advancing our knowledge of these molecular changes is vital for identifying therapeutic targets for age-related diseases and extending the human health span. The study is highly comprehensive, and the authors are commendable for their attempted thorough analysis of both gene expression and alternative splicing - an area often overlooked in similar studies. 

We thank this reviewer for the insightful review and recognition of our study's significance.  We agree with the reviewer on how to examine sex-dimorphic gene expression and alternative splicing in aging by using the GTEx dataset.  This is indeed an essential aspect of developing potential therapeutic targets for agerelated diseases to promote human health span.

Weaknesses:  

Due to the inherent noise within the GTEx dataset - which includes numerous variables beyond aging and sex - there are significant technical concerns surrounding this study. Additionally, the lack of crossvalidation with independent, existing data raises questions about whether the observed gene expression changes genuinely reflect those associated with human aging. For instance, the break-point analysis in this study identifies two major breakpoints in males around ages 35 and 50, and one breakpoint in females at age 45; however, these findings contradict a recent multi-omics longitudinal study involving 108 participants aged 25 to 75 years, where breakpoint at 44 and 60 years was observed in both male and females (Shen et al, 2024). These issues cast doubt on the robustness of the study's conclusions. Specific concerns are outlined below: 

References: 

Ferreira PG, Muñoz-Aguirre M, Reverter F, Sá Godinho CP, Sousa A, Amadoz A, Sodaei R, Hidalgo MR, Pervouchine D, Carbonell-Caballero J et al (2018) The effects of death and post-mortem cold ischemia on human tissue transcriptomes. Nature Communications 9: 490. 

Shen X, Wang C, Zhou X, Zhou W, Hornburg D, Wu S, Snyder MP (2024) Nonlinear dynamics of multiomics profiles during human aging. Nature Aging. 

Wucher V, Sodaei R, Amador R, Irimia M, Guigó R (2023) Day-night and seasonal variation of human gene expression across tissues. PLOS Biology 21: e3001986. 

(1) The primary method used in this study is linear regression, incorporating age, sex, and age-by-sex interactions as covariates, alongside other confounding factors (such as ethnicity) as unknown variables. However, the analysis overlooks two critical known variables in the GTEx dataset: time of death (TOD) and postmortem interval (PMI). Both TOD and PMI are recorded for each sample and account for substantial variance in gene expression profiles. A recent study by Wucher et al.(Wucher et al, 2023) demonstrated the powerful impact of TOD on gene expression by using it to reconstruct human circadian and even circannual datasets. Similarly, Ferreira et al. (Ferreira et al, 2018) highlighted PMI's influence on gene expression patterns. Without properly adjusting for these two variables, confidence in the study's conclusions remains limited at best. 

We appreciate the reviewer for raising this important point regarding the impact of post-mortem interval (PMI) and time of death (TOD) on gene expression, including the death seasons (DTHSEASON) and daytime (DTHTIME). To address this point, we carefully evaluated whether our linear model controlled for these factors as potential confounders. 

Our results showed that PMI and TOD significantly correlated with the estimated covariates in most tissues, suggesting that their effects could be effectively regressed out using our model (Figure S4).  As the reviewers and editors suggested, we have now included this correlation analysis in the updated Figure S4C-E and the text in the Results section, citing relevant literature [1,2] (Page 5). 

Author response image 1.

The results of differential gene expression analysis with vs without the inclusion of PMI correction as a known covariate. The scatter plots show the correlations of significance levels (pvalues, left panel) and effect sizes (coefficients, right panel) of sex (A) and age (B). Whole-blood tissue is used as an example.

In addition, we did the differential analysis that incorporated PMI as a covariate in the regression models and re-evaluated the age- and sex-related transcriptomic changes. Using WholeBlood gene expression as an example, our revised analysis shows that the inclusion of PMI in the covariates has minimal impact on the significance levels and effects of sex and age (i.e., p-values and coefficients, respectively), indicating that our findings are robust using confounding factors (Author response image 1). 

(2) To demonstrate that their analysis is robust and that the covariates TOD and PMI are otherwise negligible - the authors should cross-validate their findings with independent datasets to confirm that the identified gene expression changes are reproducible for some tissues. For instance, the recent study by Shen et al. (Shen et al., 2024) in Nature Aging offers an excellent dataset for cross-validation, particularly for blood samples. Comparing the GTEx-derived results with this longitudinal transcriptome dataset would enable verification of gene expression changes at both the individual gene and pathway levels. Without such validation, confidence in the study's conclusions remains limited. 

We thank the reviewer for the insightful suggestion regarding cross-validation with independent datasets. We understand that validating findings across datasets is crucial for ensuring robustness. As the reviewers suggested, we see whether there are some shared findings in the GTEx data with the study by Shen et al. (2024) in Nature Aging. However, after performing comparisons with our GTEx results in whole blood tissue, we found that the overlaps of differentially expressed genes are limited (Fig. 3). In our results, we found a large proportion of age-associated genes in the GTEx data, whereas just 54 genes are age-associated from Shen et al.’s PBMC data. 3 in 7 genes are differentially expressed in both datasets (Fig. 3A). Additionally, we performed the functional enrichment analysis on the GTEx-specific age-associated genes.

We observed a strong enrichment in the biological pathways related to neutrophil functions and innate immune responses, which are specific to the cell compositions in whole blood rather than PBMC (Fig. 3B).

Author response image 2.

The comparison between the gene expression of whole blood tissue from GTEx and PBMCs from Shen et al. (A) The bar plot shows the number of age (left panel) or sex-associated  (right panel) genes in the two datasets. The grey bars highlight the proportion of overlapped genes in both datasets. (B) The top 10 significantly enriched biological processes in the GTEx-specific age-associated genes. The color bar shows the number of age-associated genes in specific pathways.

These discrepancies highlighted the crucial factors in cross-dataset comparison:

• Cell compositions: GTEx used whole blood, which contains all blood components, including neutrophils and erythrocytes, whereas PBMCs contain lymphocytes and monocytes. Under the influence of granulocytes and red blood cells in whole blood, the gene expression profiles between these two datasets are different.

• Biological functions: Whole blood includes both innate and adaptive immune components; thus, aging-related gene expression changes in whole blood may include a broader systemic response than those in PBMCs. This difference in biological context contributes to the observed variation in the differentially expressed genes, as demonstrated by our functional enrichment analysis (Fig. 3B). 

• Sequencing biases and data processing: The two datasets were generated using different RNAseq processing pipelines, including distinct normalization, batch correction, and quantification methodologies. These technical differences may introduce systematic variations that complicate direct cross-validation.

Due to these fundamental problems, a direct one-to-one validation between the two datasets is challenging. We understand the importance of independent dataset validation and appreciate the reviewer’s suggestion. However, future studies could be performed more precisely if comparable whole-blood-based datasets are available. In addition, GTEx data provides nearly thousands of samples in whole blood, which is a largescale, comprehensive, and clinically relevant dataset for studying aging-related changes, particularly in innate immunity and inflammation, which are not well captured in PBMCs.

(3) As a demonstration of the lack of such validation, in the Shen et al. study (Shen et al., 2024), breakpoints at 44 and 60 years were observed in both males and females, while this study identifies two major breakpoints in males around ages 35 and 50, and one breakpoint in females at age 45. What caused this discrepancy? 

We thank the reviewer and the editors for both coming up with the non-linear multi-omic aging patterns observed by Shen et al.  They observed two prominent crests around the ages of 45 and 60 from omics data.

Similarly, we also identified two breakpoints in our analysis, with some differences in specific age breakpoints. These could be the result of sample preparation methods and breakpoint definition. These responses are also included in the editor’s recommendations.

Definition of breakpoints vs crests:

• Crests represent age-related molecular changes at each time point across the human lifespan. They indicate the number of molecules that are differentially expressed during aging (q < 0.05), without considering individual expression levels.

• Our breakpoints, in contrast, are identified after filtering the chronological trends using the Autoregressive Integrated Moving Average (ARIMA) model. We calculated the rate of change at each age point using the smooth approach and sliding windows. Breakpoints are defined as local maxima where the distance to the nearest minimum, relative to the global maximum. We indeed found some local wide peaks around 60 in some tissues, shown in Figure S10, however, we excluded these due to our strict cutoffs to remove noise.

Differences and similarities between sequenced tissues: 

• Whole-blood vs PBMC: In the GTEx RNA-seq data used in our study, whole blood samples from donors were sequenced, whereas their study used PBMCs. Whole blood contains all blood components, including red blood cells, platelets, granulocytes (e.g., neutrophils), lymphocytes, and monocytes, while PBMCs represent a subset of white blood cells, primarily consisting of lymphocytes (T cells, B cells, NK cells) and monocytes, excluding granulocytes and erythrocytes. As we mentioned in the previous responses, the gene expression changes observed in whole blood capture the contributions of neutrophils and other granulocytes, which are neglected in the PBMC profile (also shown in Figure S11C). 

• For the shared tissues in two studies – skin, we looked at the non-linear changes during aging and found the same two breakpoints: 43 and 58. 

Novelties in our study:

• Whole blood can serve as a readily accessible resource for testing age-related disease biomarkers without cell separation, making it more practical for clinical applications.

• Our analysis was performed on females and males, respectively. The main object of our analysis is to compare the differences in aging rates between sexes. Our results reveal clear sex-specific differences across multiple human tissues. Therefore, the identified breakpoints may differ when sex effects are not taken into account, highlighting the specificity of our analysis. 

• Additionally, our breakpoints are integrated across multiple tissues. Our results showed that there is a large diversity of aging patterns in different tissues.

As the reviewers and editors suggested, we have added the following statements to clarify this distinction in the Discussion section: ‘Our analysis observed the non-linear aging patterns with two breakpoints, which is consistent with recent findings, with differences in specific age points due to sex differences as well as tissue diversities 3.’ (Page 14), and ‘These breakpoints could represent key junctures in the aging process that align with the non-linear patterns of aging and disease progression.’ (Page 15)

(4) Although the alternative splicing analysis is intriguing, the authors did not differentiate between splicing events that alter the protein-coding sequence and those that do not. Many splicing changes occurring in the 5' UTR and 3' UTR regions do not impact protein coding, so it is essential to filter these out and focus specifically on alternative splicing events that can modify protein-coding sequences. 

The reviewer raises an important point. In our study, we included the AS events in protein-coding genes to gain a comprehensive understanding of sex-biased age-associated splicing. As the reviewer suggested, focusing on coding-sequence-altering events is particularly relevant to protein function. To address this, we performed an additional analysis to specifically annotate sBASEs occurring within the coding sequence (defeined as CDS-altering sBASEs) and reanalyzed their functional pathways and AD-associations (Author response image 3).  

Our analysis revealed that most of the sBASEs are relevant to protein-coding sequences (CDS) across multiple tissues (Author response image 3A).  We then confirmed our findings using CDS-altering sBASEs. We found that those sBASEs in brain regions were significantly enriched in pathways related to amyloid-beta formation and actin filament organization (Author response image 3B). Notably, male-biased sBASEs in decision-related brain regions were particularly associated with dendrite development and regulation of cell morphogenesis, highlighting the sex-specific roles of sBASEs in brain functions. Additionally, we performed a random forest classification using only CDS-altering sBASEs in AD datasets (Author response image 3C-D), again confirming the malebiased association between aging and AD.

Overall, we found that most of the identified sBASEs could modify protein-coding sequences, and our main conclusions remain consistent even after filtering out non-coding events. 

Nevertheless, in addition to AS events that impact protein sequences, alternative splicing in untranslated regions (UTRs) also plays a critical regulatory role. Splicing events in the 5′ UTR can influence translation efficiency by modifying upstream open reading frames (uORFs) or RNA secondary structures, while splicing in the 3′UTR can affect mRNA stability, localization, and translation by altering microRNA binding sites and RNA-binding protein interactions. Given these functional implications, we believe that UTR-targeted AS events should also be considered to supplement the understanding of post-transcriptional gene regulation in future research.

Author response image 3.

The distribution and functional relevance of sBASEs with coding effects. (A) The number of sBASEs and CDS-altering sBASEs across multiple tissues. The deeper bars show the number of sBASEs whose alternative splice sites are located at protein-coding regions. (B) GO biological pathways in each sex and brain region. Heatmap shows the sex-specific pathways that are significantly enriched by CDS-altering sBASEs in more than 2 brain regions and sex. (C) Correlation between ADassociated and age-associated AS changes across the CDS-altering sBASEs that alter protein-coding sequences in females and males. (D) Performances of sex-stratified models predicted by CDS-altering sBASEs in 100 iterations using the random forest approach

(5) One of the study's main conclusions - that "male-biased age-associated AS events have a stronger association with Alzheimer's disease" - is not supported by the data presented in Figure 4A, which shows an association with "regulation of amyloid precursor formation" only in female, not male, alternative splicing genes. Additionally, the gene ontology term "Alzheimer's disease" is absent from the unbiased GO analysis in Figure S6. These discrepancies suggest that the focus on Alzheimer's disease may reflect selective data interpretation rather than results driven by an unbiased analysis. 

We thank the reviewer for this point. In our functional analysis, we identified distinct biological processes enriched in female- and male-biased AS genes, such as the regulation of amyloid precursor formation in females and structural constituents of the cytoskeleton in males. However, Alzheimer’s disease (AD) is a complex neurodegenerative disorder with multiple pathological mechanisms beyond amyloid-beta (Aβ) formation, many of which are strongly age-related in both sexes. This complexity motivates us to explore novel relationships between splicing and AD in distinct sexes.

Although Figure 4A shows the enrichment of “regulation of amyloid precursor formation” in female-biased AS events, this does not contradict the broader enrichment of AD-related processes in male-biased AS events. Our disease ontology analysis supports this finding, as male-biased age-associated AS events are enriched in neurodegenerative diseases, including cognitive disorders. Additionally, we considered not only individual GO terms but also the disease-associated transcriptomic signatures from AD-related datasets, which collectively indicate a stronger association in males. 

Regarding Figure S6 mentioned by the reviewer, the GO term “Alzheimer’s disease” is not explicitly listed in the heatmap because we filtered the pathways that are consistently enriched in multiple tissues. As noted in the figure legend, we only displayed sex-specific GO terms that were significant in at least 15 tissues. Then, since the brain is highly affected by age-related processes and neurological conditions show sex differences, the sex-biased AS events could help explain differential susceptibility to age-related cognitive decline and neurodegeneration. That’s why we chose the brain data for detailed analysis.

To improve clarity, we have revised the text to describe the purpose of our analysis in brain rather than other tissues (Page 6-7). We appreciate the reviewer’s feedback, and we will consider additional analyses to further explore the sex-biased AS as well as disease risk in other tissues.

(6) The experimental data presented in Figures 5E - I merely demonstrate that estrogen receptor regulates the expression of two splicing factors, SRSF1 and SRSF7, in an estradiol-dependent manner. However, this finding does not support the notion that this regulation actually contributes to sex-dimorphic alternative splicing changes during human aging. Notably, the authors do not provide evidence that SRSF1 and SRSF7 expression changes actually occur in a sex-dependent manner with human aging (in a manner similar to TIA1). As such, this experimental dataset is disconnected from the main focus of the study and does not substantiate the conclusions on sex-dimorphic splicing during human aging. The authors performed RNAseq in wild-type and ER mutant cells, and they should perform a comprehensive analysis of ER-dependent alternative splicing and compare the results with the GTEx data. It should be straightforward. 

Thanks for the reviewer’s feedback. The main purpose of the analyses in Figures 5E-I was to explore which factors affect the sex-biased expression of splicing factors during aging and substantially regulate alternative splicing (AS). To address the reviewer’s concerns, we have included additional analysis and explained the challenge of linking estrogen receptor (ER)-regulated splicing factors to sex-dimorphic AS changes during human aging in specific human cell types. 

• As suggested by the reviewer, we first examined the expression changes of SRSF1 and SRSF7 during aging in males and females, like TIA1 in decision-related brain regions (Fig. 5I).

• Secondly, the regulation is based on a highly complex regulatory network involving multiple splicing factors and cell heterogeneity. Due to these complexities, we did not overlap ER-dependent AS changes with sBASEs from GTEx datasets directly. As far as the reviewer is concerned, we supplemented the AS analysis in the GSE89888 dataset (Fig. 5H) and identified the estrogenregulated AS events mediated by ESR1. We found that ~6% (26/396) of female-specific ageassociated AS events were regulated by ESR1, of which 6 sBASEs can be regulated by femalebiased splicing factors. The low overlaps could be represented by the limited coverage of different RNA-seq datasets and cell types used across these analyses. Notably, the results indicated that only a fraction of AS could be directly accounted for by estrogen via ESR1, suggesting the complexity of transcriptional and splicing regulatory networks during aging. 

• Meanwhile, we downloaded independent experimental datasets to discover the regulation by our candidate splicing factors. Due to SRSF1 is identified as a potential regulator of sex-biased splicing, we analyzed RNA-seq data with SRSF1 knock-down (KD) glioblastoma cell lines (U87MG and U251), a type of brain cancer formed from astrocytes that support nerve cells 4.  As a result, we indeed found that some sBASEs are regulated by SRSF1 during aging through this experiment using brain cell lines (Author response image 4). Together, these results suggested that some of the SF-RNA regulatory relationships can be observed in another cellular system, further supporting our findings. 

Due to the limitations of cell-based models and the complexity in the splicing regulatory network, it is challenging to directly validate aging regulation, particularly between different sexes, based on ER treatments in vivo. However, our findings still provide valuable mechanistic insights into ER-regulated splicing factors, implying their potential role in sex-biased aging.

Author response image 4.

SRSF1 regulations on specific sBASEs using SRSF1 knock-down RNA-seq data in GBM cells. Three examples are shown to be regulated during aging with significant changes between SRSF1 KD vs control in U251 and U87MG cell lines. The splicing diagrams are shown below.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors): 

The authors found that alternative splicing was affected by both sex and age across many tissues, with gene expression differences affected by both parameters only present in some tissues. This trend was consistent when the effects of sex chromosomes were subtracted from the analysis. The effect of aging on differential gene expression and alternative splicing was more prevalent in male than female samples. For analysis purposes, young subjects were deemed to be anyone under 40, and old subjects were over 60 years old. The authors then investigated if specific genes or alternative splicing events were responsible for these effects. Some candidate genes or splicing events were identified but there was little overlap between tissues, suggesting no universal gene or event as a driver of aging. Surrogate variables like the ethnic backgrounds of donors were also investigated. Ultimately the authors found that alternative splicing events showed a stronger sexual dimorphic effect with age than did differential gene expression and that at least for the brain, alternative splicing changes showed a bias for Alzheimer's disease in male samples. This was highlighted by examples of exon skipping in SCL43A2 and FAM107A in males that were associated respectively with plaques and tangles. 

The authors go on to identify sexual dimorphic differences in splicing factors in particular brain regions during age. Finally, the authors performed analysis for aging-modulated genes, identifying nearly 1000 across the tissues, nearly 70% of which are sex-specific. Their work suggests that further analysis of these aging-modulated genes could be differentially modulating the transcriptome based on sex. The work is novel and interesting, especially investigating sexual dimorphism in alternative splicing. However, the work is still preliminary, and these assumptions need to be applied to other data sets beyond GTEx for validation as well as some other phenomena that need to be considered. I recommend major revisions to address the points below. 

(1) At the beginning of the results section, the authors state that the brain is stratified into four functional regions. It would be useful to explicitly state those four regions in the text at that point. 

We agree that specifying these regions early in the text will improve clarity and provide the reader with a clear understanding of the analysis. As the reviewer’s suggestion, we revised the Results section (Page 3) to explicitly state the four functional brain regions as follows: ‘Due to data sparseness, the brain tissues were recombined into four functional regions (table S1), including hormone- or emotion-related region, movement-related region, memory-related region, and decision-related region (See Methods).’. This ensures that the regions are clearly defined before the subsequent analysis is presented. 

(2) The manuscript becomes a bit confusing when the authors shift from all the tissues as a whole specifically to the brain and then back to the larger tissue set to make assumptions. This can be a bit confusing and should be better delineated.

We thank the reviewer and editor for the feedback regarding the transitions between the analysis of all tissues and the brain-specific analysis. In our study, we first conducted a broad analysis of alternative splicing (AS) and gene expression (GE) across all tissues. For the AS analyses, we did sBASEs analysis in all tissues and then focused on specific tissue (i.e., brain) whose splicing changes are functionally enriched with age-related diseases.  For the GE analyses, we also analyzed the aging rate across tissues and identified the tissue-specific/shared patterns. 

We agree that the shifts of the tissues for AS and GE may cause some confusion, and have made the following revisions to delineate why we focused on different tissues for distinct analyses:

• We have added clear statements to better delineate when we shift focus from the analysis of all tissues to the region-specific analysis and vice versa. For instance, in the Results section (Page 67), we include a transitional phrase: ‘Having established patterns across all tissues, we now turn to a more focused analysis to investigate tissue-specific alternative splicing changes.’

• To improve the overall structure, we have reorganized the Results section, adding distinct subheadings for the analysis of all tissues and the brain (Page 6), which should make the transition between these sections smoother and more intuitive for the reader.

We believe that these revisions will make the manuscript’s structure clearer and allow the reader to better follow the flow of the analysis and the subsequent conclusions.

(3) Gene-length-dependent transcription decline (GLTD) is another phenomenon that occurs with aging and is known to be associated with Alzheimer's disease [PMID38519330]. The authors should make some statement if this is present in their dataset and if any sexual dimorphism in tissues is present. 

We thank the editors and reviewers for bringing up the possible connection of gene-length-dependent transcription decline (GLTD), which was reported to be associated with both aging and Alzheimer’s disease (AD). We appreciate the reviewer’s suggestion and have addressed whether GLTD is present in our dataset and whether any sex differences are observed in this context.

We evaluated GLTD using the correlation between gene length with age-associated changes (i.e., the coefficients of the ‘age’ term in the linear regression model) in GTEx data. We did observe strong evidence of GLTD, particularly in the brain, heart, muscle, pancreas, spleen, skin, muscle, etc (Author response image 5A). In brain, we performed the functional enrichment analysis on the genes with Foldchange > 2 and length > 10⁵ bp (Author response image 5B). We found that these extremely long genes are significantly relevant to synapse and neuron functions. These findings align with previous studies showing that GLTD can occur with aging in the tissues that are relevant to Alzheimer’s disease, cardiovascular diseases, and common failures of metabolism (e.g., diabetes) [5,6]. Additionally, it was not a ubiquitous phenomenon across all tissues. The correlations could be positive in tissues like adipose and artery.  These findings suggested the GLTD could be varied and tissuespecific in its manifestation during aging. 

Author response image 5.

(A) The correlation between gene length and age-associated changes across GTEx tissues in human samples. The correlation tests are evaluated using Spearman’s approach. The color bar indicates the -log10 transformed p-values in the correlation test. (B) The results of GO enrichment analysis using the genes with Foldchange > 2 and length > 10⁵ bp. The parent terms calculated by ‘rrvgo’ with a similarity threshold of 0.9 are shown.

Regarding sexual dimorphism, we conducted this analysis in females and males, respectively (Author response image 6). We found GLTD exists in both females and males in most tissues, such as brain, whole blood, muscle, etc, consistent with the previous results without considering the sex groups. Interestingly, we observed sexbiased patterns in certain tissues. In particular, the left ventricle, pancreas, and hippocampus showed notable male-biased patterns in the degree of transcriptional decline with gene length, whereas skin, liver, small intestine, and esophagus showed that in females. These findings suggest that GLTD could be relevant to aging and age-related diseases; the levels of expression and sexual dimorphism may vary depending on the tissue type. We hope this clarification addresses the reviewer’s concern and provides a more comprehensive understanding of the GLTD and sex differences observed in our dataset. 

Author response image 6.

The correlation between gene length and age-associated changes across tissues in females and males, respectively. The correlation tests are evaluated using the Spearman’s approach. The red dots indicate the significant correlations in females, while the navy dots show those in males.

(4) Because the majority of this work has been performed in the GTEx dataset, applying this analysis to another publicly available dataset would be useful validation. For instance, the authors have interesting findings in the brain and correlations to Alzheimer's disease. Analysis of an existing RNAseq dataset from Alzheimer's disease patients and controls (with functional outcomes) would provide more evidence beyond the preliminary findings from GTEx. 

We appreciate the reviewer’s suggestion on the validation of our findings by applying our analysis to independent RNA-seq datasets from Alzheimer’s disease patients. 

• We have used two Alzheimer’s disease datasets, GEO and ROSMAP, to investigate the correlation between aging and Alzheimer’s disease (AD) and included these analyses in our study (Fig. 4B-C and Figure S8C).

• In the Results section (Page 7), we have presented the results of this validation, where we identified correlations between sex-biased aging-related splicing changes and AD-related changes. These findings support the conclusions from the GTEx dataset and further strengthen the relevance of our results to AD.

As suggested, we have updated the manuscript to more explicitly highlight this validation in the Discussion section (Page 12), noting: ‘We further validated our findings using Alzheimer’s disease dataset, ROSMAP, where we observed consistent correlations between aging-related splicing changes and Alzheimer’s disease-related changes, providing additional evidence for the robustness of our results.’ 

Reviewer #2 (Recommendations for the authors): 

(1) In the text (Introduction and Discussion), the authors mention analyzing 54 tissues, the abstract states 35 tissues, Table S1 lists 48, and Figure 2A-B shows 33. Could the authors please clarify exactly how many tissues they used? I am also confused by the sample numbers in Table S1. For example: for adiposesubcutaneous tissue, the total number of females is listed as 218 but the sum of young and old females is only 110. Does this mean some samples were excluded? What is the exclusion criterion? 

We thank the reviewers and editors for pointing out the discrepancies regarding the number of tissues analyzed and the sample numbers in Table S1. We appreciate the opportunity to clarify these points:

Number of tissues analyzed:

• We downloaded and analyzed 17,382 samples in 54 tissues from GTEx in total (31 tissues and 13 brain regions), as mentioned in the Results, Methods, and Discussion sections. Table S1 lists 48 tissues (31 tissues, 13 brain regions, and 4 merged brain regions), which include a refined classification of the tissues we analyzed, accounting for the variations in brain region categorization in the dataset.

• The discrepancy also arises from the different sample size cutoffs in specific analyses. For pcSVR analysis (Figure 2A-B), we did the subsampling for the permutation analysis for certain key findings, so we filtered a subset of 33 tissues (29 tissues and 4 merged brain regions), which included at least 3 samples in each age group in females or males. 

• To resolve this, we have clarified the total number of tissues analyzed and aligned the numbers across the manuscript. In the revised manuscript, we now explicitly state in both the Abstract and Methods sections that 54 tissues were analyzed in the context of this study. We added a note in Methods to clarify that 35 tissues are 31 tissues and 4 merged brain regions (Page 16). In Figure 2A-B, we clarified that the 33 tissues are filtered due to the usage in this analysis (Page 17).

Sample numbers in Table S1:

• Regarding the sample sizes of age groups, the discrepancy occurred due to the classification of the age groups. We classify the samples into three: Young, Middle, and Old, as mentioned in the Results section (Page 4). 

• Additionally, we excluded the sample sizes in 13 single brain regions. We aligned the total tissue number to 35 with our texts.

We hope this resolves the confusion regarding the number of tissues and the sample sizes used in the analysis. These clarifications have been incorporated into the revised manuscript to ensure consistency.

(2) Was post-mortem interval (PMI) or manner of death considered in the model? For example, traumatic death may have major consequences on gene expression. Similarly, a few tissues have low sample numbers, for example, kidney cortex and brain. The pooling of brain samples is explained and the kidney cortex is excluded, so why is it listed in Table S1? 

Thank you for raising this important point regarding the potential impact of post-mortem interval (PMI) and manner of death (DTHMNNR) on gene expression. We carefully considered both factors as potential confounders in our analysis. 

Specifically, to evaluate their impacts, we calculated the correlations between the coefficients of PMI or manner of death, with the confounding factors. Our results showed that PMI and DTHMNNR are significantly correlated with the covariates in most tissues, suggesting that their effects could be effectively regressed in our model (Figure S4). As we have mentioned in Figure S4 and Author response image 1, we conducted a differential analysis that incorporated PMI as a covariate in the regression models and re-evaluated the age- and sex-related transcriptomic changes to address this concern. The high correlations showed the minor effect size of PMI when including the covariates in the model. As suggested by the reviewers and editors, we have now included this correlation analysis in Figure S4C-E and updated the text in the results section (Page 5).

Additionally, as the responses above, Table S1 provides the general sample sizes of all GTEx tissues without filtering. We have modified the table to include a total of 35 tissues, including 31 non-brain tissues and 4 brain regions.

(3) It might be important to show a simple visual of cohort details such as age ranges, sexes, ethnicities, PMIs, etc. 

To address this, we added summary figures to illustrate the distributions of key demographic variables, including age, sex, BMI, ethnicity, post-mortem intervals (PMIs), and manner of death (DTHMNNR) (Author response image 7 and Author response image 8). This will provide readers with a clearer overview of the dataset composition and potential covariates affecting the analysis. 

Author response image 7.

Age (left panel), BMI (Body Mass Index) (middle panel), and PMI (Post-Mortem Interval) (right panel) distribution in GTEx v8 cohort.

Author response image 8.

Sex (left panel), ethnicity (middle panel), and manner of death (DTHMNNR) (right panel) distribution in GTEx v8 cohort.

(4) Since this study is highly correlative, it is impossible to determine if the findings hold true without an independent cohort validation or experimental validation. They used the ROSMAP cohort for AD samples, and some splicing factors regulation but the generalizability to the age and sex effects have not been independently tested.

The reviewer raises an important point regarding the independent validation of sex- and age-associated splicing changes associated with AD. We used GTEx primarily because it includes approximately 17,000 RNA-seq samples across multiple human tissues, making it the most comprehensive public resource for studying population-level differences in age and sex. In particular, its large-scale brain samples provide a unique opportunity to analyze transcriptomic changes in sex-dimorphic aging.

We understand the reviewer’s concern that our findings are mainly supported by correlative evidence, which could be affected by dataset-specific biases. However, there are several technical issues in crossvalidation with transcriptomes across different datasets, including limited comparability due to cell type heterogeneity, postmortem artifacts, and sequencing biases.

Specifically, GTEx data is bulk RNA-seq that does not capture cell-type-specific transcriptomic changes. Given the cellular complexity of the brain and other tissues, observed differences in gene expression and splicing may be influenced by shifts in cellular composition rather than intrinsic transcriptional regulation. For example, we compared our results from GTEx whole blood with the analysis using an external dataset from Peripheral Blood Mononuclear Cells (PBMCs) provided by Shen et al. (2024) [3] (Author response image 2).  We observed limited overlap in differentially expressed genes between these datasets (probably because the whole blood contains diverse immune cell populations), highlighting the challenges in cross-dataset validation due to differences in tissue composition and sample processing.

Therefore, we applied surrogate variable analysis (SVA) to minimize technical and biological confounders. This approach helped reduce biases from genetic background to hidden batch effects, including postmortem artifacts, sequencing biases (Figure S4), and other covariates. This approach could help us identify whether sex-biased splicing events are biologically meaningful rather than technical artifacts.  

In addition, to address the reviewer’s concern on the splicing factor regulation, we managed to find a dataset in decision-related brain regions. Due to the limitation of human brain data covering different age and sex groups, we used mouse hippocampus datasets, including young and old, as well as female and male groups [7].  The analysis of protein levels from MS data identified sex-biased age-associated splicing factors, including Srsf1 and Srsf7.  We found that the changes are consistent with the findings from GTEx (Author response image 9), aligning with our sex-biased splicing factor expression during aging in the same region of the human brain. This cross-species consistency supports the robustness of our findings in human brain aging.

Author response image 9.

Protein levels of some male-specific splicing factors in human hippocampus quantified using MS data. The Y-axis shows the protein intensity. Different facets mean different sample batch sets. The yellow boxes indicate the protein levels in the young group, while the brown boxes indicate those in the old group.

In summary, despite the inherent limitations of RNA-seq studies in sex- and age-related transcriptomics, we have made our best efforts to address these concerns through comparisons with external datasets, statistical corrections, and validation using proteomic data. We appreciate the reviewer’s feedback and include additional discussion on these points (Page 13). 

(5) Are AS predictions from short-read data accurate enough to make the predictions the authors report? 

The reviewer is correct that the short-read sequencing has inherent limitations in reconstructing full-length isoforms.  However, the higher sequencing depth for short reads makes it a better choice in quantifying the relative change of each AS event across different conditions.  As a result, short-read data are extensively used in the splicing field to quantitatively measure the AS changes.  For this reason, we focused on the levels of alternative splicing events, rather than the quantification of full-length isoforms.  We used a series of stringent filters in our analyses to increase the reliability of our results.

Specifically, we filtered the read counts of the junction read counts (JC) of most differential AS events that were higher than 10, as mentioned in the Methods section. Also, we used our GPU-based gene expression quantification tool, Paean, which performed better in cross-validation with quantitative RT-PCR results. The results of Paean are consistent with other pipelines. We cited an updated version of Paean that included the comparison with other tools in analyzing AS for consistency.  The manuscript on the new Paean version is being reviewed in another journal, and we included the PDF of that manuscript (Fig. 3 in the Paean manuscript) in the revised documents. 

(6) Along the same lines, the finding that male age-related AS events are linked to Alzheimer's disease somewhat contradicts epidemiological studies that show that even after adjusting for age, women still have a greater risk of developing Alzheimer's than men. The authors show a significant overlap with AD GE events in females but don't explain the discrepancy. 

We appreciate the editor’s comment regarding these discrepancies with the epidemiological studies. Previous studies suggested that the disease manifestations of Alzheimer’s Disease (AD) showed sex differences in AD phenotypes, including cognitive decline and brain atrophy [8].  The analyses on the sex/age effect of AD are indeed pretty complex, depending on the molecular criteria (GE or AS vs epidemiological data) in distinct studies, probably due to the difficulty in capturing how environmental exposures interact with biological pathways.  We hope to bring up three related points regarding this concern, which were also discussed in the revised manuscript. 

• As we have mentioned in the Discussion section, an early study investigated the relationship between age, sex, and cognitive function in a large cohort of 17,127 UK Biobank participants [9]. Their study highlighted more apparent age-related changes in cognitive function among men, suggesting a potential vulnerability of men to cognitive decline with age.  Their main conclusion is consistent with our findings. 

• While men and women can both suffer from Alzheimer's disease, women are more likely to be diagnosed, possibly due to longer lifespans and potential differences in brain structure or other factors. Although women exhibit a higher overall risk of AD, they may also have distinct molecular compensatory mechanisms that influence disease progression. 

• To avoid the age effect, in our AD datasets, including ROSMAP, we filtered the samples over 90 years old to match the number of both sexes and the age distribution between the AD and control groups. Our analysis avoided the age biases in comparing AD and control, suggesting the crucial roles of sBASEs in AD during male aging.

Moreover, for gene expression (GE), we showed distinct patterns of AD-related genes in females with AS. These two molecular processes do not necessarily have the same functional impact. AS changes may precede or contribute to disease onset in different ways compared to GE alterations. Our study came up with the underlying mechanisms linking cognitive disorders and alternative splicing (AS) at a higher molecular resolution.   

(7) Could the authors explain which sBASE subset they used for their random forest prediction model and what was the rationale? 

We are sorry for missing the details in selecting sBASEs (sex-biased age-associated splicing events) for the random forest prediction model. We specifically used sBASEs that exhibited specific sex-biased changes in splicing associated with aging. This subset of sBASEs was chosen in terms of those that could also be detected in the ROSMAP AD dataset due to different sequencing depths or technical biases across datasets. These sBASEs were further input to a prediction model with the feature selection algorithm RFE, and then evaluated their contributions. In the revised manuscript, we added the details of this selection in the Methods (Page 7).

(8) The breakpoint analysis is particularly interesting. Can this be speculated to correlate with the recent non-linear multi-omic aging patterns observed by Shen et al in Nature Aging? 

Thank you for highlighting the interesting aspects of our breakpoint analysis and suggesting its potential correlation with the non-linear aging patterns observed by Shen et al. 

Shen et al. observed two prominent crests around the ages of 45 and 60 using omics data. Similarly, we also identified the non-linear aging patterns with two breakpoints in our analysis. However, there are some notable differences in specific breakpoints between these two studies, resulting from the breakpoint definition, as well as the sample preparations. According to the response in Author response image 2, the differences come from the following aspects:

The definition of breakpoints vs crests:

• Crests represent age-related molecular changes at each time point across the human lifespan. They indicate the number of molecules that are differentially expressed during aging (q < 0.05), without considering individual expression levels.

• Our breakpoints, in contrast, are identified after filtering the chronological trends based on the expression levels and calculating the rate of change at each age point using sliding windows. Breakpoints are defined as local maxima where the distance to the nearest minimum, relative to the global maximum, exceeds 10%. We indeed found some local wide peaks around 60 in some tissues, shown in Figure S10, however, we excluded these due to our strict cutoffs.

The sequenced biosamples: 

• Whole-blood vs Peripheral Blood Mononuclear Cells (PBMC): As mentioned in previous responses, in GTEx, whole blood samples from donors were sequenced, whereas their study used PBMCs. Whole blood contains all blood components, including red blood cells, platelets, granulocytes (e.g., neutrophils), lymphocytes, and monocytes, while PBMCs only represent a subset of white blood cells, primarily consisting of lymphocytes (T cells, B cells, NK cells) and monocytes, excluding granulocytes and erythrocytes. Gene expression changes observed in whole blood capture the contributions from neutrophils and other granulocytes, which are absent in PBMC analyses (as shown in Figure S11C and Author response image 2). Additionally, whole blood can serve as a readily accessible biomarker source for testing age-related diseases without the need for cell separation, making it a more practical option for clinical applications.

• For both studies, we share a tissue, which is skin, we looked at the non-linear changes during aging and found the same two breakpoints: 43 and 58. 

Sex-specific analysis in females and males:

• The main object of our analysis is to compare the differences in aging rates between sexes. Notably, the identified breakpoints may differ when sex effects are not taken into account, highlighting the importance of analyzing males and females separately.

We have added the following statements to further clarify this connection: ‘Our analysis observed the nonlinear aging patterns with two breakpoints, which is consistent with recent findings (Nature Aging, 2024), with differences in specific age points due to the sex differences as well as tissue diversities.’ (Page 14), and ‘These breakpoints could represent key junctures in the aging process that align with the non-linear patterns of aging and disease progression.’ (Page 15)

(9) Minor - the authors should refer to figures in the Discussion. They do so in some cases but this needs to be more extensive. 

Thank you for pointing this out. In response, we have reviewed the Discussion section and added references to relevant figures where appropriate. In the section discussing the discrepancies between the profiles of GE vs. AS, we now refer to Figure 3 to highlight the earlier onset of different transcriptomic resolutions (Page 12); When describing the sex-specific age-associated AS changes and their associations with Alzheimer’s disease, we have added references to Figure 4 (Page 12); In the discussion of estrogen-mediated regulation of splicing factors, we have referred to Figure 5A, which detail the construction of RBP-RNA regulatory network integrating muti-dimensional data obtained through several orthogonal state-of-the-art approaches (Page 14).

Reference:

(1) Ferreira, P.G. et al. The effects of death and post-mortem cold ischemia on human tissue transcriptomes. Nature communications 9, 490 (2018).

(2) Wucher, V., Sodaei, R., Amador, R., Irimia, M. & Guigó, R. Day-night and seasonal variation of human gene expression across tissues. PLoS Biology 21, e3001986 (2023).

(3) Shen, X. et al. Nonlinear dynamics of multi-omics profiles during human aging. Nature aging, 116 (2024).

(4) Zhou, X. et al. Splicing factor SRSF1 promotes gliomagenesis via oncogenic splice-switching of MYO1B. The Journal of clinical investigation 129, 676-693 (2019).

(5) Soheili-Nezhad, S., Ibáñez-Solé, O., Izeta, A., Hoeijmakers, J.H. & Stoeger, T. Time is ticking faster for long genes in aging. Trends in Genetics 40, 299-312 (2024).

(6) Brouillette, M. Gene length could be a critical factor in the aging of the genome. Proceedings of the National Academy of Sciences 121, e2416630121 (2024).

(7) Keele, G.R. et al. Global and tissue-specific aging effects on murine proteomes. Cell reports 42(2023).

(8) Ferretti, M.T. et al. Sex differences in Alzheimer disease—the gateway to precision medicine. Nature Reviews Neurology 14, 457-469 (2018).

(9) Foo, H. et al. Age-and sex-related topological organization of human brain functional networks and their relationship to cognition. Frontiers in aging neuroscience 13, 758817 (2021).

Author Response

The following is the authors’ response to the original reviews.

eLife assessment

The work is a useful contribution towards understanding the role of archaeal and plant D-aminoacyl-tRNA deacylase 2 (DTD2) in deacylation and detoxification of D-Tyr-tRNATyr modified by various aldehydes produced as metabolic byproducts in plants. It integrates convincing results from both in vitro and in vivo experiments to address the long-standing puzzle of why plants outperform bacteria in handling reactive aldehydes and suggests a new strategy for stress-tolerant crops. The impact of the paper is limited by the fact that only one modified D-aminoacyl tRNA was examined, in lack of evidence that plant eEF1A mimics EF-Tu in protecting L-aminoacyl tRNAs from modification, and in failure to measure accumulation of toxic D-aminoacyl tRNAs or impairment of translation in plant cells lacking DTD2.

We have now addressed all the drawbacks as follows:

‘only one modified D-aminoacyl tRNA was examined’

We wish to clarify that only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009) and D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007) and it also recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure: 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

Author response image 1.

DTD2 recycles multiple D-aa-tRNAs with different side chain chemistry and size. Growth of wildtype (WT), dtd null strain (∆dtd), and Pyrococcus horikoshii DTD2 (PhoDTD2) complemented ∆dtd strains of E. coli K12 cells with 500 µM IPTG along with A) no D-amino acids, B) 2.5 mM D-tyrosine, C) 30 mM D-aspartate and D) 5 mM D-tryptophan.

‘lack of evidence that plant eEF1A mimics EF-Tu in protecting L-aminoacyl tRNAs from modification’

To understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). This suggests that eEF-1A will mimic EF-Tu in protecting L-aa-tRNAs from reactive aldehydes. Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

Author response image 2.

Elongation factor enantio-selects L-aa-tRNAs through D-chiral rejection mechanism. A) Surface representation showing the cocrystal structure of EF-Tu with L-Phe-tRNAPhe. Zoomed-in image showing the binding of L-phenylalanine with side chain projected outside of binding site of EF-Tu (PDB id: 1TTT). B) Zoomed-in image of amino acid binding site of EF-Tu bound with L-phenylalanine showing the selection of amino group of amino acid through main chain atoms (PDB id: 1TTT). C) Modelling of D-phenylalanine in the amino acid binding site of EF-Tu shows severe clashes with main chain atoms of EF-Tu. Modelling of smallest chiral amino acid, alanine, in the amino acid binding site of EF-Tu shows D) no clashes with L-alanine and E) clashes with D-alanine. F) Modelling of D-alanine in the amino acid binding site of eEF-1A shows clashes with main chain atoms. (*Represents modelled molecule). G) Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and animals) showing conserved amino acid binding site residues. (Key residues are marked with red star).

Author response image 3.

Elongation factor protects L-aa-tRNAs from aldehyde modification. A) Cartoon representation showing the cocrystal structure of EF-Tu with L-Phe-tRNAPhe (PDB id: 1TTT). B) Zoomed-in image of amino acid binding site of EF-Tu bound with L-phenylalanine (PDB id: 1TTT). C) Cartoon representation showing the cryoEM structure of eEF-1A with tRNAPhe (PDB id: 5LZS). D) Image showing the overlap of EF-Tu:L-Phe-tRNAPhe crystal structure and eEF-1A:tRNAPhe cryoEM structure (r.m.s.d. of 1.44 Å over 292 Cα atoms). E) Zoomed-in image of amino acid binding site of eEF-1A with modelled L-alanine (PDB id: 5ZLS). (*Modelled) F) Overlap showing the amino acid binding site residues of EF-Tu and eEF-1A. (EF-Tu residues are marked in black and eEF-1A residues are marked in red).

‘failure to measure accumulation of toxic D-aminoacyl tRNAs or impairment of translation in plant cells lacking DTD2’

We agree that measuring the accumulation of D-aa-tRNA adducts from plant cells lacking DTD2 is important. We tried to characterise the same with dtd2 mutant plants extensively through Northern blotting as well as mass spectrometry. However, due to the lack of information about the tissue getting affected (root or shoot), identity of aa-tRNA as well as location of aa-tRNA (cytosol or organellar), we are so far unsuccessful in identifying them from plants. Efforts are still underway to identify them from plant system lacking DTD2. However, we have used a bacterial surrogate system, E. coli, as used earlier in Mazeed M. et al., Science Advances, 2021 to show the accumulation of D-aa-tRNA adducts in the absence of dtd. We could identify the accumulation of both formaldehyde and MG modified D-aa-tRNA adducts via mass spectrometry (Figure: 4). These results are now included in the revised manuscript (in line no: 190-197 and Figure: S5).

Author response image 4.

Loss of DTD results in accumulation of modified D-aminoacyl adducts on tRNAs in E. coli. Mass spectrometry analysis showing the accumulation of aldehyde modified D-Tyr-tRNATyr in A) Δdtd E. coli, B) formaldehyde and D-tyrosine treated Δdtd E. coli, and C) MG and D-tyrosine treated Δdtd E. coli. ESI-MS based tandem fragmentation analysis for unmodified and aldehyde modified D-Tyr-tRNATyr in D) Δdtd E. coli, E) and F) formaldehyde and D-tyrosine treated Δdtd E. coli, G) and H) MG and D-tyrosine treated Δdtd E. coli.

Response to Public Reviews:

We are grateful for the reviewers’ positive feedback and their comments and suggestions on this manuscript. Reviewer 1 has indicated two weaknesses and Reviewer 2 has none. We have now addressed all the concerns of the Reviewers.

Reviewer #1 (Public Review):

Summary:

This work is an extension of the authors' earlier work published in Sci Adv in 2001, wherein the authors showed that DTD2 deacylates N-ethyl-D-aminoacyl-tRNAs arising from acetaldehyde toxicity. The authors in this study, investigate the role of archaeal/plant DTD2 in the deacylation/detoxification of D-Tyr-tRNATyr modified by multiple other aldehydes and methylglyoxal (produced by plants). Importantly, the authors take their biochemical observations to plants, to show that deletion of DTD2 gene from a model plant (Arabidopsis thaliana) makes them sensitive to the aldehyde supplementation in the media especially in the presence of D-Tyr. These conclusions are further supported by the observation that the model plant shows increased tolerance to the aldehyde stress when DTD2 is overproduced from the CaMV 35S promoter. The authors propose a model for the role of DTD2 in the evolution of land plants. Finally, the authors suggest that the transgenic crops carrying DTD2 may offer a strategy for stress-tolerant crop development. Overall, the authors present a convincing story, and the data are supportive of the central theme of the story.

We are happy that reviewer found our work convincing and would like to thank the reviewer for finding our data supportive to the central theme of the manuscript.

Strengths:

Data are novel and they provide a new perspective on the role of DTD2, and propose possible use of the DTD2 lines in crop improvement.

We are happy for this positive comment on the manuscript.

Weaknesses:

(a) Data obtained from a single aminoacyl-tRNA (D-Tyr-tRNATyr) have been generalized to imply that what is relevant to this model substrate is true for all other D-aa-tRNAs (term modified aa-tRNAs has been used synonymously with the modified Tyr-tRNATyr). This is not a risk-free extrapolation. For example, the authors see that DTD2 removes modified D-Tyr from tRNATyr in a chain-length dependent manner of the modifier. Why do the authors believe that the length of the amino acid side chain will not matter in the activity of DTD2?

We thank the reviewer for bringing up this important point. As mentioned above, we wish to clarify that only half of the aminoacyl-tRNA synthetases are known to charge D-amino acids and only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009). D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007). Moreover, we have previously shown that it recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

(b) While the use of EFTu supports that the ternary complex formation by the elongation factor can resist modifications of L-Tyr-tRNATyr by the aldehydes or other agents, in the context of the present work on the role of DTD2 in plants, one would want to see the data using eEF1alpha. This is particularly relevant because there are likely to be differences in the way EFTu and eEF1alpha may protect aminoacyl-tRNAs (for example see description in the latter half of the article by Wolfson and Knight 2005, FEBS Letters 579, 3467-3472).

We thank the reviewer for bringing up this important point. As mentioned above, to understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

Reviewer #2 (Public Review):

In bacteria and mammals, metabolically generated aldehydes become toxic at high concentrations because they irreversibly modify the free amino group of various essential biological macromolecules. However, these aldehydes can be present in extremely high amounts in archaea and plants without causing major toxic side effects. This fact suggests that archaea and plants have evolved specialized mechanisms to prevent the harmful effects of aldehyde accumulation.

In this study, the authors show that the plant enzyme DTD2, originating from archaea, functions as a D-aminoacyl-tRNA deacylase. This enzyme effectively removes stable D-aminoacyl adducts from tRNAs, enabling these molecules to be recycled for translation. Furthermore, they demonstrate that DTD2 serves as a broad detoxifier for various aldehydes in vivo, extending its function beyond acetaldehyde, as previously believed. Notably, the absence of DTD2 makes plants more susceptible to reactive aldehydes, while its overexpression offers protection against them. These findings underscore the physiological significance of this enzyme.

We thank the reviewer for the positive comments the manuscript.

Response to recommendation to authors:

Reviewer #1 (Recommendations For The Authors):

I enjoyed reading the manuscript entitled, "Archaeal origin translation proofreader imparts multi aldehyde stress tolerance to land plants" from the Sankaranarayanan lab. This work is an extension of their earlier work published in Sci Adv in 2001, wherein they showed that DTD2 deacylates N-ethyl-D-aminoacyl-tRNAs arising from acetaldehyde toxicity. Now, the authors of this study (Kumar et al.) investigate the role of archaeal/plant DTD2 in the deacylation/detoxification of D-Tyr-tRNATyr modified by multiple other aldehydes and methylglyoxal (which are produced during metabolic reactions in plants). Importantly, the authors take their biochemical observations to plants, to show that deletion of DTD2 gene from a model plant (Arabidopsis thaliana) makes them sensitive to the aldehyde supplementation in the media especially in the presence of D-Tyr. These conclusions are further supported by the observation that the model plant shows increased tolerance to the aldehyde stress when DTD2 is overproduced from the CaMV 35S promoter. The authors propose a model for the role of DTD2 in the evolution of land plants. Finally, the authors suggest that the transgenic crops carrying DTD2 may offer a strategy for stress-tolerant crop development. Overall, the authors present a convincing story, and the data are supportive of the central theme of the story.

We are happy that reviewer enjoyed our manuscript and found our work convincing. We would also like to thank reviewer for finding our data supportive to the central theme of the manuscript.

I have the following observations that require the authors' attention.

1) The title of the manuscript will be more appropriate if revised to, "Archaeal origin translation proofreader, DTD2, imparts multialdehyde stress tolerance to land plants".

Both the reviewer’s suggested to change the title. We have now changed the title based on reviewer 2 suggestion.

2) Abstract (line 19): change, "physiologically abundantly produced" to "physiologically produced".

As per the reviewer’s suggestion, we have now changed it to "physiologically produced".

3) Introduction (line 50): delete, 'extremely'.

We have removed the word 'extremely' from the Introduction.

4) Line 79: change, "can be utilized" to "may be explored".

We have changed "can be utilized" to "may be explored" as suggested by the reviewers.

5) Results in general:

(a) Data obtained from a single aminoacyl-tRNA (D-Tyr-tRNATyr) have been generalized to imply that what is relevant to this model substrate is true for all other D-aa-tRNAs (term modified aa-tRNAs has been used synonymously with the modified D-Tyr-tRNATyr). This is a risky extrapolation. For example, the authors see that DTD2 removes modified D-Tyr from tRNATyr in a chain-length dependent manner of the modifier. Why do the authors believe that the length of the amino acid side chain will not matter in the activity of DTD2?

We thank the reviewer for bringing up this important point. As mentioned above, we wish to clarify that only half of the aminoacyl-tRNA synthetases are known to charge D-amino acids and only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009). D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007). Moreover, we have previously shown that it recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

(b) Interestingly, the authors do suggest (in the Materials and Methods section) that the experiments were performed with Phe-tRNAPhe as well as Ala-tRNAAla. If what is stated in Materials and Methods is correct, these data should be included to generalize the observations.

We regret for the confusing statement. We wish to clarify that L- and D-Tyr-tRNATyr were used for checking the TLC-based aldehyde modification, EF-Tu based protection assays and deacylation assays, D-Phe-tRNAPhe was used to characterise aldehyde-based modification by mass spectrometry and L-Ala-tRNAAla was used to check the modification propensity of multiple aldehydes. We used multiple aa-tRNAs to emphasize that aldehyde-based modifications are aspecific towards the identity of aa-tRNAs. All the data obtained with respective aa-tRNAs are included in manuscript.

(c) While the use of EFTu supports that the ternary complex formation by the elongation factor can resist modifications of L-Tyr-tRNATyr by the aldehydes or other agents, in the context of the present work on the role of DTD2 in plants, one would want to see the data using eEF1alpha. This is particularly relevant because there are likely to be differences in the way EFTu and eEF1alpha may protect aminoacyl-tRNAs (for example see description in the latter half of the article by Wolfson and Knight 2005, FEBS Letters 579, 3467-3472).

We thank the reviewer for bringing up this important point. As mentioned above, to understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

6) Results (line 89): Figure: 1C-G (not B-G).

As correctly pointed out by the reviewer(s), we have changed it to Figure: 1C-G.

7) Results (line 91): Figure: S1B-G (not C-G).

We wish to clarify that this is correct.

8) Line 97: change, "propionaldehyde" to "propionaldehyde (Figure: 1H)".

As per the reviewer’s suggestion, we have now changed, "propionaldehyde" to "propionaldehyde (Figure: 1H)".

9) Line 124: The statement, "DTD2 cleaved all modified D-aa-tRNAs at 50 pM to 500 nM range (Figure: 2A_D)" is not consistent with the data presented. For example, Figure 2D does not show any significant cleavage. Figure S2A-B also does not show cleavage.

We thank the reviewers for pointing this out. We have changed the sentence to “DTD2 cleaved majority of aldehyde modified D-aa-tRNAs at 50 pM to 500 nM range".

10) Line 131: Cleavage observed in Fig. S2E is inconsistent with the generalized statement on DTD1.

We wish to clarify that the minimal activity seen in Fig. S2E is inconsistent with the general trend of DTD1’s biochemical activity seen on modified D-aa-tRNAs. In addition, we have earlier shown that D-aa-tRNA fits snugly in the active site of DTD1 (Ahmad S. et al., eLife, 2013) whereas the modified D-aa-tRNA cannot bind due to the space constrains in the active site of DTD1 (Mazeed M. et al., Science Advances, 2021). Therefore, this minimal activity could be a result of technical error during this biochemical experiment and could be considered as no activity.

11) Lines 129-133: Citations of many figure panels particularly in the supplementary figures are inconsistent with generalized statements. This section requires a major rewrite or rearrangement of the figure panels (in case the statements are correct).

We thank the reviewers for bringing forth this point and we have accordingly modified the statement into “DTD2 from archaea recycled short chain aldehyde-modified D-aa-tRNA adducts as expected (Figure: 3E-G) and, like DTD2 from plants, it did not act on aldehyde-modified D-aa-tRNAs longer than three chains (Figure: 3H; S3C-D; S4G-L)”.

12) Line 142: I don't believe one can call PTH a proofreader. Its job is to recycle tRNAs from peptidyl-tRNAs.

We thank the reviewers for pointing out this very important point. This is now corrected.

13). Line 145: change, "DTD2 can exert its protection for" to "DTD2 may exert protection from".

As per the reviewer’s suggestion, we have now changed"DTD2 can exert its protection for" to "DTD2 may exert protection from".

14) Line 148: change, "a homozygous line (Figure: 3A) and checked for" to "homozygous lines (Figure: 3A) and checked them for".

As per the reviewer’s suggestion, we have now changed, "a homozygous line (Figure: 3A) and checked for" to "homozygous lines (Figure: 3A) and checked them for".

15) Line 148: Change, the sentence beginning with dtd2 as follows. Similar to earlier results30-32, dtd2-/- (dtd2 hereafter) plants were susceptible to ethanol (Figure: S4A) confirming the non-functionality DTD2 gene in dtd2 plants.

As per the reviewer’s suggestion, we have now changed the sentence accordingly.

16) Line 161: change, "linked" to "associated".

As per the reviewer’s suggestion, we have now changed "linked" to "associated".

17) Lines 173-176: It would be interesting to know how well the DTD2 OE lines do in comparison to the other known transgenic lines developed with, for example, ADH, ALDH, or AOX lines. Any ideas would help appreciate the observation with DTD2 OE lines!

We greatly appreciate the reviewer’s suggestion. We have not done any comparison experiment with any transgenic lines so far. However, it can be potentially done in further studies with DTD2 OE lines.

18) Line 194: change, "necessary" with "present".

As per the reviewer’s suggestion, we have now changed "necessary" with "present".

19) Line 210: what is meant by 'huge'? Would 'significant' sound better?

As per the reviewer’s suggestion, we have now changed "huge" with "significant".

20) Lines 239-243: This needs to be rephrased. Isn't alpha carbonyl of the carboxyl group that makes ester bond with the -CCA end of the tRNA required for DTD2 activity as well? Are you referring to the carbonyl group in the moiety that modifies the alpha-amino group? Please clarify. The cited reference (no. 64) of Atherly does not talk about it.

We regret for the confusing statement. To clarify, we were referencing to the carbonyl carbon of the modification post amino group of the amino acid in aa-tRNAs (Figure: 5). We have now included a figure (Figure: S4Q of revised manuscript) to show the comparison of the carbonyl group for the better clarity. The cited reference Atherly A. G., Nature, 1978 shows the activity of PTH on peptidyl-tRNAs and peptidyl-tRNAs possess carbonyl carbon at alpha position post amino group of amino acid in L-aa-tRNAs.

Author response image 5.

Figure showing the difference in the position of carbonyl carbon in acetonyl and acetyl modification on aa-tRNAs.

21) Line 261: thrive (not thrives).

As per the reviewer’s suggestion, we have now changed it to thrive.

22) In Fig3A: second last lane, it should be dtd-/-:: AtDTDH150A (not dtd-/-:: AtDTDH150A).

We thank the reviewers for pointing out this, we have corrected it.

23). Materials and methods: Please clarify which experiments used tRNAPhe, tRNAAla, PheRS, etc. Also, please carefully check all other details provided in this section.

As per the reviewer’s suggestion, we would like to provide a table below explaining the use of different substrates as well as enzymes in our experiments.

Author response table 1.

24) Figure legends (many places): p values higher than 0.05 (not less than) are denoted as ns.

We thank the reviewers for pointing out this. We have corrected it.

Reviewer #2 (Recommendations For The Authors):

I have only minor comments for the authors:

Title: I would replace "Archeal origin translation proofreader" with " A translation proofreader of archeal origin"

As per the reviewer’s suggestion, we have now changed the title.

Abstract: This section could benefit from some rewriting. For instance, at the outset, the initial logical connection between the first and second sentences of the abstract is somewhat unclear. At the very least, I would suggest swapping their order to enhance the narrative flow. Later in the text, the term "chiral proofreading systems" is introduced; however, it is only in a subsequent sentence that these systems are explained to be responsible for removing stable D-aminoacyl adducts from tRNA. Providing an immediate explanation of these systems would enhance the reader's comprehension. The authors switch from the past participle tense to the present tense towards the end of the text. I would recommend that they choose one tense for consistency. In the final sentence, I would suggest toning down the statement and replacing "can be used" with "could be explored." (https://www.nature.com/articles/d41586-023-02895-w). The same comment applies to the introduction, line 79.

As per the reviewer’s suggestion, we have now changed the abstract appropriately.

General note: Conventionally, the use of italics is reserved for the specific species "Arabidopsis thaliana," while the broader genus "Arabidopsis" is not italicized.

We acknowledge the reviewer for this pertinent suggestion. This is now corrected in revised version of our manuscript.

General note: I would advise the authors against employing bold characters in conjunction with colors in the figures.

We thank the reviewer for this suggestion. We have now changed it appropriately in revised version of our manuscript.

Figure 1A: I recommend including the concentrations of the various aldehydes used in the experiment within the figure legend. While this information is available in the materials and methods section, it would be beneficial to have it readily accessible when analyzing the figure.

As per the reviewer’s suggestion, we have now included the concentrations in figure legend.

Figure 1I, J: some error bars are invisible.

We thank the reviewers for pointing out this, we have corrected it.

Figure 2M: The table could be simplified by removing aldehydes for which it was not feasible to demonstrate activity. The letter "M" within the cell labeled "aldehydes" appears to be a typographical error, presumably indicating the figure panel.

As per the reviewer’s suggestion, we have now changed this appropriately.

Figure 3: For consistency with the other panels in the figure, I recommend including an additional panel to display the graph depicting the impact of MG on germination.

As per the reviewer’s suggestion, we have now changed this appropriately.

Figure 4: Considering that only one plant is presented, it would be beneficial to visualize the data distribution for the other plants used in this experiment, similar to what the authors have done in panel A of the same figure.

We thank the reviewer for bringing up this point. We wish to clarify that we have done experiment with multiple plants. However, for the sake of clarity, we have included the representative images. Moreover, we have included the quantitative data for multiple plants in Figure 3C-G.

Figure 5E: The authors may consider presenting a chronological order of events as they believe they occurred during evolution.

We thank the reviewer for the suggestion. However, it is very difficult to pinpoint the chronology of the events. Aldehydes are lethal for systems due to their hyper reactivity and systems would require immediate solutions to survive. Therefore, we think that both problem (toxic aldehyde production) and its solution (expansion of aldehyde metabolising repertoire and recruitment of archaeal DTD2) might have appeared simultaneously.

Figure 6: The model appears somewhat crowded, which may affect its clarity and ease of interpretation. The authors might also consider dividing the legend sentence into two separate sentences for better readability.

As per the reviewer’s suggestion, we have now changed this appropriately.

Line 149: I recommend explicitly stating that ethanol metabolism produces acetaldehyde. This clarification will help the general reader immediately understand why DTD2 mutant plants are sensitive to ethanol.

As per the reviewer’s suggestion, we have now changed this appropriately.

Line 289: there is a typographical error, "promotor" instead of the correct term "promoter.".

We thank the referee for pointing out this, we have now corrected it.

Figure S5: The root morphology of DTD2 OE plants appears to exhibit some differences compared to the WT, even in the absence of a high concentration of aldehydes. It would be valuable if the authors could comment on these observed differences unless they have already done so, and I may have overlooked it.

We thank the referee for pointing out this. We do see minor differences in root morphology, but they are more pronounced with aldehyde treatments. The reason for this phenotype remains elusive and we are trying to understand the role of DTD2 in root development in detail in further studies.

Some Curiosity Questions (not mandatory for manuscript acceptance):

1) Do DTD2 OE plants display an earlier flowering phenotype than wild-type Col-0?

We have not done detailed phenotyping of DTD2 OE plants. However, our preliminary observations suggest no differences in flowering pattern as compared to wild-type Col-0.

2) What is the current understanding of the endogenous regulation of DTD2?

We have not done detailed analysis to understand the endogenous regulation of DTD2.

3) Could the protective phenotype of DTD2 OE plants in the presence of aldehydes be attributed to additional functions of this enzyme beyond the removal of stable D-aminoacyl adducts from tRNAs?

Based on the available evidence regarding the biochemical activity and in vivo phenotypes of DTD2, it appears that removal of stable D-aminoacyl adducts from tRNA is key for the protective phenotype of DTD2 OE.

A Suggestion for Future Research (not required for manuscript acceptance):

The authors could explore the possibility of overexpressing DTD2 in pyruvate decarboxylase transgenic plants and assess whether this strategy enhances flood tolerance without incurring a growth penalty under normal growth conditions.

We thank the referee for this interesting suggestion for future research. We will surely keep this in mind while exploring the flood tolerance potential of DTD2 OE plants.

Author Response

The following is the authors’ response to the original reviews.

Major change:

All three of our reviewers raised the possibility that changes in movement during the time spent at the center ports could have contributed to changes in SWR rates. Analyses to address this possibility, based on the examination of trials with high and low speeds, were originally included in the supplement but we did not sufficiently highlight and explain these results. To rectify this, we have moved these results into a new main Figure 3 and now include a paragraph describing our interpretation of these results (page 9). We also include a more detailed description of the subjects’ behavior during port times – namely, that all subjects must remain quite stationary while at the reward ports in order to keep their nose in a specific position which keeps the port triggered. As a result, all subjects maintain head speeds well below our typical speed threshold for immobility while at the ports. This leads us to predict that any feedback based on periods of immobility alone (as requested by Reviewer 3) would show results very similar to our Control cohort and would not alter SWR rates seen during neurofeedback trials.

Minor changes:

(1) Reviewer 1 observed our that reported statistics appeared to be missing an interaction term showing that neurofeedback differentially affected the SWR rate/count pre- and postreward. We apologize for a lack of clarity here: we fit pre- and post-reward times with separate linear mixed effects models, so this interaction term is neither expected nor defined in our model. We have added a sentence clarifying this aspect of our LME approach in the Methods section: “Each model is designed to compare samples from all trials of the control group to samples from neurofeedback and delay trials from the neurofeedback cohort for a specific time period (for instance, pre-reward-delivery at the center ports).” Combining both times in the same model would require adding an additional hierarchical level in order to preserve the pairing of the pre- and post-reward time period for each trial, which we are concerned would complicate the formulation and interpretation of the model. However, the reviewer raises a good point that the comparison between these two time periods reveals an additional difference between the trial types: SWR rate remains relatively consistent between the pre- and post-reward periods during neurofeedback trials, while delay and control trials show a clear increase in SWR rate between the two time periods. To visualize and quantify this effect, we calculated the difference in SWR rates between the two time periods and now include this plot as Supplementary Figure 2F, which is referenced in page 8 of the main text.

(2) Reviewer 2 found our original title, “Neurofeedback training can modulate task-relevant memory replay in rats” to be misleading and suggestive of a manipulation to memory content. We are in complete agreement with the Reviewer in that our manipulation does not alter replay content, so to be more specific and accurate, we have changed our title to their suggestion “Neurofeedback training can modulate task-relevant memory replay rate in rats” accordingly.

(3) Reviewer 2 also requested that we include analyses quantifying baseline SWR rates for each of our experimental subjects. Although we initially considered reporting our results in measures of change relative to each individual animal’s baseline, we decided against this approach for several reasons.

First, it is important to clarify that we extensively train the animals on the task prior to implant, so we do not have access to a truly naïve, pre-behavior baseline SWR rate for any of our subjects. However, because the pre-implant training is conducted consistently between our neurofeedback and our control cohort, we have no reason to believe that the behavioral training prior to implant would introduce differences in SWR rate between the cohorts. Indeed, we find no difference in post-reward SWR rate (or SWR rate at the home well) when we quantify the first 250 trials of post-implant behavior for each subject (see panel A below). Note that we cannot compare the pre-reward SWR rate at this point, because it is influenced by the task structure which guarantees at least one SWR in each neurofeedback trial pre-reward.

Further, we do find that SWR rate is quite consistent over many days of task performance in the control cohort (show for the post-reward period in panel B below). This suggests that comparing the post-neurofeedback training SWR rates for the neurofeedback cohort to SWR rates throughout the training for the control cohort is not likely to be confounded by differing amounts of training experience. This is supported by our analyses in Figure 2 which show no differences in SWR rate between the two cohorts when considering pre- and post-reward times combined.

Author response image 1.

(A) SWR rate calculated during the post-reward period at the center port for the first 250 trials of postimplant behavior for each animal. Trials of all types are included (ie both neurofeedback trials and delay trials for the manipulation cohort). Groupwise comparison p=0.192. (B) Mean SWR rate during the post-reward period at the center port for each behavioral training epoch shows no systematic change over time across subjects within the control cohort.

Finally, within each cohort, we found the overall SWR rates to be quite consistent across animals. If each subject in the neurofeedback cohort had shown dramatically different SWR rates at the beginning of neurofeedback training, we would have needed to express the effect of neurofeedback training relative to baseline for each animal. However, since the range of SWR rates were highly comparable, we felt that it was more accessible, and easier to place our results within the context of the literature, by expressing our results as simple SWR rates themselves rather than measures of relative change. Within the neurofeedback cohort, comparing neurofeedback to delay trials is inherently matched for baseline SWR rate since these comparisons are made within the same animal.

(4) Finally, Reviewer 2 raises the possibility that older animals or those with cognitive deficits might respond to neurofeedback differently. We entirely agree with this possibility, and note this in our Discussion section: “Since the neurofeedback paradigm depends on the occurrence of at least a low endogenous rate of SWR occurrence, it would be important to implement neurofeedback training as a relatively early interventional strategy prior to extensive neurodegeneration, and training may take longer in aged or impaired subjects.”

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study, Yue et al. re-processed publicly available DNA methylation data (published in 2012 and 2017 from the Meissner lab) from pre- and post-implantation mouse embryos. Against the global wave of genome-wide reduction of DNA methylation occurring during pre-implantation development, they detected a slight increase (~1% on average) of DNA methylation at gene promoter regions during the transition from 8-cell to blastocyst stage. They claim that many such promoters are located in the X chromosome. Subsequently, they knocked down Dnmt3b (presumably because of its upregulation during the transition from the 8-cell to blastocyst stage) and detected the aberrant patterning of H3K27me3 in the mutant female embryos. Based on this observation, they claim that imprinted X-chromosome inactivation is impaired in the Dnmt3b-Kd pre-implantation embryos. Finally, they propose a model where such an increase of DNA methylation together with H3K27me3 regulates imprinted X-chromosome inactivation in the pre-implantation embryos. While their observation is of potential interest, the current version of the work fails to provide enough evidence to support their conclusions. Below are suggestions and comments on the manuscript.

Major issues:

(1) Sex of the embryos of the genome-wide bisulfite-sequencing data

The authors re-analyzed publicly available genome-wide DNA methylation data from the Meissner lab published in 2012 and 2017. The former used reduced representation bisulfite sequencing (RRBS) and the latter used whole-genome bisulfite sequencing (WGBS). Based mainly on the RRBS data, Yue et al. detected de novo DNA methylated promoters during the transition from 8-cell to blastocyst against the global wave of genome-wide DNA demethylation. They claim that such promoter regions are enriched at the "inactive" X chromosome. However, it would be difficult to discuss DNA methylation at inactive X-chromosomes as the RRBS data were derived from a mixture of male and female embryos. It would also be notable that the increase of DNA methylation at these promoter regions is ~1% on average. Such a slight increase in DNA methylation during pre-implantation development could also be due to the developmental variations between the embryos or between the sexes of embryos.

Thanks so much for your insightful comments. Whether de novo DNA methylation occurs in a sex-dimorphic manner would be of significance for our study. Based on your comments, we have added a reanalysis based on a publicly available single cell multi-omics sequencing (COOL-seq) data of mouse early embryos (Guo et al., 2017). The results showed that both male and female embryonic cells gain DNA methylation during the transition from the 8-cell to ICM (Figure 1—figure supplement 1C-D; Lines 112-115 in the revised manuscript).

With regards to the increase in the promoter region, many previous studies have revealed that promoter and overlapping CGI regions, especially high CpG promoters, always showed low levels of DNA methylation (Auclair et al., 2014; Borgel et al., 2010; Dahlet et al., 2020). The relatively lower basal levels make the increase seem relatively slight. Thus, we added relevant statements to clarify this information and rewritten the sentences in the revised manuscript (Lines 116-118, 125-127 in the revised manuscript).

In addition, using the single cell COOL-seq data, we also specifically reanalyzed the DNA methylation changes on the X chromosome in female embryos. The X chromosome showed a more notable increase than that on autosomes, and the female X chromosome showed a higher DNA methylation level than that of the male (Figure 3—figure supplement 2A-B; Lines 203-206 in the revised manuscript).

Thanks again for your insightful and constructive comments that significantly strengthen our evidence. We have added these results in the revised manuscript.

(2) Imprinted X-chromosome inactivation and evaluation of H3K27me3 (related to Figures 2C, D; 3F; Figure2-supplement 2 F, G; Figure3-supplement 3G)

Based on the slight change in the H3K27me3 signals in the Dnmt3b-Kd blastocysts, the authors claim that imprinted X-chromosome inactivation is impaired in the mutant embryo. It would be not easy to reach this conclusion from such a rough analysis of H3K27me3 presented in Figure 2C, D. Rigorous quantification/evaluation of the H3K27me3 signals in the Dnmt3b-Kd embryos should be considered. Additional evidence for the impairment of H3K27me3 in the mutant embryos should also be provided (expression of a subset of X-linked genes by RNA-FISH or RT-PCR etc.). Though technically challenging, high-resolution genome-wide approach such as ChIP-seq of H3K27me3 in the Dnmt3b-kd female embryos (with traceable SNPs between maternal and paternal X chromosome to distinguish inactive and active X-chromosome) could more precisely evaluate regions that lose H3K27me3 in the X-chromosome (de novo DNA methylated promoters from 8-cell to blastocyst, for example).

Thanks so much for your insightful comments that make our results more convincing. The H3K27me3 domain is a classic marker for establishment of XCI by achieving X chromosome wide heterochromatinization of transcriptional depression (Chow and Heard, 2009; Heard et al., 2004; Huynh and Lee, 2005). Thus, in the present study, we have performed immunostaining for H3K27me3 domains to evaluate the iXCI status in the blastocysts, as previously reported (Fukuda et al., 2014; Gontan et al., 2018; Inoue et al., 2010; Tan et al., 2016). Base on your comments, we have added another statistical method to quantify the establishment of iXCI, i.e. the percentage of H3K27me3-positive and -negative cells to total trophoblast cells in female blastocysts subject to Dnmt3b knockdown or not. The result also indicated that Dnmt3b knockdown led to a significant loss of H3K27me3 domains from total trophoblast cells. Similarly, new data based on statistical analyses of total trophoblast cells, has also been added in the results of Dnmt3b knockout and 5-aza-dC (Figure 3F; Figure 3—figure supplement 3D, H in the revised manuscript).

To clarify the significance and reliability of detecting H3K27me3 domains, we have added a schematic diagram depicting the process of iXCI initiation and establishment, as well as the experimental design and work flows, to make our results easier to be understood (Figure 3C in the revised manuscript).

In addition, we agree with your comments that additional evidence will benefit the conclusion. Thus, we have reanalyzed the RNA-seq and H3K27me3 CHIP-seq data in extraembryonic ectoderm (ExE) of E6.5 single embryos that underwent Dnm3a/3b knockout because preimplantation iXCI status maintains extraembryonic cells (Chen et al., 2019; Galupa and Heard, 2015; Schulz and Heard, 2013). The results showed that Dnmt knockout-induced chromosome-wide loss of DNA methylation led to a nearly complete loss of H3k27me3 on paternal X chromosome (specifically inactivated in iXCI), along with a notable transcriptional upregulation cross the chromosome. By contrast, these changes cannot be not observed on maternal X chromosome.

We have added this result in the revised manuscript (Lines 253-261; Figure 3—figure supplement 4A in the revised manuscript).

(3) Analysis of the developmental potential of Dnmt3b-kd embryos

While the authors claim that Dnmt3b-mediated de novo DNA methylation plays an important role in imprinted X-chromosome inactivation, it remains unclear whether the analysis presented in Figure 4 is derived from "female" embryos. This analysis seemed confusing as the authors claim that de novo DNA methylation in the promoter regions during the transition from 8-cell to blastocyst regulates imprinted X-chromosome inactivation, but this should not happen in the male embryos. Was the impairment of embryonic proliferation and differentiation observed in both male and female embryos? Or is this specific to the female embryos? We think that the sex of the embryos would be critical for the analysis presented in Figure 4.

Thanks so much for your constructive comments to make our results smoother and clearer. The Figure 4 mainly presents the developmental role of minor de novo methylation based on the integrated analysis of DNA methylation and gene expression dynamics from the 8-cell to ICM. Because our data indicated that both male and female embryos undergo minor de novo methylation (Figure 1—figure supplement 1C-D in the revised manuscript). This section mainly focused on genome wide and general changes, but not on sex dimorphic consequence.

To avoid the possible confusion, we have reorganized the RESULTS AND DISCUSSION section and presented this section as Figure 2 in the revised manuscript, before the chromosomal distribution analysis and subsequent detection relevant to iXCI.

Reviewer #2 (Public Review):

Summary:

Here, Yue et al. set out to determine if the low DNMT3B expression that is observed prior to de novo DNA methylation (before the blastocyst stage) has a function. Re-analyzing existing DNA methylation data from Smith et al. (2012) they find a small DNA methylation gain over a subset of promoters and gene bodies, occurring between the 8-cell and blastocyst stages, and refer to this as "minor de novo DNA methylation". They attempt to assess the relevance/functionality of this minor DNA methylation gain, and report reduced H3K27me3 in Dnmt3b knockdown (KD) trophoblast cells that normally undergo imprinted X-chromosome inactivation (iXCI) before the blastocyst stage. In addition, they assess the proliferation, differentiation, metabolic function, implantation rate, and live birth rate of Dnmt3b KD blastocysts.

Strengths:

Working with early embryos is technically demanding, making the well-designed experiments from this manuscript useful to the epigenetics community. Particularly, the DNMT3B expression and 5-mC staining at different embryonic stages.

Thanks for your positive evaluation, we have revised manuscript based on your comments, and the items need to be addressed in detail are explained in the point-by-point response to each comment.

Weaknesses:

- Throughout the manuscript, please represent DNA methylation changes as delta DNA methylation instead of fold change.

Thanks so much for your constructive comments. We have represented DNA methylation changes as “ΔDNA methylation” (Figure 2—figure supplement 1A; Figure 3—figure supplement 1A; Figure 3—figure supplement 3I in the revised manuscript).

- Detailed methods on the re-analysis of the DNA methylation data from Smith et al. 2012 are missing from the materials and methods section. Was a minimum coverage threshold used?

Thanks so much for your reminder. We have added relevant statements and provided the detail of the coverage criteria in the subsection of Bioinformatics analysis in the Materials and methods section as follows: RRBS data of mouse embryos (2-cell embryos, 4-cell embryos, 8-cell embryos, ICM, and E6.5 embryos) were downloaded from the published article by Smith et al (Smith et al., 2012) (accession number: GSE34864). The methylation level was calculated as the number of “methylated” reads (reporting as C), divided by the total number of “methylated” and “unmethylated” read, which reporting as C or T. The genomic region information was downloaded from the mm9 Repeat Masker. As described in the published article, promoters were defined as 1 kb up- and downstream of the TSS and classified into high-density CpG promoter (HCP), intermediate-density CpG promoter (ICP) and low-density CpG promoter (LCP). Only CpG sites with at least fivefold coverage were included in the methylation analysis. We have added relevant information in the revised manuscript (Lines 462-470 in the revised manuscript).

- Detailed methods on the establishment and validation of Dnmt3b KO blastocysts and 5-aza-dC treated blastocysts are missing (related to Figure 2).

Thanks so much for your detailed reminder. In the present study, we used a well-established Dnmt3b-deficient mouse model (Okano et al., 1999) to validate the role of minor de novo DNA methylation in iXCI establishment. Heterozygous Dnmt3b^+/- mice that carry one mutant locus of Dnmt3b, were obtained from the Mutant Mouse Resource & Research Centers (MMRRC, NIH). Homozygous embryos were obtained by intercrossing Dnmt3b^+/- male and female mice. Genotyping assays of collected embryos was performed by PCR using primers that were designed based on the gene targeting strategy following the MMRRC genotyping protocol (https://www.med.unc.edu/mmrrc/genotyping-protocols/mmrrc-center-protocol-29886/). We have provided the detailed methods in the revised manuscript (Lines 350-354; 391-393 in the revised manuscript). In addition, we added a schematic diagram depicting the processes of embryo collection and detection (Figure 3—figure supplement 3A in the revised manuscript).

Similarly, we have provided relevant details of 5-aza-dC supplementation in the revised manuscript (Lines 412-415 in the revised manuscript) and added a schematic diagram depicting the details of experimental design and processes (Figure 3—figure supplement 3E in the revised manuscript).

- Detailed methods on the re-analysis of the ChIPseq data from Liu et al. 2016 are missing from the materials and methods section.

Thank you for pointing this out. The bigwig files of H3K27me3 ChIP-seq data were downloaded from the published article by Liu et al (Liu et al., 2016)(accession number: GSE73952). These signal tracks were generated using the MACS2 (v2.0.10.20131216) pileup function and normalized to 1 million reads for visualization, as described in the original publication. We have added relevant information to the MATERIALS AND METHODS section in the revised manuscript (Lines 474-479 in the revised manuscript).

- Some of the data represented in bar graphs does not look convincing/significant. Maybe this data can be better represented differently, such as in box plots or violin plots, which would better represent the data.

Thanks so much for your comments that improve our result presentation, relevant results have been changed into box plots in the revised manuscript (Figure 3E; Figure 3—figure supplement 3C; Figure 3—figure supplement 3G in the revised manuscript). In addition, to strengthen our evidence, we have added alternative statistical method to quantify the establishment of iXCI, i.e. the percentage of H3K27me3-positive and -negative cells to total trophoblast cells in female blastocysts subject to Dnmt3b knockdown or not. (Figure 3F; Figure 3—figure supplement 3D, H in the revised manuscript).

- The relevance and rationale for experiments using 5-aza-dC treatment is unclear.

Thanks so much for reminding us to make our results more informative and convincing. 5-aza-dC is a well-established global DNA hypomethylating agent that efficiently inhibit the activity of all DNMTs, and thus has been frequently used to study the maintenance of DNA methylation and de novo DNA methylation (Maslov et al., 2012; Oka et al., 2005).

In our study, to validate the function of minor de novo DNA methylation in iXCI, we take advantage of 5-aza-dC-induced DNMT inhibition, which allows us, despite its inhibitory effect common to various DNMTs, to transiently treat embryos specifically during the window of minor de novo DNA methylation (from the 8-cell to blastocyst stage). We have added these statements, as well as a schematic diagram depicting the experimental design, in the revised manuscript to make our experiments more rational and easier to be understood (Lines 183-188; Figure 3—figure supplement 3E in the revised manuscript).

References

Auclair, G., Guibert, S., Bender, A. and Weber, M. (2014). Ontogeny of CpG island methylation and specificity of DNMT3 methyltransferases during embryonic development in the mouse. Genome Biol. 15, 545.

Borgel, J., Guibert, S., Li, Y., Chiba, H., Schubeler, D., Sasaki, H., Forne, T. and Weber, M. (2010). Targets and dynamics of promoter DNA methylation during early mouse development. Nat. Genet. 42, 1093-1100.

Chen, Z., Yin, Q., Inoue, A., Zhang, C. and Zhang, Y. (2019). Allelic H3K27me3 to allelic DNA methylation switch maintains noncanonical imprinting in extraembryonic cells. Sci Adv 5, eaay7246.

Chow, J. and Heard, E. (2009). X inactivation and the complexities of silencing a sex chromosome. Curr. Opin. Cell Biol. 21, 359-366.

Dahlet, T., Argueso Lleida, A., Al Adhami, H., Dumas, M., Bender, A., Ngondo, R. P., Tanguy, M., Vallet, J., Auclair, G., Bardet, A. F., et al. (2020). Genome-wide analysis in the mouse embryo reveals the importance of DNA methylation for transcription integrity. Nat Commun 11, 3153.

Fukuda, A., Tomikawa, J., Miura, T., Hata, K., Nakabayashi, K., Eggan, K., Akutsu, H. and Umezawa, A. (2014). The role of maternal-specific H3K9me3 modification in establishing imprinted X-chromosome inactivation and embryogenesis in mice. Nat Commun 5, 5464.

Galupa, R. and Heard, E. (2015). X-chromosome inactivation: new insights into cis and trans regulation. Curr. Opin. Genet. Dev. 31, 57-66.

Gontan, C., Mira-Bontenbal, H., Magaraki, A., Dupont, C., Barakat, T. S., Rentmeester, E., Demmers, J. and Gribnau, J. (2018). REX1 is the critical target of RNF12 in imprinted X chromosome inactivation in mice. Nat Commun 9, 4752.

Guo, F., Li, L., Li, J., Wu, X., Hu, B., Zhu, P., Wen, L. and Tang, F. (2017). Single-cell multi-omics sequencing of mouse early embryos and embryonic stem cells. Cell Res. 27, 967-988.

Heard, E., Chaumeil, J., Masui, O. and Okamoto, I. (2004). Mammalian X-chromosome inactivation: an epigenetics paradigm. Cold Spring Harb. Symp. Quant. Biol. 69, 89-102.

Huynh, K. D. and Lee, J. T. (2005). X-chromosome inactivation: a hypothesis linking ontogeny and phylogeny. Nat. Rev. Genet. 6, 410-418.

Inoue, K., Kohda, T., Sugimoto, M., Sado, T., Ogonuki, N., Matoba, S., Shiura, H., Ikeda, R., Mochida, K., Fujii, T., et al. (2010). Impeding Xist expression from the active X chromosome improves mouse somatic cell nuclear transfer. Science 330, 496-499.

Liu, X. Y., Wang, C. F., Liu, W. Q., Li, J. Y., Li, C., Kou, X. C., Chen, J. Y., Zhao, Y. H., Gao, H. B., Wang, H., et al. (2016). Distinct features of H3K4me3 and H3K27me3 chromatin domains in pre-implantation embryos. Nature 537, 558-562.

Maslov, A. Y., Lee, M., Gundry, M., Gravina, S., Strogonova, N., Tazearslan, C., Bendebury, A., Suh, Y. and Vijg, J. (2012). 5-aza-2'-deoxycytidine-induced genome rearrangements are mediated by DNMT1. Oncogene 31, 5172-5179.

Oka, M., Meacham, A. M., Hamazaki, T., Rodic, N., Chang, L. J. and Terada, N. (2005). De novo DNA methyltransferases Dnmt3a and Dnmt3b primarily mediate the cytotoxic effect of 5-aza-2'-deoxycytidine. Oncogene 24, 3091-3099.

Okano, M., Bell, D. W., Haber, D. A. and Li, E. (1999). DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 99, 247-257.

Schulz, E. G. and Heard, E. (2013). Role and control of X chromosome dosage in mammalian development. Curr. Opin. Genet. Dev. 23, 109-115.

Smith, Z. D., Chan, M. M., Mikkelsen, T. S., Gu, H. C., Gnirke, A., Regev, A. and Meissner, A. (2012). A unique regulatory phase of DNA methylation in the early mammalian embryo. Nature 484, 339-344.

Tan, K., An, L., Miao, K., Ren, L., Hou, Z., Tao, L., Zhang, Z., Wang, X., Xia, W., Liu, J., et al. (2016). Impaired imprinted X chromosome inactivation is responsible for the skewed sex ratio following in vitro fertilization. Proc. Natl. Acad. Sci. U. S. A. 113, 3197-3202.

Reviewer #1 (Recommendations For The Authors):

Title

It would be hard to understand what "co"-regulates means. Does this mean DNA methylation and H3K27me3 co-regulate imprinted X- X-chromosome inactivation? If so, the title can be reworded.

Thanks for your insightful comments, the title has been corrected into “A wave of minor de novo DNA methylation initiates in mouse 8-cell embryos and co-regulates imprinted X- chromosome inactivation with H3K27me3” (Line 2 in the revised manuscript).

Text

(1) As DNA methylation analysis is a primary part of this study, how they processed DNA methylation data can be added to the "Bioinformatics analysis" in the MATERIALS AND METHODS section.

Thanks for your kind reminder. We have added relevant information in the Materials and methods section in the revised manuscript (Lines 462-474 in the revised manuscript).

(2) It seems that recent literature has not been cited in the manuscript. Specifically, none of the papers after 2018 were cited. Recent relevant papers should also be cited throughout the manuscript.

Thanks so much for your reminder. We have added more recent literature to update the relevant information, such as the evidence supporting the causal role between DNA methylation and XCI (Lines 225-228, 264-265 in the revised manuscript); the concurrent enrichment of DNA methylation and H3K27me3 in genes subject to XCI (Lines 301-303 in the revised manuscript); the dominant role of de novo methylation in X chromosome (Lines 253-256 in the revised manuscript), etc.

(3) Line 56: The first report that describes the dynamics of DNMT3B expression in pre-implantation embryonic development (Hirasawa et al., 2007) is missing. This paper should be cited.

Sorry for our carelessness, we have added relevant references and rewritten the sentence in the revised manuscript (Lines 56-57 in the revised manuscript). I think you meant the report by Hirasawa et al in 2008, in which presented expression and subcellular localization of Dnmt3a and Dnmt3b in mouse oocytes and preimplantation embryos.

(4) Line 98: It would be good to mention that the data were derived from reduced representation bisulfite sequencing as the authors used whole-genome bisulfite sequencing data from the same research group as well.

Thanks for your kind reminder. As you have suggested, we have added the description in the revised manuscript to emphasize that these data were derived from reduced representation bisulfite sequencing, while another data were derived from whole-genome bisulfite sequencing, respectively. (Lines 98-99, 111 in the revised manuscript).

(5) Line 101: We first... "the preferential target of DNMT3B (Auclair et al., 2014; Borgel et al., 2010)". More recent literature (Baubec et al., 2016, Duymich et al., 2016, for example) showed that the preferential target of DNMT3B is not a promoter but a gene body. This sentence should be reworded.

Thanks so much for your detailed reminder. As you have pointed out, “preferential target” seems to be an inaccurate statement. Besides of promoters, gene bodies and other elements also undergo de novo DNA methylation (Auclair et al., 2014; Dahlet et al., 2020; Duymich et al., 2016).

We have rewritten the sentence as follows in the revised manuscript: “Promoter regions are important target sites of DNMT3B (Choi et al., 2011). The acquisition of DNA methylation in promoters, especially in intermediate and low CpG promoters, during implantation is largely dependent on DNMT3B and plays an important role in regulating developmental genes (Auclair et al., 2014; Borgel et al., 2010; Dahlet et al., 2020). Thus, among genomic regions that may undergo de novo DNA methylation, we initially focused our analysis on DNA methylation dynamics of promoters...” (Lines 100-106 in the revised manuscript)

(6) Lines 108-109: It would be good to mention that these data were derived from whole-genome bisulfite sequencing.

Thanks for your kind reminder. As aforementioned, we have added a description in the revised manuscript to distinguish between data derived from reduced representation bisulfite sequencing and whole-genome bisulfite sequencing (Lines 98-99, 111 in the revised manuscript).

(7) Line 141: rXCI should be defined.

Thanks for your kind reminder. We have added full descriptions and more necessary information about iXCI and rXCI, to make our statements clearer and easier to be understood (Lines 210-213 in the revised manuscript). In addition, we carefully checked the relevant descriptions throughout the manuscript, and each abbreviation (such as “ICM”) has been defined at its first occurrence. Additionally, we have replaced abbreviations that appears only once in the manuscript with their full terms (Lines 122, 212 in the revised manuscript).

(8) Lines 145-149: The role of DNA methylation for imprinted X-inactivation has already been reported (Chiba et al., 2008). The relevant sentences should be reworded.

Thanks so much for reminding us the important earlier literature that explores the relationship between DNA methylation and XCI. However, the primary aim and hypothesis of the study by Chiba et al. are different from those of our study. Chiba et al focused on whether DNA methylation is the imprinting mark responsible for monoallelic expression of Xist (the initiation event of iXCI), while our study focused on the role of DNA methylation in achieving X chromosomal heterochromatinization (the late event of iXCI).

In detail, the study by Chiba et al. mainly focused on exploring why Xist is specifically expressed from paternal allele and iXCI occurs specifically on the paternal X chromosome in mouse preimplantation embryos. Because Previous studies have suggested that genomic imprinting of Xist is established during oogenesis (Oikawa et al., 2014; Tada et al., 2000), Chiba et al. wanted to test whether the DNA methylation imprinting established during oogenesis is responsible for the monoallelic expression of Xist in preimpantaiton embryos. Analyses of DNA methyltransferase maternal knockout embryos revealed that oocyte DNA methylation is dispensable for Xist imprinting (Chiba et al., 2008). Follow-up study by Inoue et al. identified a broad H3K27me3 enrichment within the Xist 5’region established during oocyte growth and persists through preimplantation development, as the imprinting mark of Xist (Inoue et al., 2017). These series of studies are very important and allows us to understand the mechanism underlying paternal allele-specific iXCI in mouse preimplantation embryos and extraembryonic tissues.

However, the hypothesis is different in our study. Based on the finding of minor de novo DNA methylation and its preferential distribution on the X chromosome, we have speculated that the minor de novo methylation, which occurs from the 8-cell to blastocyst stage, may participate in achieving X chromosomal heterochromatinization. Although DNA methylation is essential for maintaining X chromosome-wide transcriptional silence of rXCI, its role in iXCI remains controversial and it is even plausibly thought that DNA methylation is not required for achieving iXCI because preimplantation embryos undergo global and massive DNA demethylation.

We have reorganized this paragraph, relevant statements have been added to make the background and discussion clearer and easier to be understood. (Lines 217-234 in the revised manuscript)

(9) Lines 164-165: Information regarding Dnmt3b KO is missing. Did the authors generate an original KO line or use an already published one? It should be explicitly stated.

Thank you so much for your kind reminder. The Dnmt3b heterozygous mice were obtained from the Mutant Mouse Resource & Research Centers (MMRRC), and Dnmt3b knockout (KO) embryos were generated by mating Dnmt3b heterozygous females with heterozygous males. The genotyping of Dnmt3b KO embryos was performed by PCR following the MMRRC genotyping protocol (https://www.med.unc.edu/mmrrc/genotyping-protocols/mmrrc-center-protocol-29886/). The relevant information has been added to the MATERIALS AND METHODS section in the revised manuscript (Lines 350-354; 391-393 in the revised manuscript).

(10) Line 165: chemical-induced inhibition of DNMT3B. As 5-aza-dC also blocks DNMT3A and DNMT1, this sentence should be reworded.

Thank you for your valuable comments. 5-aza-dC is a well-established global DNA hypomethylating agent that efficiently inhibit the activity of all DNMTs, and has been frequently used to study the maintenance of DNA methylation and de novo DNA methylation (Maslov et al., 2012; Oka et al., 2005). Thus, despite its inhibitory effect common to various DNMTs, chemical-induced inhibition of DNMTs has the advantage of allowing us to transiently treated embryos specifically during the window of minor de novo DNA methylation (the 8-cell to blastocyst stage). We have rewritten the relevant sentences in the revised manuscript (Lines 183-188 in the revised manuscript).

(11) Lines 171-174: "The role of de novo methylation in iXCI...". This possibility was already tested in the previous study from the Sasaki lab (Chiba et al., 2008).

As mentioned above, the primary aim and hypothesis of the study by Chiba et al. are different from those of our study. Chiba et al. mainly focused on exploring why Xist is specifically expressed from paternal allele and iXCI occurs specifically on the paternal X chromosome in mouse preimplantation embryos, so they tested whether the DNA methylation imprinting established during oogenesis is responsible for this monoallelic expression of Xist in preimplantation embryos (the initiation event of iXCI).

By contrast, based on the finding of minor de novo DNA methylation and its preferential distribution on X chromosome, our study has speculated that the minor de novo DNA methylation, which occurs from the 8-cell to blastocyst stage, may participate in achieving X chromosomal heterochromatinization (the late event of iXCI).

Thanks so much for reminding us this important literature, to make our discussion more informative. We have reorganized this paragraph by rewriting or adding relevant statements to make the background and discussion clearer and easier to be understood (Lines 217-231 in the revised manuscript). In addition, to avoid repeated statement and make our discussion more concise, we have removed the similar sentences at the end of this paragraph.

(12) Lines 198-200: "Given DNA methylation...". These citations mention a general relationship between DNA methylation and H3K27me3 in cells in culture. As I believe the authors focus on X-chromosome inactivation in the female embryos, more relevant papers that discuss the order of the events for the establishment of H3K27me3 and DNA methylation in the inactive X-chromosome can be cited.

Thanks so much for your comment to improve our discussion. It has been thought that during the late phase of rXCI in fully differentiated cells, gene silencing is achieved by PRC2 complex-induced H3K27me3, and then is further stably maintained by the redundant action of multiple layers of epigenetic modifications, including DNA methylation, to reach the maximum level of chromatin compaction (Chow and Heard, 2009; Heard et al., 2004; Pintacuda and Cerase, 2015). In line with this, a recent multifaceted analysis showed that DNA methylation and H3K27me3 are concurrently enriched in genes subject to XCI (Balaton and Brown, 2021). We have added these statements in the revised manuscript (Lines 295-303 in the revised manuscript).

(13) Line 241: As 5-aza-dC blocks both de novo and maintenance DNA methylation, this sentence should be reworded.

Thank you for your kind reminder. As you have mentioned above, 5-aza-dC is a well-established global DNA hypomethylating agent that efficiently inhibit the activity of all DNMTs, and has been frequently used to study the maintenance of DNA methylation and de novo DNA methylation (Maslov et al., 2012; Oka et al., 2005). Thus, despite its inhibitory effect common to various DNMTs, chemical-induced inhibition of DNMTs has the advantage of allowing us to transiently treated embryos specifically during the window of minor de novo DNA methylation (the 8-cell to blastocyst stage). We have rewritten the relevant sentences in the revised manuscript (Lines 183-188 in the revised manuscript).

Figures

(1) Figure 1C, D: Do the rows in C and D show the corresponding genes?

Figure 1C and D represent the DNA methylation changes of promoters (C) and gene bodies (D) respectively, during the transition from the 8-cell to blastocyst stage. Two data were analyzed independently, and rows did not show the corresponding genes. Since we have focused on the minor de novo methylation in promoter regions, to avoid confusion, the results of the gene body have been removed from the revised manuscript.

(2) Figure 1G: Yy2 promoter gained DNA methylation during the transition from 8-cell to the blastocyst stage. Is this a representative locus for the de novo methylated promoters that are shown in Figure 1F where an increase of DNA methylation is about ~1% on average? Another representative locus could be shown instead of this gene promoter.

Thanks so much for you detailed reminder. The inconsistency between the global methylation change and bisulfite sequencing analysis of Yy2, may be due to the details of methodologies, such C-T conversion efficiency, the number of picked colonies, etc. Since we have confirmed the presence of minor de novo DNA methylation using different publicly available data, to avoid ambiguity, we have removed this result in revised manuscript.

(3) Figures 2C and 3A: It would be helpful to mention what the arrowheads mean.

Thanks so much for you detailed reminder. In Figure 2C, the arrowhead indicates the H3k27me3 domain and the blank arrowhead indicates the blastomere without the H3k27me3 domain. In Figure 3A, the arrowhead indicates Xist RNA domain and the blank arrowhead indicates the blastomere without Xist RNA domain. We have added the information in the revised manuscript (Lines 736-738, 747-749 in the revised manuscript).

(4) Figure 3-figure supplement 2B: It would be hard to see whether H3K27me3 is enriched at the promoter regions of presented genes. It would be helpful to show the values for the Y-axis as in panel A.

Thanks for your helpful reminder. We have added the scales to the figure to improve the result presentation (Figure 4—figure supplement 2B in the revised manuscript).

(5) Figure 4-figure supplement 2: 5-aza-dC blocks not only the activity of DNMT3B but also DNMT1, and DNMT3A (all these DNMTs are expressed during pre-implantation embryos, see Hirasawa et al., 2007). This part can be omitted from the manuscript.

Thanks for your insightful comments. As you have mentioned above, the relevance and rationale for experiments using 5-aza-dC treatment should be clarified. 5-aza-dC is a well-established global DNA hypomethylating agent that efficiently inhibit the activity of all DNMTs, and thus has been frequently used to study the maintenance of DNA methylation and de novo DNA methylation (Maslov et al., 2012; Oka et al., 2005).

In our study, to validate the function of minor de novo DNA methylation in iXCI and blastocyst development, we take advantage of 5-aza-dC-induced DNMT inhibition, which allows us to transiently treated embryos specifically during the window of minor de novo DNA methylation (the 8-cell to blastocyst stage), despite its non-specificity to various DNMTs.

Based on these considerations, we hope to retain this result, and wish to get your understanding.

We have added these statements in the revised manuscript to make our experiments more rational and easier to be understood (Lines 183-188 in the revised manuscript) and added a schematic diagram depicting the experimental design (Figure 3—figure supplement 3E in the revised manuscript).

Reviewer #2 (Recommendations For The Authors):

Recommendations/concerns in the text:

- Line 106, it is unclear what is meant by "in line with this"? Gene body DNA methylation is a characteristic of active transcription, so why would a gain in DNA methylation at promoters be in line with a gain in DNA methylation over gene bodies?

Thank you so much for your comments that pointed out our ambiguous statement. We meant both the promoter and gene body regions, albeit accounting for small proportions, gain DNA methylation during the transition from the 8-cell to blastocyst stage. Based on the comment by Reviewer#1, since we have focused on the minor de novo methylation in promoter regions, to avoid confusion, the results of the gene body have been removed from the revised manuscript.

- Line 111 & 114, can 6% DNA methylation really be considered "relatively hypermethylated" compared to 3% DNA methylation that is referred to as "more hypomethylated"?

We apologize for our unclear and ambiguous statements. Here we focused on the promoter regions. Many previous studies have revealed that compared with gene bodies and other genome elements, promoter and overlapping CGI regions, especially high CpG promoters, always showed low levels of DNA methylation. We have added relevant statements to clarify this information, and rewritten the sentences in the revised manuscript (Lines 100-106, 116-118, 121, 124 in the revised manuscript).

- Line 124, there are a number of processes identified, why only mention one in the text? Suggest changing writing to be more accurate, indicating what was included for the GO analysis and using the words "enriched for ... processes". Saying it may be linked to a process is an overstatement and not supported by further experiments/data.

Thank you so much for your detailed comments that make our results more informative. We have checked the relevant description and addressed your suggestions as follows: By performing gene ontology enrichment analysis of genes that undergo minor or major de novo DNA methylation respectively, we noticed that besides of many important basic processes common to two waves of de novo DNA methylation, genes subject to minor de novo DNA methylation were enriched in processes such as organic substance transport, chromosome organization, and cell fate specification (Lines 129-134 in the revised manuscript).

- Lines 149 - 152: sentence/message unclear.

We apologize for the ambiguous description. We have corrected the relevant descriptions as follows: To identify the biological function of minor de novo DNA methylation in iXCI, we knocked down Dnmt3b in preimplantation embryos by microinjecting Dnmt3b siRNA into zygotes (Lines 234-236 in the revised manuscript).

- Lines 162-164: the data in Figure 2C/D does not support this statement, as it does not show H3K27me3 loss specifically at the inactive X-chromosome.

Thanks so much for your insightful comments. Despite the global enrichment of H3K27me3, the H3K27me3 domain detected by immunostaining is a classic marker for establishment of XCI by achieving X chromosome wide heterochromatinization of transcriptional depression (Chow and Heard, 2009; Heard et al., 2004; Huynh and Lee, 2005). Thus, we have used immunostaining for H3K27me3 domains to evaluate the iXCI establishment in the blastocysts, as previously reported (Fukuda et al., 2014; Gontan et al., 2018; Inoue et al., 2010; Tan et al., 2016). To make our results more convincing, we have added another statistical method to quantify the establishment of iXCI, i.e., the percentage of H3K27me3-positive and -negative trophoblast cells to total trophoblast cells in female blastocysts subject to Dnmt3b knockdown or not.

In addition, we have added a schematic diagram depicting the process of iXCI initiation and establishment, as well as the experimental design and work flows, to make the result easier to be understood.

In addition, we agree with your comments that additional evidence will benefit the conclusion. To strengthen the evidence, and test whether DNA methylation loss leads to a prolonged effect on iXCI, we have reanalyzed the RNA-seq and H3K27me3 CHIP-seq data in extraembryonic ectoderm (ExE) of E6.5 single embryos that underwent Dnm3a/3b knockout because preimplantation iXCI status maintains extraembryonic cells (Chen et al., 2019; Galupa and Heard, 2015; Schulz and Heard, 2013). The results showed that chromosome-wide loss of DNA methylation led to a nearly complete loss of H3k27me3 on paternal (specifically inactivated in iXCI), along with a notable transcriptional upregulation cross the chromosome. By contrast, these changes cannot be not observed on maternal X chromosome. (Lines 253-261; Figure 3—figure supplement 4A in the revised manuscript)

- Lines 169-174: sentence/message unclear.

As aforementioned, we have reorganized this paragraph by rewriting or adding relevant statements relevant to the DNA methylation and XCI, to make the background and discussion clearer and easier to be understood (Lines 217-234 in the revised manuscript). In addition, to avoid repeated statement and make our discussion more concise, we have removed the similar sentences at the end of this paragraph.

- Lines 177-179: this statement is too bold. The data does not support "direct evidence".

Thank you for your detailed reminder. We have rewritten the sentence to avoid confusion and overstatement (Lines 262-268 in the revised manuscript).

- Line 198: these are not all enzymes, but could be referred to as chromatin modifiers.

We apologize for the ambiguous description. As you suggested, we have corrected “enzymes” to “chromatin modifiers” (Lines 284, 287 in the revised manuscript).

- Line 199: this statement is not correct in all contexts. There are many studies showing antagonism between DNA methylation and H3K27me3.

Thanks so much for you careful reviewing. As you have pointed out, the relationship of DNA methylation and H3K27me3 are divergent and largely controversial among studies. Under certain circumstances, DNA methylation shows antagonistic effect to H3K27me3 at promoters, via excluding the binding of PRC2 (the main complex responsible for H3K27me3 deposition) components to their targets (Bartke et al., 2010; Jermann et al., 2014), while other studies have presented alternative evidence that PRC2 (the main complex responsible for H3K27me3 deposition) and DNA methylation cooperate to achieve silencing (Hagarman et al., 2013; Vire et al., 2006). Thus, it has been thought that the relationship between DNA and methylation and histone modifications is complex, possibly in a cell-type and/or genomic region-specific manner. Both antagonism and coordination can be observed in different regulatory elements in mouse ES cells (King et al., 2016).

We apologize our incomplete statement because we mainly focused on their synergistic relationship. We have refined this section by rewriting relevant sentences and adding necessary statements (Lines 288-303 in the revised manuscript).

- Lines 228-230: the developmental significance of DNA methylation homeostasis is already well-established. Please reference relevant papers showing this here.

Thank you for this helpful suggestion. We have reorganized this section. Relevant references that highlight the developmental significance of DNA methylation homeostasis have added. The sentence has been rewritten and moved to the end of this paragraph, in the revised manuscript (Lines 159-161 in the revised manuscript).

- Line 238: an explanation/rationale for looking at energy metabolism is lacking.

Thank you for your comments to make our results earlier to be understood. The detection of energy metabolism is mainly based on the integrated analysis of DNA methylation and gene expression from the 8-cell embryos to ICM, to test the potential short-and long-term developmental consequences of minor de novo DNA methylation. Bioinformatic analysis suggested that many basic processes, such as cell differentiation, cell cycle and metabolic regulation, may be regulated by minor de novo DNA methylation. Among the enriched genes, several are related energy metabolism. In addition, because energy metabolism is crucial for supporting embryo differentiation and development, and oxidative phosphorylation (OXPHOS) metabolism is highly activated during the blastocyst stage (Zhao et al., 2021), we next examined the energy metabolism, particularly OXPHOS activity, of Dnmt3b-KD embryos. We have refined the section by rewritten relevant sentence and added necessary statements (Lines 175-179 in the revised manuscript).

- Lines 246-248: Looking at the data in Figure 2 figure supplement 2, this statement is simply not true with regards to DNMT3B protein, and also global DNA methylation level is reduced in the Dnmt3b KD blastocyst, which could lead to defective major de novo DNA methylation.

Thanks for your careful reviewing, we have rewritten the sentence to make our statement more accurate and avoid overstatement (Lines 188-190 in the revised manuscript).

Recommendations/concerns relating to figures:

Figure 1:

- Of all genic promoters, how many were included in the analysis (contained sufficient coverage)? What cut-off/thresholds were used to consider DNA methylation gain at a promoter?

Thanks for your comments. In total, 11662 promoters were analyzed. Given that promoter methylation is generally at low level, particularly at the 8-cell stage at which minor de novo methylation is just initiated. The relatively lower basal levels make the increase before the blastocyst, seem considerably slight. To capture the slight changes, we have used the relaxed threshold based on ΔDNA methylation. Only CpG sites with at least fivefold coverage were included in the methylation analysis based on data from Smith et al. (Smith et al., 2012)., ΔDNA methylation greater or less than 0 was defined as gain or loss of DNA methylation. We have added this information in the revised manuscript (Lines 462-470 in the revised manuscript).

- Does an average methylation level of 0.02 represent 2% DNA methylation? Presuming yes, is the average 1.5% DNA methylation gain at promoters real? And meaningful? Especially compared to the gain in DNA methylation that takes place between ICM and E6.5 (Figure 1 Figure Supplement 1 D)

As you have pointed out, an average methylation level of 0.02 represent 2% DNA methylation. As aforementioned, promoters exhibited an average of 1.5% DNA methylation gain during the transition from 8-cell stage to ICM. The slight increase may be mainly due to the relatively lower basal levels. As you expected, compared with the comprehensive de novo DNA methylation during implantation, preimplantation de novo methylation occurs more slightly, at a small proportion of promoter regions, so designated it as minor de novo DNA methylation. It should be also mentioned that a proportion of these promoters continue to gain massive DNA methylation during implantation. We have refined the relevant sentences to provide more detailed information of our results (Lines 125-127 in the revised manuscript).

- Why is there a focus on promoters (which are not the preferential target of DNMT3B)?

Thanks so much for your detailed reminder. As you have pointed out, “preferential target” seems to be an inaccurate statement. besides of promoters, gene bodies and other elements also undergo de novo DNA methylation (Auclair et al., 2014; Dahlet et al., 2020; Duymich et al., 2016). We have focused on the promoter regions based on the following considerations: (1) Promoter regions are important target sites of DNMT3B (Choi et al., 2011); (2) The acquisition of DNA methylation in promoters, especially in intermediate and low CpG promoters, during implantation is largely dependent on DNMT3B and plays an important role in regulating developmental genes (Auclair et al., 2014; Borgel et al., 2010; Dahlet et al., 2020). We have rewritten the relevant sentence in the revised manuscript (Lines 100-106 in the revised manuscript).

- Figure 1H shows that promoters that gain DNA methylation during the "minor de novo DNA methylation" continue to gain DNA methylation during "de novo DNA methylation". Is the ~1.5% DNA methylation gain just the slow start of the main de novo DNA methylation wave?

Your comments is very helpful to improve the description of our results. In the present study, our analysis indicated that a small proportion of promoters initially gain methylation during the transition from the 8-cell to ICM. The finding challenges current knowledge: (1) de novo DNA methylation occurs during implantation, by which globally hypomethylated blastocysts acquire genome-wide DNA methylation (Borgel et al., 2010; Dahlet et al., 2020; Smith et al., 2012); (2) during preimplantation development, embryos undergo massive and global DNA demethylation.

To distinguish the current knowledge of the timing and dynamics of DNA methylation during the early development, we have designated our finding during the transition from the 8-cell to blastocyst stage, as minor de novo DNA methylation.

We agree with your notion that among the promoters undergoing minor de novo methylation, most of them continue to gain DNA methylation during implantation, as revealed in Fig. 1F. We have added refine the relevant statement in revised manuscript (Lines 125-127 in the revised manuscript).

- The GO analysis performed for Figure 1H, what was used as input? Promoters of genes that gain DNA methylation as identified in 1C?

Thank you for your comments. For the GO analysis shown in Figure 1H, we used genes with promoter regions that gained or lost DNA methylation during the transition from the 8-cell to ICM respectively (identified in Figure 1C, as input), respectively. This information has been clarified in the revised manuscript to ensure accuracy (Lines 129-134 in the revised manuscript).

- Figure 1 figure supplement 1, is there only a fold change as threshold or also a calculated significance (eg. p-value/FDR)?

Thanks for your valuable comments. Considering the relatively low DNA methylation levels at promoter regions, and the slightly changes occurring during the preimplantation embryo development, we used the relaxed threshold based on ΔDNA methylation. Only CpG sites with at least fivefold coverage were included in the methylation analysis based on data from Smith et al. (Smith et al., 2012), ΔDNA methylation greater or less than 0 was defined as gain or loss of DNA methylation. We have replaced relevant figures and added this information in the revised manuscript (Figure 1—figure supplement 1D-E; Lines 125-127 in the revised manuscript).

- To confirm DNMT3B is responsible for the DNA methylation gain: DNMT3B KD/KO followed by promoter DNA methylation analysis to confirm the promoters that gain DNA methylation between 8 cell and ICM don't gain DNA methylation in the absence of DNMT3B.

We agree with your comments that additional evidence will benefit the conclusion. To strengthen the evidence, we have reanalyzed the RNA-seq and H3K27me3 CHIP-seq data in extraembryonic ectoderm (ExE) of E6.5 single embryos that underwent Dnm3a/3b knockout because preimplantation iXCI status maintains extraembryonic cells (Chen et al., 2019; Galupa and Heard, 2015; Schulz and Heard, 2013). The results showed that chromosome-wide loss of DNA methylation led to a nearly complete loss of H3k27me3 on paternal (specifically inactivated in iXCI), which showed a notable transcriptional upregulation cross the chromosome. By contrast, these changes cannot be not observed on maternal X chromosome. We have added this result in the revised manuscript (Lines 253-261; Figure 3—figure supplement 4A in the revised manuscript).

Figure 2:

- Figure 2A: label missing for what the numbers on the y-axis represent.

Thank you for pointing this out. We apologize for the oversight. We have added the label of y-axis in Figure 2A to clarify what the numbers represent, making it easier to be understood (Figure 3A in the revised manuscript).

- Figure 2B: y-axis is % of methylated promoters compared to all promoters?

Thank you for your suggestion. The y-axis in Figure 2B indeed represents the percentage of de novo methylated promoters relative to all promoters. As you have suggested, we have clarified this labeling in the revised manuscript (Figure 3B in the revised manuscript).

- What is the delta DNA methylation gain specifically for X-linked promoters?

Thanks so much for your reminder. To provide more convincing evidence. We have reanalyzed a single cell COOL-seq data, we also specifically reanalyzed the DNA methylation changes on the X chromosomal promoter in female embryos. The X chromosome showed a more notable increase in the de novo methylated promoters than that on autosomes, and the female X chromosome showed higher DNA methylation levels than that of the male (Figure 3—figure supplement 2A-B; Lines 203-206 in the revised manuscript).

- Figure 2C: include representative images of separate channels to better see the signal of CDX2 and H3K27me3. Quantification would be better represented with box plots.

Thank you for your helpful suggestions. We have added separate channel images in the revised manuscript. Additionally, we have adjusted the quantification to be represented as box plots, as you have suggested, to improve the accuracy and interpretability of the data presentation (Figure 3D-F in the revised manuscript).

- Figure 2C: Does the H3K27me3 signal overlap with the location of the inactive X-chromosome (is there maybe denser DAPI or do IF combined with Xist RNA-FISH)?

Thanks so much for your insightful comments. Despite the global enrichment of H3K27me3, the H3K27me3 domain detected by immunostaining is a classic marker for establishment of XCI by achieving X chromosome wide heterochromatinization of transcriptional depression (Chow and Heard, 2009; Heard et al., 2004; Huynh and Lee, 2005). Thus, we have used immunostaining for H3K27me3 domains to evaluate the iXCI establishment in the blastocysts, as previously reported (Fukuda et al., 2014; Gontan et al., 2018; Inoue et al., 2010; Tan et al., 2016). We have taken effort to perform co-staining of H3K27me3 IF and Xist FISH, but was hindered by the technical challenge, we wish to get your understanding. However, as we aforementioned, H3K27me3 is a well-accepted maker to clarify the XCI status.

In addition, to make our results more convincing, we have added an alternative statistical method to quantify the establishment of iXCI, i.e., the percentage of H3K27me3-positive and -negative trophoblast cells to total trophoblast cells in female blastocysts subject to Dnmt3b knockdown or not (Figure 3F; Lines 243-244 in the revised manuscript)

- Figure 2 figure supplement 2A: relative expression of Dnmt3b?

Thanks for your detailed reminder. The data represent the relative expression level of Dnmt3b, as noted in the original figure legend. Based on your comments, we have added the gene name in the label of the Y-axis. Similarly, the protein name has been also added to make the results more informative (Figure 2 figure supplement 2A, C, E in the revised manuscript).

- Figure 2 figure supplement 2B/C: in the text, line 153, it is stated that "Dnmt3b mRNA and protein levels were significantly reduced in morulae, but not in blastocysts compared to those of negative control (NC) group". These figures do not support that statement. The IF images show a loss of DNMT3B in the Dnmt3b KD blastocysts. The IF quantification seems to have fewer datapoints for the blastocyst, and looking at the bar graphs, there seems to be a trend towards reduced DNMT3B in both the morula and blastocyst, which would also explain the reduction in DNA methylation in both stages as shown in Figure 2 figure supplement 2D/E.

Thanks so much for your careful reviewing that makes our statements more accurate. We have rewritten the sentence in the revised manuscript as follows: Dnmt3b mRNA and protein levels were significantly reduced in morulae, and tended to be lower in blastocysts compared to those of the negative control (NC) group. In addition, we have removed “transient” from the original statement “The transient inhibition of Dnmt3b” (Lines 168-170 in the revised manuscript).

- Figure 2 figure supplement 2F/G: include representative IF images with separation of all channels and the merged image.

Thank you for your suggestion. We have added the representative immunofluorescence (IF) images with separate channels and merged image in the revised manuscript (Figure 3—figure supplement 3B, F in the revised manuscript).

- Figure 2 figure supplement 2H: Instead of showing log2FC in methylation levels, delta methylation would be more informative. Are these genes already inactivated at the 8-cell stage? Or are they active and become inactivated by the gain in DNA methylation? Doing qPCR for these genes, or looking at published RNAseq data would be informative. What happens to the expression of these genes in the Dnmt3b KD?

Thanks for your suggestions. We have represented DNA methylation changes as “ΔDNA methylation”. During mouse preimplantation development, iXCI is initiated in earlier cleavage female embryos dependent on Xist upregulation around 4-8-cell stage, and then Xist specifically coats paternal X chromosome and finally leads to chromosome-wide silencing via heterochromatinization in early blastocysts. Thus, these non-escaping genes, which are subject to XCI, would not be inactivated at 8-cell stage

Author response image 1.

The processes of iXCI initiation and establishment (left panel), and dynamics of total expression levels of X chromosome in male and female preimplantation embryos (right panel, note that X-dosage is balanced between sexes until the early blastocyst stage).

As you expected, most of these representative non-escaping is downregulated upon the transition of 8-cell to blastocyst stage, consistent with their gain of DNA methylation. Additionally, since preimplantation iXCI status maintains extraembryonic cells (Galupa and Heard, 2015; Schulz and Heard, 2013), we further reanalyzed the published RNA-seq data in extraembryonic ectoderm (ExE) of E6.5 single embryos that underwent DNA methyltransferase knockout (Chen et al., 2019). The results showed that chromosome-wide loss of DNA methylation led to a chromosome-wide transcriptional upregulation, including the locus of these non-escaping genes, on paternal X chromosome. We have added this result in the revised manuscript (Figure 3—figure supplement 3J; Figure 3—figure supplement 4A-B; Lines 253-261 in the revised manuscript).

Figure 3:

- Figure 3 figure supplement 1: representative IF image missing.

Thanks for your kind reminder. We have added the representative IF images in the revised manuscript to provide a clearer illustration of the data (Figure 4—figure supplement 1A in the revised manuscript).

- Figure 3 figure supplement 2B: scales are missing for the H3K27me3 ChIP-seq data (are the 8-cell and ICM tracks set to the same scale?). It looks like the ICM track is cut off at the top (peaks not fully displayed) and the data looks very sparse. A more informative analysis would be to do peak calling over promoters and compare 8-cell with ICM.

Thanks for your detailed reminder. We apologize for the missing of scale bars in the H3K27me3 ChIP-seq data. The 8-cell and ICM tracks were set to the same scale, and we have now added scales to the figure in the revised manuscript to improve the result presentation. As you have speculated, the visual effect of the flatted peak is not caused by track cutting off, but rather by zooming into a specific region in the extended IGV files.

These results are based on the reanalysis of publicly available data of pooled embryos, which just provided suggestive but not direct evidence to support the role of DNA methylation in promoting X-linked H3K27me3 enrichment in iXCI.

To provide more convincing evidence. we have reanalyzed the RNA-seq and H3K27me3 CHIP-seq data in extraembryonic ectoderm (ExE) of E6.5 female embryos that underwent Dnmt3a/3b knockout because preimplantation iXCI status maintains extraembryonic cells (Chen et al., 2019; Galupa and Heard, 2015; Schulz and Heard, 2013). The results showed that Dnmt knockout led to a nearly complete loss of H3k27me3 on paternal (specifically inactivated in iXCI), which showed a notable transcriptional upregulation cross the chromosome. By contrast, these changes cannot be not observed on maternal X chromosome (Figure 3—figure supplement 4 in the revised manuscript). We have added these results in the revised manuscript.

- Figure 3E: Given all tested proteins give a positive signal, it would have been good to include a negative control chromatin protein that is known to not interact with DNMT3B. Given both PRC2 and DNMT3B are chromatin-binding proteins, can the signal be a result of close proximity instead of a direct interaction?

In the present study, to test the interaction between DNMT3B and PRC2 core components, we have used in situ proximity ligation assay (PLA), an increasingly popular technique for detecting the close proximity of two proteins in fixed samples using two primary antibodies (Alsemarz et al., 2018).

Author response image 2.

Schematic diagram of the principle of the in situ PLA.

Compared with classical co-Immunoprecipitation (Co-IP) method, in situ PLA has advantages in (1) detecting low input samples or proteins expressed at low levels, which is extremely difficult using Co-IP; (2) providing in situ or subcellular information of protein-protein interaction. However, it should be noted that the maximal distance allowing this reaction is 40 nm, which is not quite small enough to demonstrate a physical interaction between the two antigens, but sufficient to support a very close “proximity”.

In our study, in situ PLA, including the experimental design of negative control, was performed in the accordance with the manufacturer’s instruction of Duolink® In Situ Red Starter Kit (MilliporeSigma): “Technical negative controls included incubation with each primary antibody separately and no primary antibody”. We have refined the relevant sentence in the revised manuscript (Lines 308-310 in the revised manuscript)

- Figure 3G: It would have been good to include a negative control, and DNase/benzonase to exclude DNA/RNA-mediated protein interaction.

- (Of note, there have been previous studies reporting an interaction between PRC2 and DNMT3B in other cell types, such as in Weigert et al. 2023, but unfortunately, they don't seem to use DNase/benzonase either).

The Co-IP analysis of DNMT3B and PRC2 core components in differentiated female ES cells was presented as additional supportive evidence. Because the Co-IP analysis is extremely difficult for preimplantation embryos, we have used in situ PLA to detect their interaction. However, the maximal distance allowing in situ PLA reaction is 40 nm, which is not quite small enough to demonstrate a physical interaction (Alsemarz et al., 2018). Thus, we have added a Co-IP analysis using differentiated female ES cells, in which rXCI occurs upon the differentiation.

Based on this consideration of the importance and contribution of this result, we have moved this result from the main figure, to the supplemental figure (Figure 4—figure supplement 3H in the revised manuscript).

- Figure 3 figure supplement 3G: what were the ESCs differentiated into? Did the Dnmt3b KO or Dnmt3a/b DKO show any differentiation defect?

The mouse ESC line PGK12.1 was a well-established ex vivo model of rXCI. Under the standard culture condition, PGK12.1 is normally fated to neuroectodermal commitment.

Author response image 3.

Immunostaining of NESTIN, a neuroectodermal stem cell marker molecule, and NANOG in undifferentiated and differentiated PGK12.1 ESCs respectively.

No differentiation defects have been observed in either Dnmt3b KO or Dnmt3a/3b DKO ESCs in our study. Dnmt KO/DKO/TKO ES cell lines have been successfully used as the model of interaction of DNA methylation and H3K27me3 deposition (King et al., 2016).

Figure 4:

- Figure 4B: Is there an explanation for seeing similar total cell numbers in Figure 4B, but showing decreased proliferation in Figure 4A?

Thank you for your insightful comments. The EdU cell proliferation assays labels cells during the S phase of cell cycle, as the 5-ethynyl 2´-deoxyuridine (EdU) is incorporated into newly synthesized DNA. This labeling identifies cells undergoing DNA synthesis, but these cells may not have completed mitosis at the time of detection. As a result, the total cell number may not immediately reflect the decrease in proliferation observed in the treated group. To address this point, we have rewritten the sentences in the revised manuscript (Lines 174-175 in the revised manuscript).

References

Alsemarz, A., Lasko, P. and Fagotto, F. J. B. (2018). Limited significance of the in situ proximity ligation assay. bioRxiv, 411355.

Auclair, G., Guibert, S., Bender, A. and Weber, M. (2014). Ontogeny of CpG island methylation and specificity of DNMT3 methyltransferases during embryonic development in the mouse. Genome Biol. 15, 545.

Balaton, B. P. and Brown, C. J. (2021). Contribution of genetic and epigenetic changes to escape from X-chromosome inactivation. Epigenetics Chromatin 14, 30.

Bartke, T., Vermeulen, M., Xhemalce, B., Robson, S. C., Mann, M. and Kouzarides, T. (2010). Nucleosome-interacting proteins regulated by DNA and histone methylation. Cell 143, 470-484.

Borgel, J., Guibert, S., Li, Y., Chiba, H., Schubeler, D., Sasaki, H., Forne, T. and Weber, M. (2010). Targets and dynamics of promoter DNA methylation during early mouse development. Nat. Genet. 42, 1093-1100.

Chen, Z., Yin, Q., Inoue, A., Zhang, C. and Zhang, Y. (2019). Allelic H3K27me3 to allelic DNA methylation switch maintains noncanonical imprinting in extraembryonic cells. Sci Adv 5, eaay7246.

Chiba, H., Hirasawa, R., Kaneda, M., Amakawa, Y., Li, E., Sado, T. and Sasaki, H. (2008). De novo DNA methylation independent establishment of maternal imprint on X chromosome in mouse oocytes. Genesis 46, 768-774.

Choi, S. H., Heo, K., Byun, H. M., An, W., Lu, W. and Yang, A. S. (2011). Identification of preferential target sites for human DNA methyltransferases. Nucleic Acids Res. 39, 104-118.

Chow, J. and Heard, E. (2009). X inactivation and the complexities of silencing a sex chromosome. Curr. Opin. Cell Biol. 21, 359-366.

Dahlet, T., Argueso Lleida, A., Al Adhami, H., Dumas, M., Bender, A., Ngondo, R. P., Tanguy, M., Vallet, J., Auclair, G., Bardet, A. F., et al. (2020). Genome-wide analysis in the mouse embryo reveals the importance of DNA methylation for transcription integrity. Nat Commun 11, 3153.

Duymich, C. E., Charlet, J., Yang, X. J., Jones, P. A. and Liang, G. N. (2016). DNMT3B isoforms without catalytic activity stimulate gene body methylation as accessory proteins in somatic cells. Nat Commun 7, 11453.

Fukuda, A., Tomikawa, J., Miura, T., Hata, K., Nakabayashi, K., Eggan, K., Akutsu, H. and Umezawa, A. (2014). The role of maternal-specific H3K9me3 modification in establishing imprinted X-chromosome inactivation and embryogenesis in mice. Nat Commun 5, 5464.

Galupa, R. and Heard, E. (2015). X-chromosome inactivation: new insights into cis and trans regulation. Curr. Opin. Genet. Dev. 31, 57-66.

Gontan, C., Mira-Bontenbal, H., Magaraki, A., Dupont, C., Barakat, T. S., Rentmeester, E., Demmers, J. and Gribnau, J. (2018). REX1 is the critical target of RNF12 in imprinted X chromosome inactivation in mice. Nat Commun 9, 4752.

Hagarman, J. A., Motley, M. P., Kristjansdottir, K. and Soloway, P. D. (2013). Coordinate regulation of DNA methylation and H3K27me3 in mouse embryonic stem cells. PLoS One 8, e53880.

Heard, E., Chaumeil, J., Masui, O. and Okamoto, I. (2004). Mammalian X-chromosome inactivation: an epigenetics paradigm. Cold Spring Harb. Symp. Quant. Biol. 69, 89-102.

Huynh, K. D. and Lee, J. T. (2005). X-chromosome inactivation: a hypothesis linking ontogeny and phylogeny. Nat. Rev. Genet. 6, 410-418.

Inoue, A., Jiang, L., Lu, F. and Zhang, Y. (2017). Genomic imprinting of Xist by maternal H3K27me3. Genes Dev. 31, 1927-1932.

Inoue, K., Kohda, T., Sugimoto, M., Sado, T., Ogonuki, N., Matoba, S., Shiura, H., Ikeda, R., Mochida, K., Fujii, T., et al. (2010). Impeding Xist expression from the active X chromosome improves mouse somatic cell nuclear transfer. Science 330, 496-499.

Jermann, P., Hoerner, L., Burger, L. and Schubeler, D. (2014). Short sequences can efficiently recruit histone H3 lysine 27 trimethylation in the absence of enhancer activity and DNA methylation. Proc. Natl. Acad. Sci. U. S. A. 111, E3415-3421.

King, A. D., Huang, K., Rubbi, L., Liu, S., Wang, C. Y., Wang, Y., Pellegrini, M. and Fan, G. (2016). Reversible Regulation of Promoter and Enhancer Histone Landscape by DNA Methylation in Mouse Embryonic Stem Cells. Cell Rep. 17, 289-302.

Maslov, A. Y., Lee, M., Gundry, M., Gravina, S., Strogonova, N., Tazearslan, C., Bendebury, A., Suh, Y. and Vijg, J. (2012). 5-aza-2'-deoxycytidine-induced genome rearrangements are mediated by DNMT1. Oncogene 31, 5172-5179.

Oikawa, M., Inoue, K., Shiura, H., Matoba, S., Kamimura, S., Hirose, M., Mekada, K., Yoshiki, A., Tanaka, S., Abe, K., et al. (2014). Understanding the X chromosome inactivation cycle in mice: a comprehensive view provided by nuclear transfer. Epigenetics-Us 9, 204-211.

Oka, M., Meacham, A. M., Hamazaki, T., Rodic, N., Chang, L. J. and Terada, N. (2005). De novo DNA methyltransferases Dnmt3a and Dnmt3b primarily mediate the cytotoxic effect of 5-aza-2'-deoxycytidine. Oncogene 24, 3091-3099.

Pintacuda, G. and Cerase, A. (2015). X Inactivation Lessons from Differentiating Mouse Embryonic Stem Cells. Stem Cell Rev Rep 11, 699-705.

Schulz, E. G. and Heard, E. (2013). Role and control of X chromosome dosage in mammalian development. Curr. Opin. Genet. Dev. 23, 109-115.

Smith, Z. D., Chan, M. M., Mikkelsen, T. S., Gu, H. C., Gnirke, A., Regev, A. and Meissner, A. (2012). A unique regulatory phase of DNA methylation in the early mammalian embryo. Nature 484, 339-344.

Tada, T., Obata, Y., Tada, M., Goto, Y., Nakatsuji, N., Tan, S., Kono, T. and Takagi, N. (2000). Imprint switching for non-random X-chromosome inactivation during mouse oocyte growth. Development 127, 3101-3105.

Tan, K., An, L., Miao, K., Ren, L., Hou, Z., Tao, L., Zhang, Z., Wang, X., Xia, W., Liu, J., et al. (2016). Impaired imprinted X chromosome inactivation is responsible for the skewed sex ratio following in vitro fertilization. Proc. Natl. Acad. Sci. U. S. A. 113, 3197-3202.

Vire, E., Brenner, C., Deplus, R., Blanchon, L., Fraga, M., Didelot, C., Morey, L., Van Eynde, A., Bernard, D., Vanderwinden, J. M., et al. (2006). The Polycomb group protein EZH2 directly controls DNA methylation. Nature 439, 871-874.

Zhao, J., Yao, K., Yu, H., Zhang, L., Xu, Y., Chen, L., Sun, Z., Zhu, Y., Zhang, C., Qian, Y., et al. (2021). Metabolic remodelling during early mouse embryo development. Nat Metab 3, 1372-1384.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1

(1) In the "Introduction" section, an important aspect that requires attention pertains to the discussion surrounding the heterodimerization of CXCR4 and CCR5. Notably, the manuscript overlooks a recent study (https://doi.org/10.1038/s41467-023-42082-z) elucidating the mechanism underlying the formation of functional dimers within these G protein-coupled receptors (GPCRs)…The inclusion of this study within the manuscript would significantly enrich the contextual framework of the work, offering readers a comprehensive understanding of the current knowledge surrounding the structural dynamics and functional implications of CXCR4 and CCR5 heterodimerization.

We thank the reviewer for his/her recommendation to enrich the contextual framework of our study. The Nature Communications paper by Di Marino et al. was published after we sent the first version of our manuscript to eLife, and therefore was not included in the discussion. As the reviewer rightly indicates, this paper elucidates the mechanism underlying the formation of functional dimers within CCR5 and CXCR4. Using metadynamics approaches, the authors emphasize the importance of distinct transmembrane regions for dimerization of the two receptors. In particular, CXCR4 shows two low energy dimer structures and the TMVI-TMVII helices are the preferred interfaces involved in the protomer interactions in both cases. Although the study uses in silico techniques, it also includes the molecular binding mechanism of CCR5 and CXCR4 in the membrane environment, as the authors generate a model in which the receptors are immersed in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) phospholipid bilayer with 10% cholesterol. This is an important point in this study, as membrane lipids also interact with membrane proteins, and the lipid composition affects CXCR4 oligomerization (Gardeta S.R. et al. Front. Immunol. 2023). In particular, Di Marino et al. find a cholesterol molecule placed in-between the two CXCR4 protomers where it engages a series of hydrophobic interactions with residues including Leu132, Val214, Leu216 and Phe249. Then, the polar head of cholesterol forms an H-bond with Tyr135 that further stabilizes protomer binding. In our hands, the F249L mutation in CXCR4 reverted the antagonism of AGR1.137, suggesting that the compound binds, among others, this residue. We should, nonetheless, indicate that we analyzed receptor oligomerization and not CXCR4 dimerization, which was the main object of the Di Marino et al. study. It is therefore also plausible that other residues than those described as essential for CXCR4 dimerization might participate in receptor oligomerization. We can speculate that AGR1.137 might affect cholesterol binding to CXCR4 and, therefore, alter dimerization/oligomerization. Additionally, the CXCR4 x-ray structure with PDB code 3ODU (Wu B. et al. Science, 2010) experimentally shows the presence of two fatty acid molecules in contact with both TMV and TMVI. These molecules closely interact with hydrophobic residues in the protein, thereby stabilizing it in a hydrophobic environment. Although more experiments will be needed to clarify the mechanism involved, our results suggest that cholesterol and/or other lipids also play an important role in CXCR4 oligomerization and function, as seen for other GPCRs (Jakubik J. & ElFakahani E.E. Int J Mol Sci. 2021). However, we should also consider that other factors not included in the analysis by Di Marino et al. can also affect CXCR4 oligomerization; for instance, the co-expression of other chemokine receptors and/or other GPCRs that heterodimerize with CXCR4 might affect CXCR4 dynamics at the cell membrane, similar to other membrane proteins such as CD4, which also forms complexes with CXCR4 (Martinez-Muñoz L. et al. Mol. Cell 2018).

The revised discussion contains references to the study by Di Marino et al. to enrich the contextual framework of our data.

(2) In "various sections" of the manuscript, there appears to be confusion surrounding the terminology used to refer to antagonists. It is recommended to provide a clearer distinction between allosteric and orthosteric antagonists to enhance reader comprehension. An orthosteric antagonist typically binds to the same site as the endogenous ligand, directly blocking its interaction with the receptor. On the other hand, an allosteric antagonist binds to a site distinct from the orthosteric site, inducing a conformational change in the receptor that inhibits the binding of the endogenous ligand. By explicitly defining the terms "allosteric antagonist" and "orthosteric antagonist" within the manuscript, readers will be better equipped to discern the specific mechanisms discussed in the context of the study.

The behavior of the compounds described in our manuscript (AGR1.35 and AGR1.137) fits with the definition of allosteric antagonists, as they bind on a site distinct from the orthosteric site, although they only block some ligand-mediated functions and not others. This would mean that they are not formally antagonists and should be not considered as allosteric compounds, as their binding on CXCR4 does not alter CXCL12 binding, although they might affect its affinity. In this sense, our compounds respond much better to the concept of negative allosteric modulators (Gao Z.-G. & Jacobson K.A. Drug Discov. Today Technol. 2013). They act by binding on a site distinct from the orthosteric site and selectively block some downstream signaling pathways but not others induced by the same endogenous agonist.

To avoid confusion and to clarify the role of the compounds described in this study, we now refer to them as negative allosteric modulators along the manuscript.

(3) In the Results section, the computational approach employed for "screening small compounds targeting CXCR4, particularly focusing on the inhibition of CXCL12-induced CXCR4 nanoclustering", requires clarification due to several points of incomprehension. The following recommendations aim to address these concerns and enhance the overall clarity of the section:

(1) Computational Approach and Binding Mode Description: 

-Explicitly describe the methodology for identifying the pocket/clef area in angstroms (Å) on the CXCR4 protein structure. Include details on how the volume of the cleft enclosed by TMV and TMVI was determined, as this information is not readily apparent in the provided reference (https://doi.org/10.1073/pnas.1601278113).

The identification of the cleft was based on the observations by Wu et al. (Wu B. et al. Science 2010) who described the presence of bound lipids in the area formed by TMV and VI, and those of Wescott et al. (Wescott M.P. et al. Proc. Natl. Acad. Sci. 2016) on the importance of TMVI in the transmission of conformational changes promoted by CXCL12 on CXCR4 towards the cytoplasmic surface of the receptor to link the binding site with signaling activation. Collectively, these results, and our previous data on the critical role of the N-terminus region of TMVI for CXCR4 oligomerization (Martinez-Muñoz L. et al. Mol. Cell 2018), focused our in silico screening to this region. Once we detected that several compounds bound CXCR4 in this region, the cleavage properties were calculated by subtracting the compound structure. The resulting PDB was analyzed using the PDBsum server (Laskowski R.A. et. al. Protein Sci. 2018). Volume calculations were obtained using the server analyzing surface clefts by SURFNET (Laskowski R. A. J. Mol. Graph. 1995). The theoretical interaction surface between the selected compounds and CXCR4 and the atomic distances between the protein residues and the compounds was calculated using the PISA server (Krissinel E. & Henrick K. J. Mol. Biol. 2007) (Fig. I, only for review purposes). The analysis of the cleft occupied by AGR1.135 showed two independent cavities of 434 Å3 and 1,381 Å3 that were not connected to the orthosteric site. In the case of AGR1.137, the data revealed two distinct clefts of 790 Å3 and 580 Å3 (Fig. I, only for review purposes). These details have been included in the revised manuscript (New Fig. 1A, Supplementary Fig 8A, B).

(4) Clarify the statement regarding the cleft being "surface exposed for interactions with the plasma membrane," particularly in the context of its embedding within the membrane.

For GPCRs, transmembrane domains represent binding sites for bioactive lipids that play important functional and physiological roles (Huwiler A. & Zangemeister-Wittke U. Pharmacol. Ther. 2018). The channel between TMV and TMVI connects the orthosteric chemokine binding pocket to the lipid bilayer and is occupied by an oleic acid molecule, according to the CXCR4 structure published in 2010 (Wu B. et al. Science 2010). In addition, the target region contains residues involved in cholesterol (and perhaps other lipids) engagement (Di Marino et al. Nat. Commun. 2023). Taken together, these data support our statement that the cleft supports interactions between CXCR4 molecules and the plasma membrane. 

Moreover, the data of Di Marino et al. also support that CCR5 and CXCR4 have a symmetric and an asymmetric binding mode. Therefore, either dimeric structure has the possibility to form trimers, tetramers, and even oligomers by using the free binding interface to complex with another protomer. This hypothesis suggests that the interaction of dimers to form oligomers should involve residues distinct from those included in the dimeric conformation.

The sentence has been modified in the revised manuscript to clarify comprehension.

(5) Discuss the rationale behind targeting the allosteric binding pocket instead of the orthosteric pocket, outlining potential advantages and disadvantages.

The advantages and disadvantages of using negative allosteric modulators vs orthosteric antagonists have been now included in the revised discussion. 

The majority of GPCR-targeted drugs function by binding to the orthosteric site of the receptor, and are agonists, partial agonists, antagonists or inverse agonists. These orthosteric compounds can have off-target effects and poor selectivity due to highly homologous receptor orthosteric sites and to abrogation of spatial and/or temporal endogenous signaling patterns. 

The alternative is to use allosteric modulators, which can tune the functions associated with the receptors without affecting the orthosteric site. They can be positive, negative or neutral modulators, depending on their effect on the functionality of the receptor (Foster D.J. & Conn P.J. Neuron 2017). For example, the use of a negative allosteric modulator of a chemokine receptor to dampen pathological signaling events, while retaining full signaling for non-pathological activities might limit adverse effects (Kohout T.A.et al. J. Biol. Chem. 2004). In this case, the negative allosteric modulator 873140 blocks CCL3 binding on CCR5 but does not alter CCL5 binding (Watson C. et al. Mol. Pharmacol. 2005). In other cases, allosteric modulators can stabilize a particular receptor conformation and block others. The mechanism of action of the anti-HIV-1, FDAapproved, CCR5 allosteric modulator, maraviroc (Jin J. et al. Sci. Signal. 2018) is attributed to its ability to modulate CCR5 dimer populations and their subsequent subcellular trafficking and localization to the cell membrane (Jin J .et al. Sci. Signal. 2018). Two CCR5 dimeric conformations that are imperative for membrane localization were present in the absence of maraviroc; however, an additional CCR5 dimer conformation was discovered after the addition of maraviroc, and all homodimeric conformations were further stabilized. This finding is consistent with the observation that CCR5 dimers and oligomers inhibit HIV host-cell entry, likely by preventing the HIV-1 co-receptor formation.

It is well known that GPCRs activate G proteins, but they also recruit additional proteins (e.g., β-arrestins) that induce signaling cascades which, in turn, can direct specific subsets of cellular responses independent of G protein activation (Eichel K. et al. Nature 2018) and are responsible for either therapeutic or adverse effects. Allosteric modulators can thus be used to block these adverse effects without influencing the therapeutic benefits. This was the case in the design of G protein-biased agonists for the kappa opioid receptor, which maintain the desirable antinociceptive and antipruritic effects and eliminate the sedative and dissociative effects in rodent models (Brust T.F. et al. Sci. Signal 2016).

(6) Provide the PDB ID of the CXCR4 structure used as a template for modeling with SwissModel. Explain the decision to model the structure from the amino acid sequence and suggest an alternative approach, such as utilizing AlphaFold structures and performing classical molecular dynamics with subsequent clustering for the best representative structure.

The PDB used as a template for modeling CXCR4 was 3ODU. This information was already included in the material and methods section. At the time we performed these analyses, there were several crystallographic structures of CXCR4 in complex with different molecules and peptides deposited at the PDB. None of them included a full construct containing the complete receptor sequence to provide a suitable sample for Xray structure resolution, as the N- and C-terminal ends of CXCR4 are very flexible loops. In addition, the CXCR4 constructs contained T4 lysozyme inserted between helices TMV and TMVI to increase the stability of the protein––a common strategy used to facilitate crystallogenesis of GPCRs (Zou Y. et al. PLoS One 2012). Therefore, we generated a CXCR4 homology model using the SWISS-MODEL server (Waterhouse A. et al. Nucleic Acids Res. 2018). This program reconstructed the loop between TMV and TMVI, a domain particularly important in this study that was not present in any of the crystal structure available in PDB. The model structure was, nonetheless, still incomplete, as it began at P27 and ended at S319 because the terminal ends were not resolved in the crystal structure used as a template. Nevertheless, we considered that these terminal ends were not involved in CXCR4 oligomerization. 

As Alphafold was not available at the time we initiated this project, we didn’t use it. However, we have now updated our workflow to current methods and predicted the structure of the target using AlphaFold (Jumper J. et al. Nature 2021) and the sequence available under UniProt entry P61073. We prepared the ligands using OpenBabel (O’Boyle N.M. et al., J. Cheminformatics 2011), with a gasteiger charge assignment, and generated 10 conformers for each input ligand using the OpenBabel genetic algorithm. We then prepared the target structure with Openmm, removing all waters and possible heteroatoms, and adding all missing atoms. We next predicted the target binding pockets with fPocket (Le Guilloux V. et al. BMC Bioinformatics 2009), p2rank (Krivak R. & Hoksza, J. Cheminformatics 2018), and AutoDock autosite (Ravindranath P.A. & Sanner M.F. Bioinformatics 2016). We chose only those pockets between TMV and TMVI (see answer to point 3). We merged the results of the three programs into so-called consensus pockets, as two pockets are said to be sufficiently similar if at least 75% of their surfaces are shared (del Hoyo D. et al. J. Chem. Inform. Model. 2023). From the consensus pockets, there was one pocket that was significantly larger than the others and was therefore selected. We then docked the ligand conformers in this pocket using AutoDock GPU (Santos-Martins D. et al. J. Chem. Theory Comput. 2021), LeDock (Liu N & Xu Z., IOP Conf. Ser. Earth Environ. Sci. 2019), and Vina (Eberhardt J. et al. J. Chem. Inf. Model. 2021). The number of dockings varied from 210 to 287 poses. We scored each pose with the Vina score using ODDT (Wójcikowski M. et al. J. Cheminform. 2015). Then, we clustered the different solutions into groups whose maximum RMSD was 1Å. This resulted in 40 clusters, the representative of each cluster was the one with maximum Vina score and confirmed that the selected compounds bound this pocket (Author response image 1). When required, we calculated the binding affinity using Schrodinger’s MM-GBSA procedure (Greenidge P.A. et al. J. Chem. Inf. Model. 2013), in two ways: first, assuming that the ligand and target are fixed; second, with an energy minimization of all the atoms within a distance of 3Å from the ligand. This information has now been included in the revised version of the manuscript.

Author response image 1.

AGR1.135 docking in CXCR4 using the updated protocol for ligand docking. Cartoon representation colored in gray with TMV and TMVI shown in blue and pink, respectively. AGR1.135 is shown in stick representation with carbons in yellow, oxygens in red and nitrogens in blue.

(7) Specify the meaning of "minimal interaction energy" and where (if present) the interaction scores are reported in the text.

We refer to minimal interaction energy, the best docking score, that is, the best score obtained in our docking studies. These data were not included in the previous manuscript due to space restrictions but are now included in the reviewed manuscript.

(8) You performed docking studies using GLIDE to identify potential binding sites for the small compounds on the CXCR4 protein. The top-scoring binders were then subjected to further refinement using PELE simulations. However, I realize that a detailed description of the specific binding modes of these compounds was not provided in the text. Please make the description of binding poses more detailed

Firstly, to assess the reliability of this method, a PELE study was carried out for the control molecule IT1t, which is a small drug-like isothiourea derivative that has been crystallized in complex with CXCR4 (PDB code: 3ODU). IT1t is a CXCR4 antagonist that binds to the CXCL12 binding cavity and inhibits HIV-1 infection (Das D. Antimicrob. Agents Chemother. 2015; Dekkers S. et al. J. Med. Chem. 2023). From the best five trajectories, two of them had clearly better binding energies, and corresponded to almost the same predicted pose of the molecule. Although the predicted binding mode was not exactly the same as the one in the crystal structure, the approximation was very good, giving validation to the approach. Although PELE is a suitable technique to find potential binding sites, the predicted poses must be subsequently refined using docking programs.

Analyzing the best trajectories for the remaining ligands, at least one of the best-scored poses was always located at the orthosteric binding site of CXCR4. Even though these poses showed good binding energies, they were discarded as the in vitro biological experiments indicated that the compounds were unable to block CXCL12 binding or CXCL12-mediated inhibition of cAMP release or CXCR4 internalization. Collectively, these data indicated that the selected compounds did not behave as orthosteric inhibitors of CXCR4. The CXCL12 binding pocket is the biggest cavity in CXCR4, and so PELE may tend to place the molecules near it. However, all the compounds presented other feasible binding sites with a comparable binding energy.

AGR1.135 and AGR1.137 showed interesting poses between TMV and TMVI with very good binding energy (-51.4 and -37.2 kcal/mol, respectively). This was precisely the region we had previously selected for the in silico screening, as previously described (see response to point 3).

AGR1.131 showed two poses with low binding energy that were placed between helices TMI and TMVII (-43.6 kcal/mol) and between helices TMV and TMVI (-39.8 kcal/mol). This compound was unable to affect CXCL12-mediated chemotaxis and was therefore used as an internal negative control as it was selected in the in silico screening with the same criteria as the other compounds but failed to alter any CXCL12-mediated functions. PELE studies nonetheless provided different binding sites for each molecule, which had to be further studied using docking to obtain a more accurate binding mode. In agreement with the previous commentary, we repeated the analysis using AlphaFold and the rest of the procedure described (see our response to point 6) and calculated the binding energies for all the compounds using Schrodinger’s MM-GBSA procedure (Greenidge P.A. et al. J. Chem. Inf. Model. 2013). Calculations were performed in two ways: first, assuming that the ligand and target are fixed; second, with an energy minimization of all the atoms within a distance of 3Å from the ligand. The results using the first method indicated that AGR1.135 and AGR1.137 showed poses between TMV and TMVI with - 56.4 and -62.4 kcal/mol, respectively and AGR1.131 had a pose between TMI and TMVII with -61.6kcal/mol.  In the second method AGR1.135 and AGR1.137 showed poses between TMV and TMVI with -57.9, and -67.6 kcal/mol, respectively, and AGR1.131 of -62.2 kcal/mol between TMI and TMVII.

This information is now included in the text.

(9) (2) Experimental Design:-Justify the choice of treating Jurkat cells with a concentration of 50 μM of the selected compound. Consider exploring different concentrations and provide a rationale for the selected dosage. Additionally, clearly identify the type of small compound used in the initial experiment.

The revised version contains a new panel in Fig. 1B to show a more detailed kinetic analysis with different concentrations (1-100 µM) of the compounds in the Jurkat migration experiments. In all cases, 100 µM nearly completely abrogated cell migration, but in order to reduce the amount of DMSO added to the cells we selected 50 µM for further experiments, as it was the concentration that inhibits 50-75% of ligand-induced cell migration. Regarding the type of small compounds used in the initial experiments, they were compounds included in the library described in reference #24 (Sebastian-Pérez V. et al Med. Biol. Chem. 2017), which contains heterocyclic compounds. We would note that we do not consider AGR1.137 a final compound. We think that there is scope to develop AGR1.137-based second-generation compounds with greater solubility in water, greater specificity or affinity for CXCR4, and to evaluate delivery methods to hopefully increase activity.  

(10) Avoid reporting details in rounded parentheses within the text; consider relocating such information to the Materials and Methods section or figure captions for improved readability.

Most of the rounded parentheses within the text have been eliminated in the revised version of the manuscript to improve readability.

(11) Elaborate on the virtual screening approach using GLIDE software, specifying the targeted site and methodology employed.

For the virtual screening, we used the Glide module (SP and XP function scoring) included in the Schrödinger software package, utilizing the corresponding 3D target structure and our MBC library (Sebastián-Pérez V et al. J. Chem. Inf. Model. 2017).  The center of the catalytic pocket was selected as the centroid of the grid. In the grid generation, a scaling factor of 1.0 in van der Waals radius scaling and a partial charge cutoff of 0.25 were used. A rescoring of the SP poses of each compound was then performed with the XP scoring function of the Glide. The XP mode in Glide was used in the virtual screening, the ligand sampling was flexible, epik state penalties were added and an energy window of 2.5 kcal/mol was used for ring sampling. In the energy minimization step, the distance-dependent dielectric constant was 4.0 with a maximum number of minimization steps of 100,000. In the clustering, poses were considered as duplicates and discarded if both RMS deviation is less than 0.5 Å and maximum atomic displacement is less than 1.3 Å.

(12) Provide clarity on the statement that AGR1.131 "theoretically" binds the same motif, explaining the docking procedure used for this determination.

In the in silico screening, AGR1.131 was one of the 40 selected compounds that showed, according to the PELE analysis (see answer to point 8), a pose with low binding energy (-39.8 kcal/mol) between TMV and TMVI helices, which is the selected area for the screening. It, nonetheless, also showed a best pose placed between helices TM1 and TM7 (-43.7 kcal/mol) using the initial workflow. In conclusion, although AGR1.131 also faced to the TMV-TMVI, the most favorable pose was in the area between TMI and TMVII. In addition, the compound was included in the biological screening, where it did not affect CXCL12-mediated chemotaxis. We thus decided to use it as an internal negative control, as it has a skeleton very similar to AGR1.135 and AGR1.137 and can interact with the TM domains of CXCR4 without promoting biological effects. This statement has been clarified in the revised text.

(13) Toxicity Testing:

-Enhance the explanation of the approach to testing the toxicity of the compound in Jurkat cells. Consider incorporating positive controls to strengthen the assessment and clarify the experimental design.

All the selected compounds in the in silico screening were initially tested for propidium iodide incorporation in treated cells in a toxicity assay, and some of them were discarded for further experiments (e.g., AGR1.103 and VSP3.1).

Further evaluation of Jurkat cell viability was determined by cell cycle analysis using propidium iodide.  Supplementary Fig. 1B included the percentage of each cell cycle phase, and data indicated no significant differences between the treatments tested. Nevertheless, at the suggestion of the reviewer, and to clarify this issue, positive controls inducing Jurkat cell death (staurosporine and hydrogen peroxide) have also been included in the new Supplementary Fig. 2. The new figure also includes a table showing the percentage of cells in each cell-cycle phase.  

(14) In the Results section concerning "AGR1.135 and AGR1.137 blocking CXCL12-mediated CXCR4 nanoclustering and dynamics", several points can be improved to enhance clarity and coherence: 1. Specificity of Low Molecular Weight Compounds:  

-Clearly articulate how AGR1.135 and AGR1.137 specifically target homodimeric CXCR4 and provide an explanation for their lack of impact on heterodimeric CXCR4-CCR5 in that region.

First of all, we should clarify that when we talk about receptor nanoclustering, oligomers refer to complexes including 3 or more receptors and, therefore, the residues involved in these interactions can differ from those involved in receptor dimerization. Moreover, our FRET experiments did not indicate that the compounds alter receptor dimerization (see new Supplementary Fig. 7). Of note, mutant receptors unable to oligomerize can still form dimers (Martínez-Muñoz L. et al. Mol. Cell 2018; García-Cuesta E.M .et al. Proc. Natl. Acad. Sci. USA 2022). Additionally, we believe that these oligomers can also include other chemokine receptors/proteins expressed at the cell membrane, which we are currently studying using different models and techniques.

We have results supporting the existence of CCR5/CXCR4 heterodimers (Martínez-Muñoz L et al. Proc. Natl. Acad. Sci. USA 2014), in line with the data published by Di Marino et al. However, in the current study we have not evaluated the impact of the selected compounds on other CXCR4 complexes distinct from CXCR4 oligomers. Our Jurkat cells do not express CCR5 and, therefore, we cannot discuss whether AGR1.137 affects CCR5/CXCR4 heterodimers. The chemokine field is very complex and most receptors can form dimers (homo- and heterodimers) as well as oligomers (Martinez-Muñoz L., et al Pharmacol & Therap. 2011) when co-expressed. To evaluate different receptor combinations in the same experiment is a complex task, as the number of potential combinations between distinct expressed receptors makes the analysis very difficult. We started with CXCR4 as a model, to continue later with other possible CXCR4 complexes. In addition, for the analysis of CCR5/CXCR4 dynamics, it is much better to use dual-TIRF techniques, which allow the simultaneous detection of two distinct molecules coupled to different fluorochromes.

Regarding the data of Di Marino et al., it is possible that the compounds might also affect heterodimeric conformations of CXCR4. This aspect has also been broached in the revised discussion. We would again note that we evaluated CXCR4 oligomers and not monomers or dimers; this is especially relevant when we compare the residues involved in these processes as they might differ depending on the receptor conformation considered. This issue was also hypothesized by Di Marino et al. (see our response to point 4).

(15) When referring to "unstimulated" cells, provide a more detailed explanation to elucidate the experimental conditions and cellular state under consideration.

Unstimulated cells refer to the cells in basal conditions, that is, cells in the absence of CXCL12. For TIRF-M experiments, transiently-transfected Jurkat cells were plated on glass-bottomed microwell dishes coated with fibronectin; these are the unstimulated cells. To observe the effect of the ligand, dishes were coated as above plus CXCL12 (stimulated cells). We have clarified this point in the material and methods section of the revised version.

(16) 2. Paragraph Organization

-Reorganize the second paragraph to eliminate redundancy and improve overall flow. A more concise and fluid presentation will facilitate reader comprehension and engagement.

The second paragraph has been reorganized to improve overall flow.

(17) Ensure that each paragraph contributes distinct information, avoiding repetition and redundancy.

We have carefully revised each paragraph of the manuscript to avoid redundancy.

(18) 3. Claim of Allosteric Antagonism:

-Exercise caution when asserting that "AGR1.135 and AGR1.137 behave as allosteric antagonists of CXCR4" based on the presented results. Consider rephrasing to reflect that the observed effects suggest the potential allosteric nature of these compounds, acknowledging the need for further investigations and evidence.

To avoid misinterpretations on the effect of the compounds on CXCR4, as we have commented in our response to point 2, we have substituted the term allosteric inhibitors with negative allosteric modulators, which refer to molecules that act by binding a site distinct from the orthosteric site, and selectively block some downstream signaling pathways, whereas others induced by the same endogenous or orthosteric agonist are unaffected (Gao Z.-G. & Jacobson K.A. Drug Discov. Today Technol. 2013). Our data indicate that the selected small compounds do not block ligand binding or G protein activation or receptor internalization, but inhibit receptor oligomerization and ligand-mediated directed cell migration.

(19) In the Results section discussing the "incomplete abolition of CXCR4-mediated responses in Jurkat cells by AGR1.135 and AGR1.137", several points can be refined for better clarity and completeness:  1. Inclusion of Positive Controls: 

-Consider incorporating positive controls in relevant experiments to provide a comparative benchmark for assessing the impact of AGR1.135 and AGR1.137. This addition will strengthen the interpretation of results and enhance the experimental rigor. 

The in vivo experiments (Fig. 7E,F) used AMD3100, an orthosteric antagonist of CXCR4, as a positive control. We also included AMD3100, as a positive control of inhibition when evaluating the effect of the compounds on CXCL12 binding (Fig. 3, new Supplementary Fig. 3). The revised version of the manuscript also includes the effect of this inhibitor on other relevant CXCL12-mediated responses such as cell migration (Fig. 1B), receptor internalization (Fig. 3A), cAMP production (Fig. 3C), ERK1/2 and AKT phosphorylation (Supplementary Fig. 4), actin polymerization (Fig. 4A), cell polarization (Fig. 4B, C) and cell adhesion (Fig. 4D), to facilitate the interpretation of the results and improve the experimental rigor.

(20) 2. Clarification of Terminology: 

-Clarify the term "CXCR4 internalizes" by providing context, perhaps explaining the process of receptor internalization and its relevance to the study.

We refer to CXCR4 internalization as a CXCL12-mediated endocytosis process that results in reduction of CXCR4 levels on the cell surface. We use CXCR4 internalization in this study with two purposes: First, for CXCR4 and other chemokine receptors, internalization processes are mediated by ligand-induced clathrin vesicles (Venkatesan et al 2003) a process that triggers CXCR4 aggregation in these vesicles. We have previously determined that the oligomers of receptors detected by TIRF-M remain unaltered in cells treated with inhibitors of clathrin vesicle formation and of internalization processes (Martinez-Muñoz L. et al. Mol. Cell 2018). Moreover, we have described a mutant CXCR4 that cannot form oligomers but internalizes normally in response to CXCL12 (Martinez-Muñoz L. et al. Mol. Cell 2018). The observation in this manuscript of normal CXCL12-mediated endocytosis in the presence of the negative allosteric inhibitors of CXCR4 that abrogate receptor oligomerization reinforces the idea that the oligomers detected by TIRF are not related to receptor aggregates involved in endocytosis; Second, receptor internalization is not affected by the allosteric compounds, indicating that they downregulate some CXCL12-mediated signaling events but not others (new Fig. 3).

All these data have been included in the revised discussion of the manuscript.

(21) Elaborate on the meaning of "CXCL12 triggers normal CXCR4mut internalization" to enhance reader understanding.

We have previously described a triple-mutant CXCR4 (K239L/V242A/L246A; CXCR4mut). The mutant residues are located in the N-terminal region of TMVI, close to the cytoplasmic region, thus limiting the CXCR4 pocket described in this study (see our response to point 3). This mutant receptor dimerizes but neither oligomerizes in response to CXCL12 nor supports CXCL12-induced directed cell migration, although it can still trigger some Ca2+ flux and is internalized after ligand activation (Martinez-Muñoz L. et al. Mol. Cell 2018).  We use the behavior of this mutant (CXCR4mut) to show that the CXCR4 oligomers and the complexes involved in internalization processes are not the same and to explain why we evaluated CXCR4 endocytosis in the presence of the negative allosteric modulators.

As we indicated in a previous answer to the reviewer, these issues have been re-elaborated in the revised version.

(22) 3. Discrepancy in CXCL12 Concentration:

-Address the apparent discrepancy between the text stating, "...were stimulated with CXCL12 (50 nM, 37{degree sign}C)," and the figure caption (Fig. 3A) reporting a concentration of 12.5 nM. Rectify this inconsistency and provide an accurate and clear explanation.

We apologize for this error, which is now corrected in the revised manuscript. With the exception of the cell migration assays in Transwells, where the optimal concentration was established at 12.5 nM, in the remaining experiments the optimal concentration of CXCL12 employed was 50 nM. These concentrations were optimized in previous works of our laboratory using the same type of experiment. We should also remark that in the experiments using lipid bilayers or TIRF-M experiments, CXCL12 is used to coat the plates and therefore it is difficult to determine the real concentration of the ligand that is retained in the surface of the plates after the washing steps performed prior to adding the cells. In addition, we use 100 nM CXCL12 to create the gradient in the chambers used to perform the directed-cell migration experiments.

(23) 4. Speculation on CXCL12 Binding:

-Refrain from making speculative statements, such as "These data suggest that none of the antagonists alters CXCL12 binding to CXCR4," unless there is concrete evidence presented up to that point. Clearly outline the results that support this conclusion.

Figure 3B and Supplementary Figure 3 show CXCL12-ATTO700 binding by flow cytometry in cells pretreated with the negative allosteric modulators. We have also included AMD3100, the orthosteric antagonist, as a control for inhibition. While these experiments showed no major effect of the compounds on CXCL12 binding, we cannot discard small changes in the affinity of the interaction between CXCL12 and CXCR4. In consequence we have re-written these statements.

(24) 5. Corroboration of Data:

-Specify where the corroborating data from immunostaining and confocal analysis are reported, ensuring readers can access the relevant information to support the conclusions drawn in this section.

In agreement with the suggestion of the reviewer, the revised manuscript includes data from immunostaining and confocal analysis to complement Fig. 4B (new Fig. 4C). The revised version also includes some representative videos for the TIRF experiments showed in Figure 2 to clarify readability.

(25) In the Results section concerning "AGR1.135 and AGR1.137 antagonists and their direct binding to CXCR4", several aspects need clarification and refinement for a more comprehensive and understandable presentation: 1. Workflow Clarification:

-Clearly articulate the workflow used for assessing the binding of AGR1.135 and AGR1.137 to CXCR4. Address the apparent contradiction between the inability to detect a direct interaction and the utilization of Glide for docking in the TMV-TMVI cleft.

To address the direct interaction of the compounds with CXCR4, we intentionally avoided the modification of the small compounds with different labels, which could affect their properties. We therefore attempted a fluorescence a spectroscopy strategy to formally prove the ability of the small compounds to bind CXCR4, but this failed because the AGR1.135 is yellow in color, which interfered with the determinations. We also tried a FRET strategy (see new Supplementary Fig. 7) and detected a significant increase in FRET efficiency of CXCR4 homodimers when AGR1.135 was evaluated, but again the yellow color interfered with FRET determinations. Moreover, AGR1.137 did not modify FRET efficiency of CXCR4 dimers. Therefore, we were unable to detect the interaction of the compounds with CXCR4.

We elected to develop an indirect strategy; in silico, we evaluated the binding-site using docking and molecular dynamics to predict the most promising CXCR4 binding residues involved in the interaction with the selected compounds. Next, we generated point mutant receptors of the predicted residues and re-evaluated the behavior of the allosteric antagonists in a CXCL12-induced cell migration experiment. Obviously, we first discarded those CXCR4 mutants that were not expressed on the cell membrane as well as those that were not functional when activated with CXCL12. Using this strategy, we eliminated the interference due to the physical properties of the compounds and demonstrated that if the antagonism of a compound is reversed in a particular CXCR4 mutant it is because the mutated residue participates or interferes with the interaction between CXCR4 and the compound, thus assuming (albeit indirectly) that the compound binds CXCR4. 

To select the specific mutations included in the analysis, our strategy was to generate point mutations in residues present in the TMV-TMVI pocket of CXCR4 that were not directly proposed as critical residues involved in chemokine engagement, signal initiation, signal propagation, or G protein-binding, based on the extensive mutational study published by Wescott MP et. al. (Wescott M.P. et. al. Proc. Natl. Acad. Sci. U S A. 2016).

(26) Provide a cohesive explanation of the transition from docking evaluation to MD analysis, ensuring a transparent representation of the methodology.

Based on the aim of this work, the workflow shown in Author response image 2, was proposed to predict the binding mode of the selected molecules. Firstly, a CXCR4 model was generated to reconstruct some unresolved parts of the protein structure; then a binding site search using PELE software was performed to identify the most promising binding sites; subsequently, docking studies were performed to refine the binding mode of the molecules; and finally, molecular dynamics simulations were run to determine the most stable poses and predict the residues that we should mutate to test that the compounds interact with CXCR4. 

Author response image 2.

Workflow followed to determine the binding mode of the  studied compounds.

(27) 2. Choice of Software and Techniques:

-Justify the use of "AMBER14" and the PELE approach, considering  their potential obsolescence.

These experiments were performed five years ago when the project was initiated. As the reviewer indicates, AMBER14 and PELE approaches might perhaps be considered obsolescent. Thus, we have predicted the structure of the target using AlphaFold (Jumper J. et al, Nature 2021) and the sequence available under UniProt entry P61073. The complete analysis performed (see our response to point 4) confirmed that the compounds bound the selected pocket, as we had originally determined using PELE. These new analyses have been incorporated into the revised manuscript.

(28)-Discuss the role of the membrane in the receptor-ligand interac7on. Elaborate on how the lipidic double layer may influence the binding of small compounds to GPCRs embedded in the membrane.

Biological membranes are vital components of living organisms, providing a diffusion barrier that separates cells from the extracellular environment, and compartmentalizing specialized organelles within the cell. In order to maintain the diffusion barrier and to keep it electrochemically sealed, a close interaction of membrane proteins with the lipid bilayer is necessary. It is well known that this is important, as many membrane proteins undergo conformational changes that affect their transmembrane regions and that may regulate their activity, as seen with GPCRs (Daemen F.J. & Bonting S.L., Biophys. Struct. Mech. 1977; Gether U. et al. EMBO J. 1997). The lateral and rotational mobility of membrane lipids supports the sealing function while allowing for the structural rearrangement of membrane proteins, as they can adhere to the surface of integral membrane proteins and flexibly adjust to a changing microenvironment. In the case of the first atomistic structure of CXCR4 (Wu B. et al. Science 2010), it was indicated that for dimers, monomers interact only at the extracellular side of helices V and VI, leaving at least a 4-Å gap between the intracellular regions, which is presumably filled by lipids. In particular, they indicated that the channel between TMV and TMVI that connects the orthosteric chemokine binding pocket to the lipid bilayer is occupied by an oleic acid molecule. Recently, Di Marino et al., analyzing the dimeric structure of CXCR4, found a cholesterol molecule placed in between the two protomers, where it engages a series of hydrophobic interactions with residues located in the area between TMI and TMVI (Leu132, Val214, Leu216, Leu246, and Phe249). The polar head of cholesterol forms an H-bond with Tyr135 that further stabilizes its binding mode. This finding confirms that cholesterol might play an important role in mediating and stabilizing receptor dimerization, as seen in other GPCRs (Pluhackova, K., et al. PLoS Comput. Biol. 2016). In addition, we have previously observed that, independently of the structural changes on CXCR4 triggered by lipids, the local lipid environment also regulates CXCR4 organization, dynamics and function at the cell membrane and modulates chemokine-triggered directed cell migration. Prolonged treatment of T cells with bacterial sphingomyelinase promoted the complete and sustained breakdown of sphingomyelins and the accumulation of the corresponding ceramides, which altered both membrane fluidity and CXCR4 nanoclustering and dynamics. Under these conditions, CXCR4 retained some CXCL12-mediated signaling activity but failed to promote efficient directed cell migration (Gardeta S.R. et al. Front. Immunol. 2022). Collectively, these data demonstrate the key role that lipids play in the stabilization of CXCR4 conformations and in regulating its lateral mobility, influencing their associated functions. These considerations have been included in the revised version of the manuscript. 

(29) 3. Stable Trajectories and Binding Mode Superimposi7on -Specify the criteria for defining "stable trajectories" to enhance reader understanding

There could be several ways to describe the stability of a MD simulation, based on the convergence of energies, distances or ligand-target interactions, among others. In this work, we use the expression “stable trajectories” to refer to simulations in which the ligand trajectory converges and the ligand RMSD does not fluctuate more than 0.25Å. This definition is now included in the revised text.

(30)  Clarify the meaning behind superimposing the two small compounds and ensure that the statement in the figure caption aligns with the information presented in the main text.

We apologize for the error in the previous Fig. 5A and in its legend. The figure was created by superimposing the protein component of the poses for the two compounds, AGR1.135 and AGR1.137, rather than the compounds themselves. As panel 5A was confusing, we have modified all Fig. 5 in the revised manuscript to improve clarity.

(31) 4. Volume Analysis and Distances:

-Provide details on how the volume analysis was computed and how distances were accounted for. Consider adding a figure to illustrate these analyses, aiding reader comprehension.

The cleft search and analysis were performed using the default settings of SURFNET (Laskowski R.A. J. Mol. Graph. 1995) included in the PDBsum server (Laskowski R.A. et. al. Trends Biochem. Sci. 1997). The first run of the input model for CXCR4 3ODU identified a promising cleft of 870 Å3 in the lower half of the region flanked by TMV and TMVI, highlighting this area as a possible small molecule binding site (Fig. I, only for review purposes). Analysis of the cleft occupied by AGR1.135 showed two independent cavities of 434 Å3 and 1381 Å3 that were not connected to the orthosteric site. The same procedure for AGR1.137 revealed two distinct clefts of 790 Å3 and 580 Å3, respectively (Fig. I, only for review purposes). Analysis of the atomic distances between the protein residues and the compounds was performed using the PISA server. Krissinel E. & Henrick K. J. Mol. Biol. 2007). (Please see our response to point 3 and the corresponding figure).

(32) 5. Mutant Selection and Relevance:

-Clarify the rationale behind selecting the CXCR4 mutants used in the study. Consider justifying the choice and exploring the possibility of performing an alanine (ALA) scan for a more comprehensive mutational analysis.  

The selection of the residues to be mutated along the cleft was first based on their presence in the proposed cleft and the direct interaction of the compounds with them, either by hydrogen bonding or by hydrophobic interactions. Secondly, all mutated residues did not belong to any of the critical residues involved in transmitting the signal generated by the interaction of CXCL12 with the receptor. In any case, mutants producing a non-functional CXCR4 at the cell membrane were discarded after FACS analysis and chemotaxis experiments. Finally, the length and nature of the resulting mutations were designed mainly to occlude the cleft in case of the introduction of long residues such as lysines (I204K, L208K) or to alter hydrophobic interactions by changing the carbon side chain composition of the residues in the cleft. Indeed, we agree that the alanine scan mutation analysis would have been an alternative strategy to evaluate the residues involved in the interactions of the compounds. 

(33) Reevaluate the statement regarding the relevance of the Y256F muta7on for the binding of AGR1.137. If there is a significant impact on migra7on in the mutant (Fig. 6B), elaborate on the significance in the context of AGR1.137 binding.

In the revised discussion we provide more detail on the relevance of Y256F mutation for the binding of AGR1.137 as well as for the partial effect of G207I and R235L mutations. The predicted interactions for each compound are depicted in new Fig. 6 C, D after LigPlot+ analysis (Laskowski R.A. & Swindells M.B. J. Chem. Inf. Model. 2011), showing that AGR1.135 interacted directly with the receptor through a hydrogen bond with Y256. When this residue was mutated to F, one of the anchor points for the compound was lost, weakening the potential interaction in the region of the upper anchor point.

It is not clear how the Y256F mutation will affect the binding of AGR1.137, but other potential contacts cannot be ruled out since that portion of the compound is identical in both AGR1.135 and AGR1.137. This is especially true for its neighboring residues in the alpha helix, F249, L208, as shown in 3ODU structure (Fig. 6D), which are shown to be directly implicated in the interaction of both compounds. Alternatively, we cannot discard that Y256 interacts with other TMs or lipids stabilizing the overall structure, which could reverse the effect of the mutant at a later stage (Author response image 3).

Author response image 3.

Cartoon representation of Y256 and its intramolecular interactions in the CXCR4 Xray solved structure 3ODU. TMV helix is colored in blue and TMVI in pink.

(34) Address the apparent discrepancy in residue involvement between AGR1.135 and AGR1.137, particularly if they share the same binding mode in the same clef.

AGR1.135 and AGR1.137 exhibit comparable yet distinct binding modes, engaging with CXCR4 within a molecular cavity formed by TMV and TMVI. AGR1.135 binds to CXCR4 through three hydrogen bonds, two on the apical side of the compound that interact with residues TMV-G207 and TMVI-Y256 and one on the basal side that interacts with TMVI-R235 (Fig. 5A). This results in a more extended and rigid conformation when sharing hydrogen bonds, with both TMs occupying a surface area of 400 Å2 and a length of 20 Å in the cleft between TMV and TMVI (Supplementary Fig. 8A). AGR1.137 exhibits a distinct binding profile, interacting with a more internal region of the receptor. This interaction involves the formation of a hydrogen bond with TMIIIV124, which induces a conformational shift in the TMVI helix towards an active conformation (Fig. 5B; Supplementary Fig. 13). Moreover, AGR1.137 may utilize the carboxyl group of V124 in TMIII and overlap with AGR1.135 binding in the cavity, interacting with the other 19 residues dispersed between TMV and VI to create an interaction surface of 370 Å2 along 20 Å (Supplementary Fig. 8B). This is illustrated in the new Fig. 5B. AGR1.137 lacks the phenyl ring present in AGR1.135, resulting in a shorter compound with greater difficulty in reaching the lower part of TMVI where R235 sits. 

Author response image 4.

AGR1.135 and AGR1.137 interaction with TMV and TMVI.  The model shows the location of the compounds within the TMV-VI cleft, illustrated by a ribbon and stick representation. The CXCR4 segments of TMV and TMVI are represented in blue and pink ribbons respectively, and side chains for some of the residues defining the cavity are shown in sticks. AGR1.135 and AGR1.137 are shown in stick representation with carbon in yellow, nitrogen in blue, oxygen in red, and fluorine in green. Hydrogen bonds are indicated by dashed black lines, while hydrophobic interactions are shown in green. The figure reproduces the panels A, B of Fig. 5 in the revised manuscript.

(35) In the Results sec7on regarding "AGR1.137 treatment in a zebrafish xenograf model", the following points can be refined for clarity and completeness: 1. Cell Line Choice for Zebrafish Xenograft Model:

-Explain the rationale behind the choice of HeLa cells for the zebrafish xenograft model when the previous experiments primarily focused on Jurkat cells. Address any specific biological or experimental considerations that influenced this decision.

As far as we know, there are no available models of tumors in zebrafish using Jurkat cells. We looked for a tumoral cell system that expresses CXCR4 and could be transplanted into zebrafish. HeLa cells are derived from a human cervical tumor, express a functional CXCR4, and have been previously used for tumorigenesis analyses in zebrafish (Brown H.K. et al. Expert Opin. Drug Discover. 2017; You Y. et al Front. Pharmacol. 2020). These cells grow in the fish and disseminate through the ventral area and can be used to determine primary tumor growth and metastasis. Nonetheless, we first analyzed in vitro the expression of a functional CXCR4 in these cells (Supplementary Fig. 10A), whether AGR1.137 treatment specifically abrogated CXCL12-mediated direct cell migration (Fig. 7A, B), as whether it affected cell proliferation (Supplementary Fig. 10B). As HeLa cells reproduce the in vitro effects detected for the compounds in Jurkat cells, we used this model in zebrafish. These issues were already discussed in the first version of our manuscript. 

(36) 2. Toxicity Assessment in Zebrafish Embryos: 

-Clarify the basis for stating that AGR1.137 is not toxic to zebrafish embryos. Consider referencing the Zebrafish Embryo Acute Toxicity Test (ZFET) and provide relevant data on lethal concentration (LC50) and non-lethal toxic phenotypes such as pericardial edema, head and tail necrosis, malformation, brain hemorrhage, or yolk sac edema.

Tumor growth and metastasis kinetics within the zebrafish model have been extensively evaluated in many publications (White R. et al. Nat. Rev. Cancer. 2013; Astell K.R. and Sieger D. Cold Spring Harb. Perspect. Med. 2020; Chen X. et al. Front. Cell Dev. Biol. 2021; Weiss JM. Et al. eLife 2022; Lindhal G. et al NPJ Precis. Oncol. 2024). Our previous experience using this model shows that tumors start having a more pronounced proliferation and lower degree of apoptosis from day 4 onwards, but we cannot keep the tumor-baring larvae for that long due to ethical reasons and also because we don’t see much scientific benefit of unnecessarily extending the experiments. Anti-proliferative or pro-apoptotic effects of drugs can still be observed within the three days, even if this is then commonly seen as larger reduction (instead of a smaller growth as it is commonly seen in for example mouse tumor models) compared to controls. Initially we characterized the evolution of implanted tumors in our system and how much they metastasize over time in the absence of treatment before to test the compounds (Author response image 5).

The in vivo experiments were planned to validate efficacious concentrations of the investigated drugs rather than to derive in vivo IC50 or other values, which require testing of multiple doses. We have, however, included an additional concentration to show concentration-dependence and therefore on-target specificity of the drugs in the revised version of the manuscript (data also being elaborated in ongoing experiments). At this stage, we believe that adding the LC50 does not provide interesting new knowledge, and it is standard to only show results from the experimental endpoint (in our case 3 days post implantation). We agree that showing these new data points strengthens the manuscript and facilitates independent evaluation and conclusions to be drawn from the presented data. We have created new graphs where datapoints for each compound dose are shown.  

Author response image 5.

Evolution of the tumors and metastasis along the time in the absence of any treatment. HeLa cells were labeled with 8 µg/mL Fast-DiI™ oil and then implanted in the dorsal perivitelline space of 2-days old zebrafish embryos. Tumors were imaged within 2 hours of implantation and re-imaged each 24 h for three days. Changes in tumor size was evaluated as tumor area at day 1, 2 and 3 divided by tumor area at day 0, and metastasis was evaluated as the number of cells disseminated to the caudal hematopoietic plexus at day 1, 2 and 3 divided by the number of cells at day  3.

Regarding the statement that AGR1.137 was not toxic, this was based on visual inspection of the zebrafish larvae at the end of the experiment, which also revealed a lack of drug-related mortality in these experiments. There are a number of differences in how our experiment was run compared with the standardized ZFET. ZFET evaluates toxicity from 0 hours post-fertilization to 1 or 2 days post-fertilization, whereas here we exposed zebrafish from 2 days post-fertilization to 5 days post-fertilization. The ZFET furthermore requires that the embryos are raised at 26ºC whereas kept the temperature as close as possible to a physiologically relevant temperature for the tumor cells (36ºC). In the ZFET, embryos are incubated in 96-well plates whereas for our studies we required larger wells to be able to manipulate the larvae and avoid well edge-related imaging artefacts, and we therefore used 24-well plates. As such, the ZFET was for various reasons not applicable to our experimental settings. As we were not interested in rigorously determining the LD50 or other toxicity-related measurements, as our focus was instead on efficacy and we found that the targeted dose was tolerated, we did not evaluate multiple doses, including lethal doses of the drug, and are therefore not able to determine an LD50/LC50. We also did not find drug-induced non-lethal toxic phenotypes in this study, and so we cannot elaborate further on such phenotypes other than to simply state that the drug is well tolerated at the given doses. Therefore, the reference to ZFET in the manuscript was eliminated.

(37) If supplementary information is available, consider providing it for a comprehensive understanding of toxicity assessments. 

The effective concentration used in the zebrafish study was derived from the in vitro experiments. That being said, and as elaborated in our response to comment 36, we have added data for one additional dose to show the dose-dependent regulation of tumor growth and metastasis. 

(38) 3. Optimization and Development of AGR1.137: 

-Justify the need for further optimization and development of AGR1.137 if it has a comparable effect to AMD3100. Explain the specific advantages or improvements that AGR1.137 may offer over AMD3100. 

AGR1.137 is highly hydrophobic and is very difficult to handle, particularly in in vivo assays; thus, for the negative allosteric modulators to be used clinically, it would be very important to increase their solubility in water. Contrastingly, AMD3100 is a water-soluble compound. Before using the zebrafish model, we performed several experiments in mice using AGR1.137, but the inhibitory results were highly variable, probably due to its hydrophobicity. We also believe that it would be important to increase the affinity of AGR1.137 for CXCR4, as the use of lower concentrations of the negative allosteric modulator would limit potential in vivo side effects of the drug. On the other hand, we are also evaluating distinct administration alternatives, including encapsulation of the compounds in different vehicles. These alternatives may also require modifications of the compounds. 

AMD3100 is an orthosteric inhibitor and therefore blocks all the signaling cascades triggered by CXCL12. For instance, we observed that AMD3100 treatment blocked CXCL12 binding, cAMP inhibition, calcium flux, cell adhesion and cell migration (Fig. 3, Fig. 4), whereas the effects of AGR1.137 were restricted to CXCL12-mediated directed cell migration. Although AMD3100 was well tolerated by healthy volunteers in a singledose study, it also promoted some mild and reversible events, including white blood cells count elevations and variations of urine calcium just beyond the reported normal range (Hendrix C.W. et al. Antimicrob. Agents Chemother. 2000). To treat viral infections, continuous daily dosing requirements of AMD3100 were impractical due to severe side effects including cardiac arrhythmias (De Clercq E. Front Immunol. 2015). For AMD3100 to be used clinically, it would be critical to control the timing of administration. In addition, side effects after long-term administration have potential problems. Shorter-term usage and lower doses would be fundamental keys to its success in clinical use (Liu T.Y. et al. Exp. Hematol. Oncol. 2016). The use of a negative allosteric modulator that block cell migration but do not affect other signaling pathways triggered by CXCL12 would be, at least in theory, more specific and produce less side effects. These ideas have been incorporated into the revised discussion to reflect potential advantages or improvements that AGR1.137 may offer over AMD3100.

(39) 4. Discrepancy in AGR1.137 and AMD3100 Effects:

-Discuss the observed discrepancy where AGR1.137 exhibits similar effects to AMD3100 but only after 48 hours. Provide insights into the temporal dynamics of their actions and potential implications for the experimental design.

Images and data shown in Fig. 7E, F correspond to days 0 and 3 after HeLa cell implantation (tumorigenesis) and only to day 3 in the case of metastasis data. The revised version contains the effect of two distinct doses of the compounds (10 and 50 µM, for AGR1.135 and AGR1.137 and 1 and 10 µM for AMD3100). 

(40) In the "Discussion" section, there are several points that require clarifica7on and refinement to enhance the overall coherence and depth of the analysis:  1. Reduction of Side-Effects: 

-Provide a more detailed explanation of how the identified compounds, specifically AGR1.135 and AGR1.137, contribute to the reduction of side effects. Consider discussing specific mechanisms or characteristics that differentiate these compounds from existing antagonists.

The sentence indicating that AGR1.135 and AGR1.137 contribute to reduce side effects is entirely speculative, as we have no experimental evidence to support it. We have therefore corrected this in the revised version. The origin of the sentence was that orthosteric antagonists typically bind to the same site as the endogenous ligand, thus blocking its interaction with the receptor. Therefore, orthosteric inhibitors (i.e. AMD3100) block all signaling cascades triggered by the ligand and therefore their functional consequences. However, the compounds described in this project are essentially negative allosteric modulators, that is, they bind to a site distinct from the orthosteric site, inducing a conformational change in the receptor that does not alter the binding of the endogenous ligand, and therefore block some specific receptor-associated functions without altering others. We observed that AGR1.137 blocked receptor oligomerization and directed cell migration whereas CXCL12 still bound CXCR4, triggered calcium mobilization, did not inhibit cAMP release or promoted receptor internalization. This is why we speculated on the limitation of side effects. The statements have been nonetheless revised in the new version of the manuscript.

(41) 2. Binding Site Clarification:

-Address the apparent discrepancy between docking the small compounds in a narrow cleft formed by TMV and TMVI helices and the statement that AGR1.131 binds elsewhere. Clarify the rationale behind this assertion

After the in silico screening, a total of 40 compounds were selected.  These compounds showed distinct degrees of interaction with the cleft formed by TMV and TMVI and even with other potential interaction sites on CXCR4, with the exception of the ligand binding site according to the data described by Wescott et al. (PNAS 2016 113:9928-9933), as this possibility was discarded in the initial approach of the in silico screening. According to PELE analysis, AGR1.131 was one of the 40 selected compounds that showed a pose with low binding energy, -39.8 kcal/mol, between TMV and TMVI helices, that is, it might interact with CXCR4 through the selected area for the screening. It nonetheless also showed a best pose placed between helices TMI and TMVII, -43.7 kcal/mol. In any case, the compound was included in the biological screening, where it was unable to impact CXCL12-mediated chemotaxis (Fig. 1B). We then focused on AGR1.135 and AGR1.137, as showed a higher inhibitory effect on CXCL12-mediated migration, and on AGR1.131 as an internal negative control. AGR1.131 has a skeleton very similar to the other compounds (Fig. 1C) and can interact with the TM domains of CXCR4 without promoting effects. None of the three compounds affected CXCL12 binding, or CXCL12mediated inhibition of cAMP release, or receptor internalization. However, whereas AGR1.135 and AGR1.137, blocked CXCL12-mediated CXCR4 oligomerization and directed cell migration towards CXCL12 gradients, AGR1.131 had no effect in these experiments (Fig. 3, Fig.  4). 

Next, we performed additional theoretical calculations (PELE, docking, MD) to inspect in detail the potential binding modes of active and inactive molecules. Based on these additional calculations, we identified that whereas AGR1.135 and AGR1.137 showed preferent binding on the molecular pocket between TMV and TMVI, the best pose for AGR1.131 was located between TMI and TMVII, as the initial experiments indicated.  These observations and data have been clarified in the revised discussion. 

(42) 3. Impact of Chemical Modifications:

-Discuss the consequences of the distinct chemical groups in AGR1.135, AGR1.137, and AGR1.131, specifically addressing how variations in amine length and chemical nature may influence binding affinity and biological activity. Provide insights into the potential effects of these modifications on cellular responses and the observed outcomes in zebrafish. 

The main difference between AGR1.131 and the other two compounds is the higher flexibility of AGR1.131 due to the additional CH2 linker, together with the lack of a piperazine ring. The additional CH2 linking the phenyl ring increases the flexibility of AGR1.131 when compared with AGR1.135 and AGR1.137, and the absence of the piperazine ring might be responsible for its lack of activity, as it makes this compound able to bind to CXCR4 (Fig. 1C).

AGR1.137 was chosen in a second round. The additional presence of the tertiary amine (in the piperazine ring) allows the formation of quaternary ammonium salts in the aqueous medium and its substituents to increase its solubility (Fig 1C). This characteristic might be related to the absence of toxic effects of the compound in the zebrafish model.

(43) 4. Existence of Distinct CXCR4 Conformational States: 

-Provide more detailed support for the statement suggesting the "existence of distinct CXCR4 conformational states" responsible for activating different signaling pathways. Consider referencing relevant studies or experiments that support this claim.

Classical models of GPCR allostery and activation, which describe an equilibrium between a single inactive and a single signaling-competent active conformation, cannot account for the complex pharmacology of these receptors. The emerging view is that GPCRs are highly dynamic proteins, and ligands with varying pharmacological properties differentially modulate the balance between multiple conformations.

Just as a single photograph from one angle cannot capture all aspects of an object in movement, no one biophysical method can visualize all aspects of GPCR activation. In general, there is a tradeoff between high-resolution information on the entire protein versus dynamic information on limited regions. In the former category, crystal and cryo-electron microscopy (cryoEM) structures have provided comprehensive, atomic-resolution snapshots of scores of GPCRs both in inactive and active conformations, revealing conserved conformational changes associated with activation. However, different GPCRs vary considerably in the magnitude and nature of the conformational changes in the orthosteric ligand-binding site following agonist binding (Venkatakrishnan A.J.V. et al. Nature 2016). Spectroscopic and computational approaches provide complementary information, highlighting the role of conformational dynamics in GPCR activation (Latorraca N.R.V. et al. Chem. Rev 2017). In the absence of agonists, the receptor population is typically dominated by conformations closely related to those observed in inactive-state crystal structures (Manglik A. et al. Cell 2015). While agonist binding drives the receptor population towards conformations similar to those in activestate structures, a mixture of inactive and active conformations remains, reflecting “loose” or incomplete allosteric coupling between the orthosteric and transducer pockets (Dror R.O. et al. Proc. Natl. Acad. Sci. USA 2011). Surprisingly, for some GPCRs, and under some experimental conditions, a substantial fraction of unliganded receptors already reside in an active-like conformation, which may be related to their level of basal or constitutive signaling (Staus D.P. et al. J. Biol. Chem. 2019);  Ye L. et al. Nature 2016).  In our case, the negative allosteric modulators, (Staus DP, et al. J. Biol. Chem 2019); Ye L. et al. Nature 2016) did not alter ligand binding and had only minor effects on specific CXCL12-mediated functions such as inhibition of cAMP release or receptor internalization, among others, but failed to regulate CXCL12-mediated actin dynamics and receptor oligomerization. Collectively, these data suggest that the described compounds alter the active conformation of CXCR4 and therefore support the presence of distinct receptor conformations that explain a partial activation of the signaling cascade.

All these observations are now included in the revised discussion of the manuscript.

(44) 5. Equilibrium Shift and Allosteric Ligands: 

-Clarify the statement about "allosteric ligands shifting the equilibrium to favor a particular receptor conformation". Support this suggestion with references or experimental evidence

In a previous answer (see our response to point 2), we explain why we define the compounds as negative allosteric modulators. These compounds do not bind the orthosteric binding site or a site distinct from the orthosteric site that alters the ligand-binding site. Their effect should be due to changes in the active conformation of CXCR4, which allow some signaling events whereas others are blocked. Our functional data thus support that through the same receptor the compounds separate distinct receptor-mediated signaling cascades, that is, our data suggest that CXCR4 has a conformational heterogeneity. It is known that GPCRs exhibit more than one “inactive” and “active” conformation, and the endogenous agonists stabilize a mixture of multiple conformations. Biased ligands or allosteric modulators can achieve their distinctive signaling profiles by modulating this distribution of receptor conformations. (Wingler L.M. & Lefkowitz R.J. Trends Cell Biol. 2020). For instance, some analogs of angiotensin II do not appreciably activate Gq signaling (e.g., increases in IP3 and Ca2+) but still induce receptor phosphorylation, internalization, and mitogen-activated protein kinase (MAPK) signaling (Wei H, et al. Proc. Natl. Acad. Sci. USA 2003). Some of these ligands activate Gi and G12 in bioluminescence resonance energy transfer (BRET) experiments (Namkung Y. et al. Sci. Signal. 2018). A similar observation was described in the case of CCR5, where some chemokine analogs promoted G protein subtype-specific signaling bias (Lorenzen E. et al. Sci. Signal 2018). Structural analysis of distinct GPCRs in the presence of different ligands vary considerably in the magnitude and nature of the conformational changes in the orthosteric ligand-binding site following agonist binding (Venkatakrishnan A.J.V. et al. Nature 2016). Yet, these changes modify conserved motifs in the interior of the receptor core and induce common conformational changes in the intracellular site involved in signal transduction. That is, these modifications might be considered distinct receptor conformations. 

The revised discussion contains some of these interpretations to support our statement about the stabilization of a particular receptor conformation triggered by the negative allosteric modulators. 

(45) 6. Refinement of Binding Mode: 

-Clarify the workflow for obtaining the binding mode, particularly the role of GLIDE and PELE. Clearly explain how these software tools were used in tandem to refine the binding mode. 

The computational sequential workflow applied in this project included, i) Protein model construction, ii) Virtual screening (Glide), iii) PELE, iv) Docking (AutoDock and Glide) and v) Molecular Dynamics (AMBER).

Glide was applied for the structure-based virtual screening to explore which compounds could fit and interact with the previously selected binding site.

After the identification of theoretically active compounds (modulators of CXCR4), additional calculations were done to identify a potential binding site. PELE was used in this sense, to study how the compounds could bind in the whole surface of the target (TMV-TMVI). By applying PELE, we avoided biasing the calculation, and we found that the trajectories with better interaction energies identified the cleft between TMV and TMVI as the binding site for AGR1.135 and AGR1.137, and not for AGR1.131. AGR1.131 showed a pose with low binding energy, -39.8 kcal/mol, between TMV and TMVI helices, that is, it might interact with CXCR4 in the selected area for the screening. But it also showed a better pose placed between helices TMI and TMVII, - 43.7 kcal/mol (see our response to point 41). These data have been now confirmed using Schrodinger’s MM-GBSA procedure (see our response to points 6 and 8). In any case, the compound was included in the biological screening, where it was unable to affect CXCL12-mediated chemotaxis (Fig. 1B). Docking and MD simulations were then performed to study and refine the specific binding mode in this cavity. These data were important to choose the mutations on CXCR4 required, to test whether the compounds reversed its behavior. In these experiments we also confirmed that AGR1.131 had a better pose on the TMI-TMVII region. 

(46) 7. Impact of Compound Differences on CXCR4-F249L mutant: 

-Provide visual aids, such as figures, and additional experiments to support the statement about differences in the behavior of AGR1.135 and AGR1.137 on cells expressing CXCR4-F249L mutant. Elaborate on the closer interaction suggested between the triazole group of AGR1.137 and the F249 residue

At the reviewer’s suggestion, Fig. 5 has been modified to incorporate a closer view of the interactions identified and new panels in new Fig. 6 have been added to show in detail the effect of the mutations selected on the structure of the cleft between TMV and TMVI. The main difference between AGR1.135 and AGR1.137 is how the triazole group interacts with F249 and L216 (Author response image 6). In AGR1.137, the three groups are aligned in a parallel organization, which appears to be more effective: This might be due to a better adaptation of this compound to the cleft since there is only one hydrogen bond with V124. In AGR1.135, the compound interacts with the phenyl ring of F249 and has a stronger interaction at the apical edge to stabilize its position in the cleft. However, there is still an additional interaction present. When changing F249

Author response image 6.

Cartoon representation of the interaction of CXCR4 F249L mutant with AGR1.135 (A) and AGR1.137 (B). The two most probable conformations of Leucine rotamers are represented in cyan A and B conformations. Van der Waals interactions are depicted in blue cyan dashed lines, hydrogen bonds in black dashed lines. CXCR4 segments of TMV and TMVI are colored in blue and pink, respectively

to L (Fig. VIIA, B, only for review purposes) and showing the two most likely rotamers resulting from the mutation, it is observed that rotamer B is in close proximity to the compound, which may cause the binding to either displace or adopt an alternative conformation that is easier to bind into the cleft. As previously mentioned, it is likely that AGR1.135 can displace the mutant rotamer and bind into the cleft more easily due to its higher affinity.

(47) In the "Materials and Methods" section, the computational approach for the "discovery of CXCR4 modulators" requires significant revision and clarification. The following suggestions aim to address the identified issues: 1. Structural Modeling: 

-Reconsider the use of SWISS-MODEL if there is an available PDB code for the entire CXCR4 structure. Clearly articulate the rationale for choosing one method over the other and explain any limitations associated with the selected approach. 

The SWISS-model server allows for automated comparative modeling of 3D protein structures that was pioneered in the fields of automated modeling. At the time we started this project. it was the most accurate method to generate reliable 3D protein structure models.

As explained above, we have now predicted the structure of the target using AlphaFold (Jumper J. et al, Nature 2021) and performed several additional experiments that confirm that the small compounds bind the selected pocket as the original strategy indicated (see our response to point 6). (Fig. II, only for review purposes).

(48) 2. Parametriza7on of Small Compounds: 

-Provide a detailed description of the parametrization process for the small compounds used in the study. Specify the force field and parameters employed, considering the obsolescence of AMBER14 and ff14SB. Consider adopting more contemporary force fields and parameterization strategies. 

When we performed these experiments, some years ago, the force fields applied (ff14SB, AMBER14 used in MD or OPLS2004 in docking with Glide) were well accepted and were gold standards. It is, however, true that the force fields have evolved in the past few years, Moreover, in the case of the MD simulations, to consider the parameters of the ligands that are not contained within the force field, we performed an additional parameterization as a standard methodology. We then generated an Ab initio optimization of the ligand geometry, defining as basis sets B3LYP 6-311+g(d), using Gaussian 09, Revision A.02, and then a single point energy calculation of ESP charges, with HF 6311+g(d) on the optimized structure. As the last step of the parametrization, the antechamber module was used to adapt these charges and additional parameters for MD simulations.

(49) 3. Treatment of Lipids and Membrane: 

-Elaborate on how lipids were treated in the system. Clearly describe whether a membrane was included in the simulations and provide details on its composition and structure. Address the role of the membrane in the study and its relevance to the interactions between CXCR4 and small compounds 

To stabilize CXCR4 and more accurately reproduce the real environment in the MD simulation, the system was embedded in a lipid bilayer using the Membrane Builder tool (Sunhwan J. et al. Biophys. J. 2009) from the CHARMM-GUI server. The membrane was composed of 175 molecules of the fatty acid 1-palmitoyl-2-oleoyl-sn-glycero-3phosphocholine (POPC) in each leaflet. The protein-membrane complex was solvated with TIP3 water molecules. Chloride ions were added up to a concentration of 0.15 M in water, and sodium ions were added to neutralize the system. This information was previously described in detail.

(50) 4. Molecular Dynamics Protocol: 

-Provide a more detailed and coherent explanation of the molecular dynamics protocol. Clarify the specific steps, parameters, and conditions used in the simulations. Ensure that the protocol aligns with established best practices in the field.

Simulations were calculated on an Asus 1151 h170 LVX-GTX-980Ti workstation, with an Intel Core i7-6500 K Processor (12 M Cache, 3.40 GHz) and 16 GB DDR4 2133 MHz RAM, equipped with a Nvidia GeForce GTX 980Ti available for GPU (Graphics Processing Unit) computations. MD simulations were performed using AMBER14 (Case D.A. et al. AMBERT 14, Univ. of California, San Francisco, USA, 2014) with ff14SB (Maier J.A. et al. J. Chem. Theory Comput. 2015) and lipid14 (Dickson C. J. et al. J. Chem. Theory Comput. 2014) force fields in the NPT thermodynamic ensemble (constant pressure and temperature). Minimization was performed using 3500 Steepest Descent steps and 4500 Conjugate Gradient steps three times, firstly considering only hydrogens, next considering only water molecules and ions, and finally minimizing all atoms. Equilibration raises system temperature from 0 to 300 K at a constant volume fixing everything but ions and water molecules. After thermalization, several density equilibration phases were performed. In the production phase, 50 ns MD simulations without position restraints were calculated using a time step of 2 fs. Trajectories of the most interesting poses were extended to 150 ns. All bonds involving hydrogen atoms were constrained with the SHAKE algorithm (Lippert R.A. et al. J. Chem. Phys. 2007). A cutoff of 8 Å was used for the Lennard-Jones interaction and the short-range electrostatic interactions. Berendsen barostat (Berendsen H.J. et al. J. Chem. Phys.  1984) and Langevin thermostat were used to regulate the system pression and temperature, respectively. All trajectories were processed using CPPTRAJ (Roe D.R. & Cheatham III T.E. J. Chem. Theory Comput. 2013) and visualized with VMD (Visual Molecular Dynamics) (Humphrey W. et al. J. Mol. Graphics. 1996). To reduce the complexity of the data, Principal Component Analysis (PCA) was performed on the trajectories using CPPTRAJ.

(51) Consider updating the molecular dynamics protocol to incorporate more contemporary methodologies, considering advancements in simulation techniques and software.

In our answer to points 6 and 47, we describe why we use the technology based on Swiss-model and PELE analysis and how we have now used Alphafold and other more contemporary methodologies to confirm that the small compounds bind the selected pocket.

(52) Figure 1A: 

•  Consider switching to a cavity representation for CXCL12 to enhance clarity and emphasize the cleft.

Fig. 1A has been modified to emphasize the cleft.

(53) Explicitly show the TMV-TMVI cleft in the figure for a more comprehensive visualization. 

In Fig. 1A we have added an insert to facilitate TMV-TMVI visualization.

(54) Figure 1B: 

•  Clearly explain the meaning of the second DMSO barplot to avoid confusion. 

To clarify this panel, we have modified the figure and the figure legend. Panel B now includes a complete titration of the three compounds analyzed in the manuscript.  The first bar shows cell migration in the absence of both treatment with AMD3100 and stimulation with CXCL12.  The second bar shows migration in response to CXCL12 in the absence of AMD3100. The third bar shows the effect of AMD3100 on CXCL12-induced migration, as a known control of inhibition of migration.  We hope that this new representation of the data results is clearer.

(55) Figure 1C: 

•  Provide a clear legend explaining the significance of the green shading on the small compounds. 

The legend for Fig. 1C has been modified accordingly to the reviewer’s suggestion.

(56) Figure 2: 

•  Elaborate on the role of fibronectin in the experiment and explain the specific contribution of CD86-AcGFP.

The ideal situation for TIRF-M determinations is to employ cells on a physiological substrate complemented with or without chemokines. Fibronectin is a substrate widely used in different studies that allows cell adhesion, mimicking a physiological situation. Jurkat cells express alpha4beta1 and alpha5beta1 integrins that mediate adhesion to fibronectin (Seminario M.C. et al. J. Leuk. Biol. 1999).

Regarding the use of CD86-AcGFP in TIRF-M experiments. We currently determine the number of receptors in individual trajectories of CXCR4 using, as a reference, the MSI value of CD86-AcGFP that strictly showed a single photobleaching step (Dorsch S. et al. Nat Methods 2009).

We preferred to use CD86-AcGFP in cells instead of AcGFP on glass, to exclude any potential effect on the different photodynamics exhibited by AcGFP when bound directly to glass. In any case, this issue has been clarified in the revised version.

(57) Figure 3D: 

•  Include a plot for the respective band intensity to enhance data presentation 

The plot showing the band intensity analysis of the experiments shown in Fig. 3D was already included in the original version (see old Supplementary Fig. 3). However, in the revised version, we include these plots in the same figure as panels 3E and 3F.  As a control of inhibition of CXCL12 stimulation, we have also included a new figure (Supplementary Fig. 4) showing the effect of AMD3100 on CXCL12-induced activation of Akt and ERK as analyzed by western blot.

(58) Consider adding AMD3100 as a control for comparison. 

In agreement with the reviewer’s suggestion, we have added the effect of AMD3100 in most of the functional experiments performed.

(59) Figure 4: 

•  Address the lack of positive controls in Figure 4 and consider their inclusion for a more comprehensive analysis. 

DMSO bars correspond to the control of the experiment, as they represent the effect of CXCL12 in the absence of any allosteric modulator. As previously described in this point-by-point reply, DMSO bars correspond to the control performed with the solvent with which the small compounds, at maximum concentration, are diluted.  Therefore, they show the effect of the solvent on CXCL12 responses. In any case, and in order to facilitate the comprehension of the figure we have also added the controls in the absence of DMSO to demonstrate that the solvent does not affect CXCL12-mediated functions, together with the effect of the orthosteric inhibitor AMD3100. In addition, we have also included representative images of the effect of the different compounds on CXCL12-induced polarization (Fig. 4C).

(60) In Figure 4A, carefully assess overlapping error bars and ensure accurate interpreta7on. If necessary, consider alternative representation. 

We have tried alternative representations of data in Fig. 4A, but in all cases the figure was unclear. We believe that the way we represent the data in the original manuscript is the most clear and appropriate.  Nevertheless, we have now included significance values as a table annexed to the figure, as well as the effect of AMD3100, as a control of inhibition

(61) Supplementary Figure 1A: 

•  Improve the clarity of bar plots for better understanding. Consider reordering them from the most significant to the least. 

This was a good idea, and therefore Supplementary Fig. 1A has been reorganized to improve clarity.

(62) Supplementary Figure 1C: 

•  Clarify the rationale behind choosing the 12.5 nM concentration and explain if different concentrations of CXCL12 were tested. 

In old Supplementary Fig. 1C, we used untreated cells, that is, CXCL12 was not present in the assay.  These experiments were performed to test the potential toxicity of DMSO (solvent) or the negative allosteric modulators on Jurkat cells. The 12.5 nM concentration of CXCL12 mentioned in the figure legend applied only to panels A and B, as indicated in the figure legend. We previously optimized this concentration for Jurkat cells using different concentrations of CXCL12 between 5 and 100 nM.  Nevertheless, we have reorganized old supplementary fig. 1 and clarified the figure legend to avoid misinterpretations (see Supplementary Fig 1A, B and Supplementary Fig. 2A, B).

(63) Explain the observed reduction in fluorescence intensity for AGR1.135. 

The cell cycle analysis has been moved from Supplementary Fig. 1C to a new Supplementary Fig. 2.  It now includes the flow cytometry panels to show fluorescence intensity as a function of the number of cells analyzed (Panel 1A) as well as a table (panel B) with the percentage of cells in each phase of the cell cycle. We believe that the apparent reduction in fluorescence that the reviewer observes is mainly due to the number of events analyzed. However, we have changed the flow cytometry panels for others that are more representative and included a table with the mean of the different results. When we determined the percentage of cells in each cell cycle phase, we observed that it looks very similar in all the experimental conditions. That is, none of the compounds affected any of the cell cycle phases. We have also included the effect of H2O2 and staurosporine as control compounds inducing cell death and cell cycle alteration of Jurkat cells.

(64) Supplementary Table 1: 

•  Include a column specifying the scoring for each compound to provide a clear reference for readers. 

To facilitate references to readers, we have now included the inhibitory effect of each compound on Jurkat cell migration in the revised version of this table. 

(65) Minor Points 

Page 2 - Abstract: Rephrase the first sentence of the abstract to enhance fluidity. 

Although the entire manuscript was revised by a professional English editor, we appreciate the valuable comments of this reviewer and we have corrected these issues accordingly.

(66) Page 2 - Abstract: Explicitly define "CXCR4" as "C-X-C chemokine receptor type 4" the first time it appears.

We have not used C-X-C chemokine receptor type 4 the first time it appears in the abstract. CXCR4 is an acronym normally accepted to identify this chemokine receptor, and it is used as CXCR4 in many articles published in eLife. However, we introduce the complete name the first time it appears in the introduction.

(67) Page 2 - Abstract: Explicitly define "CXCL12" as "C-X-C motif chemokine 12" the first time it is mentioned. 

As we have discussed in the previous response, we have not used C-X-C motif chemokine 12 the first time CXCL12 appears in the abstract, as it is a general acronym normally accepted to identify this specific chemokine, even in eLife papers. However, we introduce the complete name the first time it appears in the introduction section.

(68) Page 2 - Abstract: Explicitly define "TMV and TMVI" upon its first mention.

The acronym TM has been defined as “Transmembrane” in the revised version

(69) Page 2 - Abstract: Review the use of "in silico" in the sentence for accuracy and consider revising if necessary.

With the term “in silico” we want to refer to those experiments performed on a computer or via computer simulation software. We have carefully reviewed its use in the new version of the manuscript.

(70) Page 2 - Abstract: Add a comma after "compound" in the sentence, "We identified AGR1.137, a small compound that abolishes...".

A comma after “compound” has been added in the revised sentence.

(71) Page 2 - Significance Statement: Rephrase the first sentence of the "Significance Statement" to avoid duplication with the abstract.

The first sentence of the Significance Statement has been revised to avoid duplication with the abstract. 

(72) Page 2 - Significance Statement: Break down the lengthy sentence, "Here, we performed in silico analyses..." for better readability. 

The sentence starting by “Here, we performed in silico analyses…” has been broken down in the revised manuscript.

(73) Page 2 - Introduction: Replace "Murine studies" with a more specific term for clarity.

The term “murine studies” is normally used to refer to experimental studies developed in mice. We have nonetheless rephrased the sentence.

(74) Page 3 - Introduction: Rephrase the sentence for clarity: "Finally, using a zebrafish model, ..."

The sentence has been now rephrased for clarity.

(75) Results-AGR1.135 and AGR1.137 block CXCL12-mediated CXCR4 nanoclustering and dynamics: 

Rephrase the sentence for clarity: "Retreatment with AGR1.135 and AGR1.137, but not with AGR1.131, substantially impaired CXCL12-mediated receptor nanoclustering.”

The sentence has been rephrased for clarity.

(76) Results - AGR1.135 and AGR1.137 incompletely abolish CXCR4-mediated responses in Jurkat cells: Clarify the sentence: "In contrast to the effect promoted by AMD3100, a binding-site antagonist of CXCR4..."

The sentence has been modified for clarity.

(77) Consider using "orthosteric" instead of "binding-site" antagonist.

The term orthosteric is now used throughout to refer to a binding site antagonist.

(78) Discussion: Use the term "in silico" only when necessary.

We have carefully reviewed the use of “in silico” in the manuscript.

(79) Discussion: Clarify the sentence: "...not affect neither CXCR2-mediated cell migration...". Confirm if "CXCL12" is intended.

The sentence refers to the chemokine receptor CXCR2, which binds the chemokine CXCL2. To test the specificity of the compounds for the CXCL12/CXCR4 axis, we evaluated CXCL2-mediated cell migration.  The results indicated that CXCL2/CXCR2 axis was not affected by the negative allosteric modulators, whereas CXCL12-mediated cell migration was blocked.  The sentence has been clarified in the new version of the manuscript.

(80) Figure 4B: Bold the "B" in the figure label for consistency.

The “B” in Fig. 4B has been bolded.

Reviewer #2

(1) Fig 2. The SPT data is sub-optimal in its presentation as well as analysis. Example images should be shown. The analysis and visualization of the data should be reconsidered for improvements. Graphs with several hundreds, in some conditions over 1000 tracks, per condition are very hard to compare. The same (randomly selected representative set) number of data points should be shown for better visualization. Also, more thorough analyses like MSD or autocorrelation functions are lacking - they would allow enhanced overall representation of the data.

In agreement with the reviewer’s commentary, we have modified the representation of Fig. 2. We have carefully read the paper published by Lord S.J. and col. (Lord S. J. et al., J. Cell Biol. 2020) and we apply their recommendations for these type of data. We have also included as supplementary material representative videos for the TIRF-M experiments performed to allow readers to visualize the original images. Regarding the MSD analyses, they were developed to determine all D1-4 values. According to the data published by Manzo & García-Parajo (Manzo C. & García-Parajo M.F. Rep.Prog. Phys. 2015) due to the finite trajectory length the MSD curve at large tlag has poor statistics and deviates from linearity. However, the estimation of the Diffusion Coefficient (D1-4) can be obtained by fitting of the short tlag region of the MSD plot giving a more accurate idea of the behavior of particles. In agreement we show D1-4 values and not MSD data. 

Due to the space restrictions, it is very difficult to include all the figures generated, but, only for review purposes, we included in this point-by-point reply some representative plots of the MSD values as a function of the time from individual trajectories showing different types of motion obtained in our experiments (Author response image 7).

Author response image 7.

Representative MSD plots from individual trajectories of CXCR4-AcGFP showing different types of motion: A) confined, B) Brownian/Free, C) direct transport of CXCR4-AcGFP particles diffusing at the cell membrane detected by SPT-TIRF in resting JKCD4 cells.

Further analysis, such as the classification based on particle motion, has not been included in this article. This classification uses the moment scaling spectrum (MSS), described by Ewers H. et al. 2005 PNAS, and requires particles with longer trajectories (>50 frames). Only for review purposes, we include a figure showing the percentage of the MSS-based particle motion classification for each condition. As expected, most of long particles are confined, with a slight increase in the percentage upon CXCL12 stimulation in all conditions, except in cell treated with AGR1.137 (Author response image 8).

Author response image 8.

Effects of the negative allosteric modulators on the Types of Motion of CXCR4. Percentage of single trajectories with different types of motion, classified by MSS (DMSO: 58 particles in 59 cells on FN; 314 in 63 cells on FN+CXCL12; AGR1.131: 102 particles in 71 cells on FN; 258in 69 cells on FN+CXCL12; AGR1.135: 86 particles in 70 cells on FN; 120 in 77 cells on FN+CXCL12; AGR1.137: 47 particles in 66 cells on FN; 74 in 64 cells on FN+CXCL12) n = 3.

(2) Fig 3. The figure legends have inadequate information on concentrations and incubation times used, both for the compounds and other treatments like CXCL12 and forskolin. For the Western blot data, also the quantification should be added to the main figure. The compounds, particularly AGR1.137 seem to lead to augmented stimulation of pAKT and pERK. This should be discussed

The Fig. 3 legend has been corrected in the revised manuscript. Fig. 3D now contains representative western blots and the densitometry evaluation of these experiments. As the reviewer indicates, we also detected in the western blot included, augmented stimulation of pAKT and pERK in cells treated with AGR1.137. However, as shown in the densitometry analysis, no significant differences were noted between the data obtained with each compound. As a control of inhibition of CXCL12 stimulation we have included a new Supplementary Fig. 4 showing the effect of AMD3100 on CXCL12-induced activation of Akt and ERK as analyzed by western blot.

(3) Fig. 4 immunofluorescence data on polarization as well as the flow chamber data lack the representative images of the data. The information on the source of the T cells is missing. Not clear if this experiment was done on bilayers or on static surfaces.

Representative images for the data shown in Figure 4B have been added in the revised figure (Fig. 4C). The experiments in Fig. 4B were performed on static surfaces. As indicated in the material and methods section, primary T cell blasts were added to fibronectin-coated glass slides and then were stimulated or not with CXCL12 (5 min at 37ºC) prior to fix permeabilize and stain them with Phalloidin. Primary T cell blasts were generated from PBMCs isolated from buffy coats that were activated in vitro with IL-2 and PHA as indicated in the material and methods section.

(4) The data largely lacks titration of different concentrations of the compounds. How were the effective concentration and treatment times determined? What happens at higher concentrations? It is important to show, for instance, if the CXCR12 binding gets inhibited at higher concentrations. most experiments were performed with 50 uM, but HeLa cell data with 100 uM. Why and how was this determined? 

The revised version contains a new panel in Fig. 1B to show a more detailed kinetic analysis with different concentrations (1-100 µM) of the compounds in the migration experiments using Jurkat cells. We choose 50 µM for further studies as it was the concentration that inhibits 50-75% of the ligand induced cell migration. 

We have also included the effect of two doses of the compounds (10 and 50 µM) in the zebrafish model as well as AMD3100 (1 and 10 µM) as control (new Fig. 7D, E).  Tumors were imaged within 2 hours of implantation and tumor-baring embryos were treated with either vehicle (DMSO) alone, AGR1.131 or AGR1.137 at 10 and 50 µM or AMD3100 at 1 and 10 µM for three days, followed by re-imaging.

Regarding the amount of CXCL12 used in these experiments, with the exception of cell migration assays in Transwells, where the optimal concentration was established at 12.5 nM, in all the other experiments the optimal concentration of CXCL12 employed was 50 nM. In the case of the directional cell migration assays, we use 100 nM to create the chemokine gradient in the device. These concentrations have been optimized in previous works of our laboratory using these types of experiments. It should also be noted that in the experiments using lipid bilayers or TIRF-M experiments, CXCL12 is used to coat the plates and therefore it is difficult to determine the real concentration that is retained in the surface after the washing steps performed prior adding the cells.

(5) The authors state that they could not detect direct binding of the compounds and the CXCR14. It should be reported what approaches were tried and discussed why this was not possible. 

We attempted a fluorescence spectroscopy strategy to formally prove the ability of AGR1.135 to bind CXCR4, but this strategy failed because the compound has a yellow color that interfered with the determinations. We also tried a FRET strategy (see supplementary Fig. 7) and detected a significant increase in FRET efficiency of CXCR4 homodimers in cells treated with AGR1.135; this effect was due to the yellow color of this compound that interferes with FRET determinations. In the same assays, AGR1.137 did not modify FRET efficiency for CXCR4 homodimers and therefore we cannot assume that AGR1.137 binds on CXCR4. All these data have been considered in the revised discussion.

(6) The proliferation data in Supplementary Figure 1 lacks controls that affect proliferation and indication of different cell cycle stages. What is the conclusion of this data? More information on the effects of the drug to cell viability would be important.

Toxicity in Jurkat cells was first determined by propidium iodide incorporation. Some compounds (i.e., AGR1.103 and VSP3.1) were discarded from further analysis as they were toxic for cells. In a deeper analysis of cell toxicity, even if these compounds did not kill the cells, we checked whether they could alter the cell cycle of the cells. New Supplementary Fig. 2 includes a table (panel B) with the percentage of cells in each cell cycle phase, and no differences between any of the treatments tested were detected. 

Nevertheless, to clarify this issue the revised version of the figure also includes H2O2 and staurosporine stimuli to induce cell death and cell cycle alterations as controls of these assays.

(7) The flow data in Supplementary Figure 2 should be statistically analysed. 

Bar graphs corresponding to the old Supplementary Fig. 2 (new Supplementary Fig. 3) are shown in Fig. 3B. We have also incorporated the corresponding statistical analysis to this figure. 

(8) In general, the authors should revise the figure legends to ensure that critical details are added. 

We have carefully revised all the figure legends in the new version of the manuscript.

(9) Bar plots are very poor in showing the heterogeneity of the data. Individual data points should be shown whenever feasible. Superplot-type of representation is strongly advised (https://doi.org/10.1083/jcb.202001064).

We have carefully read the paper published by Lord S.J. and col. (Lord S. J. et al., J. Cell Biol. 2020) and we apply their recommendations for our TIRF-M data (see revised Fig.  2).

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Recommendations For The Authors): 

- The title may not reflect the key finding of the paper. It is well established in the field that the disaggregation process is sensitive to perturbations of the levels of the disaggregating factors.

We have changed the title to better reflect the major finding of the work, the importance of the NEF during the initiation of disaggregation. The new title is: Early Steps of Protein Disaggregation by Hsp70 Chaperone and Class B J-Domain Proteins are Shaped by Hsp110.

- Abstract:

Please note that the phrases "stimulation is much limited with class A JDPs", "limited destabilization of the chaperone complex improves disaggregation", and "tuned proportion between the co-chaperones" are hard to understand. Only after having read the manuscript are the meanings of these phrases accessible.

The phrases in the abstract were changed (page 1, lines 10-14).

- The subheading "Sse1 improves aggregate modification by Hsp70" on p. 7 is unclear. What is measured is a decrease in aggregate size dependent on Hsp70-JDP as well as Sse1.

The subheading was changed to include more precise information, into “Sse1 leads to Hsp70-depenent reduction of aggregate size”.

- The subheading "Biphasic effects of Sse1 on the Hsp70 disaggregation activity" does not describe the finding clearly; "Biphasic effects" is a term that is hard to understand.

To avoid phrases that can be understood in many ways, we have changed the subheading into “Hormetic effects of Sse1 in Hsp70 disaggregation activity”

- p.5, last line. Hsp110 typo The typos have been corrected.

Reviewer #2 (Recommendations For The Authors):

(1) The article emphasises multiple times the importance of stoichiometry between the (co-)chaperones. Most figures would benefit from an indication of the used stoichiometry (or all absolute concentrations) to support the points made about the stoichiometry, especially the figures showing titrations of Sse1, Sse1-2, and Sis1 (Fig. 3D, 3E, 4A-C, S2B, S5F, S6A-E).

The information of protein concentrations has been included in all figure captions.

(2) The manuscript includes a summary model. While this model is a plausible hypothesis of the mechanism of disaggregation by Hsp70, in particular when viewed with previous data (Wyszkowski et al., 2021), it focuses rather heavily on the potential remodeling of clients by Hsp70, which is not the primary focus of the data presented in this manuscript. More emphasis could be put on the JDP class/ functional specificity observed.

The model has been changed according to the Reviewer’s comments to better reflect the findings presented in the manuscript (Figure 5).

(3) The methods section is very brief. I recommend including additional details about reaction conditions (temperature, buffer compositions, protein concentrations) even when previously reported elsewhere to improve the readability of the manuscript. Details regarding the DLS experiments performed are missing.

More detailed information on the experimental conditions has been added to the Methods section, as well as to figure legends.

(4) Many experiments incorporate BLI to assess the effect of NEFs on the binding of the Hsp70 and JDP to aggregates. Although appropriate controls are included (no ATP, Hsp70, and JDP only), a control with only Hsp70 and the NEF would be useful to determine to which extent the NEF itself alters the thickness of the (Hsp70-bound) aggregate biolayer.

The suggested controls were added (Figure 1—figure supplement 1 G) and discussed in the manuscript (page 5, lines 23-24).

Reviewer #3 (Recommendations For The Authors):

- The refolding assay makes use of Luciferase denatured in 5 M GdnHCl. These conditions lead to a spontaneous refolding yield of 20% (Figure 3C), which is very high and limits conclusions on the effect of Hsp110 but also JDPs on the refolding process. Typically this assay uses 6 M GdnHCl for Luciferase denaturation and under these conditions, spontaneous refolding of Luciferase is hardly observed (e.g. Laufen et al. PNAS 1999). The authors are therefore asked to repeat key experiments using altered (6M) GdnHCl concentrations.

We based our experiments assessing luciferase refolding on the publication by Imamoglu et al. (2020), in which the authors, using 5 M GdnHCl for luciferase denaturation, demonstrated that spontaneous and chaperone-assisted luciferase refolding strongly depends on luciferase concentration. In this work, a similar degree of luciferase refolding was reported for the same final luciferase concentration (100 nM) as we used in our experiments (Figure 1—figure supplement 1D). As an additional control, we compared the effects of 5 M and 6 M of GdnHCl during denaturation on luciferase refolding under the same conditions (100 nM, 25 °C, 2 h) and we observed no significant differences (Author response image 1).

Author response image 1.

Chaperone-assisted folding of luciferase after denaturation at 5 M or 6 M GdnHCl. Luciferase was denatured in 5 M or 6 M GdnHCl according to the protocol in the Materials and Methods section. Luminescence was monitored alone or after incubation with Luminescence was monitored alone or after incubation with Ssa1-Sis1 or Ssa1-Ydj1. Chaperones were used at 1 µM concentration. Luciferase activity was measured after 2 hours and normalized to the activity of the native protein. Error bars indicate SD from three repeats.

- Figure 1B: The authors are asked to provide binding curves for Ssa1/Sse1 (no Sis1) and Sis1/Sse1 (no Ssa1) as controls. Particularly the latter combination is required as direct cooperation between Hsp110 and JDPs has been suggested in the literature (Mattoo et al., JBC 2013).

We performed the suggested BLI experiment, and the results are presented in the new Figure 1—figure supplement 1 G (page 5, lines 23-24).

- Figure 1B (and other figure parts showing BLI data): it is unclear how often the BLI experiments have been performed. This should be stated in the figure legend. Can the authors add SDs to the respective curves?

We added detailed information about the number of replicates to the figure legends. SD bars were added to the BLI results shown in Figures1-4, apart from the results of titrations, for which, for the sake of clarity, the three replicates are represented in the plots on the right (Figure 3D). In the case of less than 3 repeats of the results presented in the Supplementary Figures, the remaining repeats are added to the provided Source Data file, information about which has been added to the captions of the respective figures. 

- The observation that Hsp110 can interrupt Hsp70 interaction with JDPs is intriguing. Do the authors envision JDP displacement from the aggregate? If so this could be shown in BLI experiments by monitoring the release of fluorescently labeled Sis1 (similar to labeled Ssa1, Fig. S3C). Or will the released JDP immediately rebind to another binding site on the aggregate? The authors should at least discuss the diverse scenarios as they are relevant to the mechanism of protein disaggregation.

The proposed experiment is challenging due to the transient nature of Sis1 binding to aggregate and high background observed with the method using the fluorescently labelled proteins. The aspect of chaperone’s re-binding after their release by Hsp110 proposed by the reviewer has been introduced into the Discussion section (pages 12/13, lines 25-4). We speculate that Hsp110 might release an Hsp70 molecule as well as a JDP molecule that had been bound to the aggregate through Hsp70 (Figure 5).  

- Figure 2B: Ssa1/Sis1/Sse1 strongly decreases the size of Luciferase-GFP aggregates. Yet this activity only allows for limited refolding of aggregated Luciferase and the reaction stays largely dependent on Hsp104. How do the authors envision the role of the hexameric disaggregase in this process? Does it act exclusively on small-sized aggregates after Hsp110-dependent fragmentation?

A question of the Hsp104 activity with the Hsp70-processed aggregates is indeed intriguing and we agree that it should have been discussed more thoroughly. We added to the manuscript the results of the reactivation of luciferase-GFP with and without Hsp104 to emphasize the role of Hsp104 in the active protein recovery (Figure 2—figure supplement 1A) (page 7, lines 24-27). We propose that aggregate fragmentation by Hsp70-JDPB-Hsp110 increases the effective aggregate surface, at which Hsp104 might become engaged. We do not think that Hsp104 acts only on small aggregates, it might be just more effective, when the number of exposed polypeptides is larger. In the cell, where Hsp104 binds to aggregates of various sizes, protein aggregates apparently also need to undergo such Hsp110-boosted pre-processing by Hsp70, based on the finding that Sse1 is not necessary for Hsp104 recruitment to aggregates, but it is required for Hsp104-dependent disaggregation (Kaimal et al., 2017). We have added a comment on this problem to the Discussion section (pages 11/12, lines 33-4) .

- Page 9: The authors state that the Sse1-2 variant is nearly as effective as Sse1 Wt in stimulating substrate dissociation and refer to published work (Polier et al., 2008). It is unclear how the variant should have Wtlike activity in triggering substrate release although its activity in catalyzing nucleotide exchange is reduced to 5% (both activities are coupled). The observation that high Sse1-2 concentrations do not inhibit protein disaggregation does not necessarily exclude the possibility that high Sse1 WT concentration inhibit the reaction by overstimulating substrate release. The latter possibility should be considered by the authors and added to the discussion section.

We agree with the Reviewer that the description of the Sse1-2 variant was misleading, as it was lacking the key information, that according to the published data (Polier et al., 2008), it was 10 times higher the concentration of the Sse1-2 variant than Sse1 WT that had a similar nucleotide-exchange activity to the wild type. We have changed the text (page 9, lines 16-22, page 13, lines 26-28) to avoid confusion as well as the model in the Figure 5, to underline the importance of substrate release as the cause of the Hsp110-dependent inhibition.

- While similar effects are observed for human class A and class B JDP co-chaperones, they are clearly less pronounced. A mechanistic explanation for the difference between yeast and human chaperones is currently missing and the authors are asked to elaborate on this aspect.

There are indeed clear differences between the human and yeasts systems, especially regarding the dependence on the NEF. Hsc70 has been reported to have a lower rate of ADP release (Dragovic et al., 2006) and thus might rely more on Hsp110 than its yeast ortholog. For the same reason, the strong Hsc70 stimulation by Hsp105 is also observed with class A JDP. We have added a comment on these effects in the Discussion section (page 12, lines 17-21).

Minor points

- Figure S1C (right): the disaggregation rate (%GFP/h) is somewhat misleading/confusing as a value of more than 150%/h is determined in the presence of the complete disaggregation system while only approx. 60% GFP is indeed refolded by the system (Figure S1C, left). Showing the rate as %GFP/min seems more rational.

We changed the units according to the Reviewer’s comment (Figure 1—figure supplement 1A, C).

- Figure S5B: Only a single data point is shown for Ssa1/Sis1/Sse1.

We changed the figure to include datapoints from all three repeats (Figure 3—figure supplement 1 B).

- There are several typos throughout the manuscript. A more careful proofreading is recommended

We have corrected the typos.

Reviewer #1 (Public Review):

The experiments differ somewhat in regard to the aggregated protein used. For example, in Figure 1A, FFL is used with only limited reactivation (10% reactivated at the last timepoint and the curve is flattening), while in Figure 2B FFL-EGFP is used to monitor microscopically what appears to be complete disaggregation. Does FFL-EGFP behave the same as FFL in assays such as the one in Figure 1A or are there major differences that may impact how the data should be interpreted?

We added the results of Luc-GFP reactivation (Figure 2—figure supplement 1 B) (discussed on page 7, lines 24-27 of the manuscipt) which agree with the results obtain with Luciferase as a substrate (Figure 1—figure supplement 1 B). They clearly show that the Ssa1-Sis1-Sse1-dependent decrease in aggregate size is not associated with the recovery of active protein.

Reviewer #2 (Public Review):

Experimental data concerning the class A JDPs should be interpreted with caution. These experiments show very small reactivation activities for luciferase in the range of 0-1% without the addition of Hsp104 and 0-15% with the addition of Hsp104. Moreover, since the assay is based on the recovery of luciferase activity, it conflates two chaperone activities, namely disaggregation and refolding. It is possible that the small degree of reactivation observed for the class A JDP reflects a minor subpopulation of the aggregated species that is particularly easy to disaggregate/refold and may thus not be representative of bulk behaviour.

The disaggregation by the Hsp70 system can be enhanced by the addition of small heat shock proteins at the step of substrate aggregation (Rampelt et al., 2012). However, sHsps compete with Hsp70 for binding to the aggregate (Żwirowski et al., 2017) and for that reason we decided not to include sHsps in the experiments presented in the manuscript, as it would introduce another level of complexity. However, as a control, we performed the disaggregation assay with Hsp70 with Ydj1 using luciferase aggregates formed in the presence or absence of sHsp (Author response image 2). In 1 h, the Hsp70 system without Hsp104 yielded 5% of recovered luciferase activity and the system with Hsp104, 23% compared to the native. The impact of Sse1 on Ssa1-Ydj1 and Ssa1-Ydj1-Hsp104 was similar as for luciferase aggregates formed without sHsps (Figure 1A, Figure 1—figure supplement 1 B). Furthermore, according to the Reviewer’s comment, we have changed the Figure 5 to underscore the more prominent role of class A JDPs in the final protein folding than in disaggregation.

Author response image 2.

Disaggregaton of heat-aggregated luciferase – impact of sHsps. Luciferase (2 μM) was denatured with (blue) or without (red) Hsp26 (20 μM) at 45 ̊C for 15 min in the buffer A (Materials and Methods). Upon 100-fold dilution with the buffer A, supplemented wih 5 mM ATP, 2 mM DTT, 1.2 μM creatine kinase, 20 mM creatine phosphate, chaperones indicated in the legend were added to the final concentration of 1 μM, except for Sse1, concentration of which was 0.1 μM. Shown is luciferase activity measured after 1 h of incubation at 25 °C, normalized to the activity of native luciferase.

Reviewer #3 (Public Review):

Enhanced recruitment of Hsp70 in the presence of Hsp110 was shown for amyloid fibrils before (Beton et al., EMBO J 2022) and should be acknowledged. 

We have added the suggested citation with a respective comment (page 11, lines 20-21).

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

This paper details a study of endothelial cell vessel formation during zebrafish development. The results focus on the role of aquaporins, which mediate the flow of water across the cell membrane, leading to cell movement. The authors show that actin and water flow together drive endothelial cell migration and vessel formation. If any of these two elements are perturbed, there are observed defects in vessels. Overall, the paper significantly improves our understanding of cell migration during morphogenesis in organisms.

Strengths:

The data are extensive and are of high quality. There is a good amount of quantification with convincing statistical significance. The overall conclusion is justified given the evidence.

Weaknesses:

There are two weaknesses, which if addressed, would improve the paper.

(1) The paper focuses on aquaporins, which while mediates water flow, cannot drive directional water flow. If the osmotic engine model is correct, then ion channels such as NHE1 are the driving force for water flow. Indeed this water is shown in previous studies. Moreover, NHE1 can drive water intake because the export of H+ leads to increased HCO3 due to the reaction between CO2+H2O, which increases the cytoplasmic osmolarity (see Li, Zhou and Sun, Frontiers in Cell Dev. Bio. 2021). If NHE cannot be easily perturbed in zebrafish, it might be of interest to perturb Cl channels such as SWELL1, which was recently shown to work together with NHE (see Zhang, et al, Nat. Comm. 2022).

(2) In some places the discussion seems a little confusing where the text goes from hydrostatic pressure to osmotic gradient. It might improve the paper if some background is given. For example, mention water flow follows osmotic gradients, which will build up hydrostatic pressure. The osmotic gradients across the membrane are generated by active ion exchangers. This point is often confused in literature and somewhere in the intro, this could be made clearer.

Reviewer #1 (Recommendations For The Authors):

(1) The paper focuses on aquaporins, which while mediating water flow, cannot drive directional water flow. If the osmotic engine model is correct, then ion channels such as NHE1 are the driving force for water flow. Indeed this water is shown in previous studies. Moreover, NHE1 can drive water intake because the export of H+ leads to increased HCO3 due to the reaction between CO2+H2O, which increases the cytoplasmic osmolarity (see Li, Zhou and Sun, Frontiers in Cell Dev. Bio. 2021). If NHE cannot be easily perturbed in zebrafish, it might be of interest to perturb Cl channels such as SWELL1, which was recently shown to work together with NHE (see Zhang, et al, Nat. Comm. 2022).

We thank Reviewer #1 for this very important comment and the suggestion to examine the function of ion channels in establishing an osmotic gradient to drive directional flow. We have taken on board the reviewer’s suggestion and examined the expression of NHE1 and SWELL1 in endothelial cells using published scRNAseq of 24 hpf ECs (Gurung et al, 2022, Sci. Rep.). We found that slc9a1a, slc9a6a, slc9a7, slc9a8, lrrc8aa and lrrc8ab are expressed in different endothelial subtypes. To examine the function of NHE1 and SWELL1 in endothelial cell migration, we used the pharmacological compounds, 5-(N-ethyl-Nisopropyl)amiloride (EIPA) and DCPIB, respectively. While we were unable to observe an ISV phenotype after EIPA treatment at 5, 10 and 50µM, we were able to observe impaired ISV formation after DCPIB treatment that was very similar to that observed in Aquaporin mutants. We were very encouraged by these results and proceeded to perform more detailed experiments whose results have yielded a new figure (Figure 6) and are described and discussed in lines 266 to 289 and 396 to 407, respectively, in the revised manuscript.

(2) In some places the discussion seems a little confusing where the text goes from hydrostatic pressure to osmotic gradient. It might improve the paper if some background is given. For example, mention water flow follows osmotic gradients, which will build up hydrostatic pressure. The osmotic gradients across the membrane are generated by active ion exchangers. This point is often confused in literature and somewhere in the intro, this could be made clearer.

Thank you for pointing out the deficiency in explaining how osmotic gradients drive water flow to build up hydrostatic pressure. We have clarified this in lines 50, 53 - 54 and 385.

The two recommendations listed above would improve the paper. They are however not mandatory. The paper would be acceptable with some clarifying rewrites. I am not an expert on zebrafish genetics, so it might be difficult to perturb ion channels in this model organism. Have the authors tried to perturb ion channels in these cells?

We hope that our attempts at addressing Reviewer’s 1 comments are satisfactory and sufficient to clarify the concerns outlined.

Reviewer #2 (Public Review):

Summary:

Directional migration is an integral aspect of sprouting angiogenesis and requires a cell to change its shape and sense a chemotactic or growth factor stimulus. Kondrychyn I. et al. provide data that indicate a requirement for zebrafish aquaporins 1 and 8, in cellular water inflow and sprouting angiogenesis. Zebrafish mutants lacking aqp1a.1 and aqp8a.1 have significantly lower tip cell volume and migration velocity, which delays vascular development. Inhibition of actin formation and filopodia dynamics further aggravates this phenotype. The link between water inflow, hydrostatic pressure, and actin dynamics driving endothelial cell sprouting and migration during angiogenesis is highly novel.

Strengths:

The zebrafish genetics, microscopy imaging, and measurements performed are of very high quality. The study data and interpretations are very well-presented in this manuscript.

Weaknesses:

Some of the mechanobiology findings and interpretations could be strengthened by more advanced measurements and experimental manipulations. Also, a better comparison and integration of the authors' findings, with other previously published findings in mice and zebrafish would strengthen the paper.

We thank Reviewer #2 for the critique that the paper can be strengthened by more advanced measurements and experimental manipulations. One of the technical challenges that we face is how to visualize and measure water flow directly in the zebrafish. We have therefore taken indirect approaches to assess water abundance in endothelial cells in vivo. One approach was to measure the diffusion of GEM nanoparticles in tip cell cytoplasm in wildtype and Aquaporin mutants, but results were inconclusive. The second was to measure the volume of tip cells, which should reflect water in/outflow. As the second approach produced clear and robust differences between wildtype ECs, ECs lacking Aqp1a.1 and Aqp8a.1 and ECs overexpressing Aqp1a.1 (revised Fig. 5), we decided to present these data in this manuscript.

We have also taken Reviewer 2 advice to better incorporate previously published data in our discussion (see below and lines 374 to 383 of the revised manuscript).

Reviewer #2 (Recommendations For The Authors):

I have a few comments that the authors may address to further improve their manuscript analysis, quality, and impact.

Major comments:

(1) Citation and discussion of published literature

The authors have failed to cite and discuss recently published results on the role of aqp1a.1 and aqp8a.1 in ISV formation and caliber in zebrafish (Chen C et al. Cardiovascular Research 2024). That study showed a similar impairment of ISV formation when aqp1a.1 is absent but demonstrated a stronger phenotype on ISV morphology in the absence of aqp8a.1 than the current manuscript by Kondrychyn I et al. Furthermore, Chen C et al show an overall decrease in ISV diameter in single aquaporin mutants suggesting that the cell volume of all ECs in an ISV is affected equally. Given this published data, are ISV diameters affected in single and double mutants in the current study by Kondrochyn I et al? An overall effect on ISVs would suggest that aquaporin-mediated cell volume changes are not an inherent feature of endothelial tip cells. The authors need to analyse/compare and discuss all differences and similarities of their findings to what has been published recently.

We apologise for having failed and discussed the recently published paper by Chen et al. This has been corrected and discussed in lines 374 to 383.

In the paper by Chen et al, the authors describe a role of Aqp1a.1 and Aqp8a.1 in regulating ISV diameter (ISV diameter was analysed at 48 hpf) but they did not examine the earlier stages of sprouting angiogenesis between 20 to 30 hpf, which is the focus of our study. We therefore cannot directly compare the ISV phenotypes with theirs. Nevertheless, we recognise that there are differences in ISV phenotypes from 2 dpf. For example, they did not observe incompletely formed or missing ISVs at 2 and 3 dpf, which we clearly observe in our study. This could be explained by differences in the mutations generated. In Chen et al., the sgRNA used targeted the end of exon 2 that resulted in the generation of a 169 amino acid truncated aqp1a.1 protein. However, in our approach, our sgRNA targeted exon 1 of the gene that resulted in a truncated aqp1a.1 protein that is 76 amino acid long. As for the aqp8a.1 zebrafish mutant that we generated, our sgRNA targeted exon 1 of the gene that resulted in a truncated protein that is 73 amino acids long. In Chen et al., the authors did not generate an aqp8a.1 mutant but instead used a crispant approach, which leads to genetic mosaicism and high experimental variability.

Following the reviewer’s suggestion, we have now measured the diameters of arterial ISVs (aISVs) and venous ISVs (vISVs) in aqp1a.1^-/-, aqp8a.1^-/- and aqp1a.1^-/-;aqp8a.1^-/- zebrafish. In our lab, we always make a distinction between aISVs and vISVs are their diameters are significantly different from each other. The results are in Fig S11A. While we corroborate a decrease in diameter in both aISVs and vISVs in single aqp1a.1^-/- and double aqp1a.1^-/-;aqp8a.1^-/-.zebrafish, we observed a slight increase in diameter in both aISVs and vISVs in aqp8a.1^-/- zebrafish at 2 dpf. We also measured the diameter of aISV and vISV in Tg(fli1ep:aqp1a.1-mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) zebrafish at 2 dpf (Fig S11B) and unlike in Chen et al., we could not detect a difference in the diameter between control and aqp1a.1- or aqp8a.1-overexpressing endothelial cells.

We also would also like to point out that, because ISVs are incompletely formed or are missing in aqp1a.1^-/-;aqp8a.1^-/- zebrafish (Fig. 3G – L), blood flow is most likely altered in the zebrafish trunk of these mutants, and this can have a secondary effect on blood vessel calibre or diameter. In fact, we often observed wider ISVs adjacent to unperfused ISVs (Fig. 3J) as more blood flow enters the lumenized ISV. Therefore, to determine the cell autonomous function of Aquaporin in mediating cell volume changes in vessel diameter regulation, one would need to perform cell transplantation experiments where we would measure the volume of single aqp1a.1^-/-;aqp8a.1^-/- endothelial cells in wildtype embryos with normal blood flow. As this is beyond the scope of the present study, we have not done this experiment during the revision process.

(2) Expression of aqp1a.1 and aqp8a.1

The quantification shown in Figure 1G shows a relative abundance of expression between tip and stalk cells. However, it seems aqp8a.1 is almost never detected in most tip cells. The authors could show in addition, the % of Tip and stalk cells with detectable expression of the 2 aquaporins. It seems aqp8a1 is really weakly or not expressed in the initial stages. Ofcourse the protein may have a different dynamic from the RNA.

We would like to clarify that aqp8a.1 mRNA is not detected in tip cells of newly formed ISVs at 20hpf. At 22 hpf, it is expressed in both tip cells (22 out of 23 tip cells analysed) and stalk cells of ISVs at 22hpf. This is clarified in lines 107 - 109. We also include below a graph showing that although aqp8a.1 mRNA is expressed in tip cells, its expression is higher in stalk cells.

Author response image 1.

Could the authors show endogenously expressed or tagged protein by antibody staining? The analysis of the Tg(fli1ep:aqp8a.1-mEmerald)rk31 zebrafish line is a good complement, but unfortunately, it does not reveal the localization of the endogenously expressed protein. Do the authors have any data supporting that the endogenously expressed aqp8a.1 protein is present in sprouting tip cells?

We tested several antibodies against AQP1 (Alpha Diagnostic International, AQP11-A; ThermoFisher Scientific, MA1-20214; Alomone Labs, AQP-001) and AQP8 (Sigma Aldrich, SAB 1403559; Alpha Diagnostic International, AQP81-A; Almone Labs, AQP-008) but unfortunately none worked. As such, we do not have data demonstrating endogenous expression and localisation of Aqp1a.1 and Aqp8a.1 proteins in endothelial cells.

Could the authors perform F0 CRISPR/Cas9 mediated knockin of a small tag (i.e. HA epitope) in zebrafish and read the endogenous protein localization with anti-HA Ab?

CRISPR/Cas9 mediated in-frame knock-in of a tag into a genomic locus is a technical challenge that our lab has not established. We therefore cannot do this experiment within the revision period.

Given the double mutant phenotypic data shown, is aqp8a.1 expression upregulated and perhaps more important in aqp1a.1 mutants?

In our analysis of aqp1a.1 homozygous zebrafish, there is a slight down_regulation in _aqp8a.1 expression (Fig. S5C). Because the loss of Aqp1a.1 leads to a stronger impairment in ISV formation than the loss of Aqp8a.1 (see Fig. S6F, G, I and J), we believe that Aqp1a.1 has a stronger function than Aqp8a.1 in EC migration during sprouting angiogenesis.

Regarding the regulation of expression by the Vegfr inhibitor Ki8751, does this inhibitor affect Vegfr/ERK signalling in zebrafish and the sprouting of ISVs significantly?

ki8751 has been demonstrated to inhibit ERK signalling in tip cells in the zebrafish by Costa et al., 2016 in Nature Cell Biology. In our experiments, treatment with 5 µM ki8751 for 6 hours from 20 hpf also inhibited sprouting of ISVs.

The data presented suggest that tip cells overexpressing aqp1a.1-mEmerald (Figure 2C) need more than 6 times longer to migrate the same distance as tip cells expressing aqp8a.1mEmerald (Figure 2D). How does this compare with cells expressing only Emerald? A similar time difference can be seen in Movie S1 and Movie S2. Is it just a coincidence? Could aqp8a.1, when expressed at similar levels than aqp1a, be more functional and induce faster cell migration? These experiments were interpreted only for the localization of the proteins, but not for the potential role of the overexpressed proteins on function. Chen C et al. Cardiovascular Research 2024 also has some Aqp overexpression data.

The still images prepared for Fig. 2 C and D were selected to illustrate the localization of Aqp1a.1-mEmerald and Aqp8a.1-mEmerald at the leading edge of migrating tip cells. We did not notice that the tip cell overexpressing Aqp1a.1-mEmerald (Figure 2C) needed more than 6 times longer to migrate the same distance as the tip cell expressing aqp8a.1-mEmerald (Figure 2D), which the reviewer astutely detected. To ascertain whether there is a difference in migration speed between Aqp1a.1-mEmerald and Aqp8a.1-mEmerald overexpressing endothelial cells, we measured tip cell migration velocity of three ISVs from Tg(fli1ep:aqp1a.1-mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) zebrafish during the period of ISV formation (24 to 29 hpf) using the Manual Tracking plugin in Fiji. As shown in the graph, there is no significant difference in the migration speed of ECs overexpressing Aqp1a.1-mEmerald and Aqp8a.1-mEmerald, suggesting that Aqp8a.1-overexpressing cells migrate at a similar rate as Aqp1a.1-overexpressing cells. As we have not generated a Tg(fli1ep:mEmerald) zebrafish line, we are unable to determine whether endothelial cells migrate faster in Tg(fli1ep:aqp1a.1mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) zebrafish compared to endothelial cell expressing only mEmerald. As for the observation that tip cells overexpressing aqp1a.1mEmerald (Figure 2C) need more than 6 times longer to migrate the same distance as tip cells expressing aqp8a.1-mEmerald, we can only surmise that it is coincidental that the images selected “showed” faster migration of one ISV from Tg(fli1ep:aqp8a.1-mEmerald) zebrafish. We do not know whether the Aqp1a.1 and Aqp8a.1 are overexpressed to the same levels in Tg(fli1ep:aqp1a.1mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) zebrafish.

We would also like to point out that when we analysed the lengths of ISVs at 28 hpf in aqp1a.1^-/- and aqp8a.1^-/- zebrafish, ISVs were shorter in aqp1a.1^-/- zebrafish compared to aqp8a.1^-/- zebrafish (Fig. S6 F to J). These results indicate that the loss of Aqp1a.1 function causes slower migration than the loss of aqp8a.1 function, and suggest that Aqp1a.1 induces faster endothelial cell migration that Aqp8a.1.

Author response image 2.

The data on Aqps expression after the Notch inhibitor DBZ seems unnecessary, and is at the moment not properly discussed. It is also against what is set in the field. aqp8a.1 levels seem to increase only 24h after DBZ, not at 6h, and still authors conclude that Notch activation inhibits aqp8a.1 expression (Line 138-139). In the field, Notch is considered to be more active in stalk cells, where aqp8a.1 expression seems higher (not lower). Maybe the analysis of tip vs stalk cell markers in the scRNAseq data, and their correlation with Hes1/Hey1/Hey2 and aqp1 vs aqp8 mRNA levels will be more clear than just showing qRT-PCR data after DBZ.

As our scRNAseq data did not include ECs from earlier during development when ISVs are developing, we have analysed of scRNAseq data of 24 hpf endothelial cells published by Gurung et al, 2022 in Scientific Reports during the revision of this manuscript. However, we are unable to detect separate clusters of tip and stalk cells. As such, we are unable to correlate hes1/hey1/hey2 expression (which would be higher in stalk cells) with that of aqp1a.1/aqp8a.1. Also, we have decided to remove the DBZ-treatment results from our manuscript as we agree with the two reviewers that they are unnecessary.

The paper would also benefit from some more analysis and interpretation of available scRNAseq data in development/injury/disease/angiogenesis models (zebrafish, mice or humans) for the aquaporin genes characterized here. To potentially raise a broader interest at the start of the paper.

We thank the reviewer for suggesting examining aquaporin genes in other angiogenesis/disease/regeneration models to expand the scope of aquaporin function. We will do this in future studies.

(3) Role of aqp1a.1 and aqp8a.1 on cytoplasmic volume changes and related phenotypes

In Figure 5 the authors show that Aqp1/Aqp8 mutant endothelial tip cells have a lower cytoplasmic volume than tip cells from wildtype fish. If aquaporin-mediated water inflow occurs locally at the leading edge of endothelial tip cells (Figure 2, line 314-318), why doesn't cytoplasmic volume expand specifically only at that location (as shown in immune cells by Boer et al. 2023)? Can the observed reduction in cytoplasmic volume simply be a side-effect of impaired filopodia formation (Figure 4F-I)?

We believe that water influx not only expands filopodia but also the leading front of tip cells (see bracket region in Fig. 4D), where Aqp1a.1-mEmerald/Aqp8a.1-mEmerald accumulate (Fig. 2), to generate an elongated protrusion and forward expansion of the tip cell. The decrease in cytoplasmic volume observed in the aqp1a.1;aqp8a.1 double mutant zebrafish is a result of decreased formation of these elongated protrusions at the leading front of migration tip cells as shown in Fig. 4E (compare to Fig. 4D), not from just a decrease in filopodia number. In fact, in the method used to quantify cell volume, mEmerald/EGFP localization is limited to the cytoplasm and does not label filopodia well (compare mEmerald/EGFP in green with membrane tagged-mCherry in Fig. 5A - C). The volume measured therefore reflects cytoplasmic volume of the tip cell, not filopodia volume.

Do the authors have data on cytoplasmic volume changes of endothelial tip cells in latrunculin B treated fish? The images in Figures 6 A,B suggest that there is a difference in cell volume upon lat b treatment only.

No, unfortunately we have not performed single cell labelling and measurement of tip cells in Latrunculin B-treated embryos. We can speculate that as there is a decrease in actindriven membrane protrusions in this experiment, one would also expect a decrease in cell volume as the reviewer has observed.

(4) Combined loss of aquaporins and actin-based force generation.

Lines 331-332 " we show that hydrostatic pressure is the driving force for EC migration in the absence of actin-based force generation"....better leave it more open and stick to the data. The authors show that aquaporin-mediated water inflow partially compensates for the loss of actin-based force generation in cell migration. Not that it is the key driving/rescuing force in the absence of actin-based force.

We have changed it to “we show that hydrostatic pressure can generate force for EC migration in the absence of actin-based force generation” in line 348.

(5) Aquaporins and their role in EC proliferation

In the study by Phnk LK et al. 2013, the authors have shown that proliferation is not affected when actin polymerization or filopodia formation is inhibited. However, in the current manuscript by Kondrychyn I. et al. this has not been analysed carefully. In Movie S4 the authors indicate by arrows tip cells that fail to invade the zebrafish trunk demonstrating a severe defect of sprouting initiation in these mutants. Yet, when only looking at ISVs that reach the dorsal side in Movie S4, it appears that they are comprised of fewer EC nuclei/ISV than the ISVs in Movie S3. At the beginning of DLAV formation, most ISVs in control Movie S3 consist of 3-4 EC nuclei, while in double mutants Movie S4 it appears to be only 2-3 EC nuclei. At the end of the Movie S4, one ISV on the left side even appears to consist of only a single EC when touching the dorsal roof. The authors provide convincing data on how the absence of aquaporin channels affects sprouting initiation and migration speed, resulting in severe delay in ISV formation. However, the authors should also analyse EC proliferation, as it may also be affected in these mutants, and may also contribute to the observed phenotype. We know that effects on cell migration may indirectly change the number of cells and proliferation at the ISVs, but this has not been carefully analysed in this paper.

We thank the reviewer for highlighting the lack of information on EC number and division in the aquaporin mutants. We have now quantified EC number in ISVs that are fully formed (i.e. connecting the DA or PCV to the DLAV) at 2 and 3 dpf and the results are displayed in Figure S10A and B. At 2 dpf, there is a slight but significant reduction in EC number in both aISVs and vISVs in aqp1a.1^-/- zebrafish and an even greater reduction in the double aqp1a. aqp1a.1^/-;aqp8a.1^-/- zebrafish. No significant change in EC number was observed in aqp8a.1^-/- zebrafish. EC number was also significantly decreased at 3 dpf for aqp1a.1^-/-, aqp8a.1^-/- and aqp1a.1^-/-;aqp8a.1^-/- zebrafish. The decreased in EC number per ISV may therefore contribute to the observed phenotype.

We have also quantified the number of cell divisions during sprouting angiogenesis (from 21 to 30 hpf) to assess whether the lack of Aquaporin function affects EC proliferation. This analysis shows that there is no significant difference in the number of mitotic events between aqp1a.1^+/-; aqp8a.1^+/- and aqp1a.1^-/-;aqp8a.1^-/- zebrafish (Figure S10 C), suggesting that the reduction in EC number is not caused by a decrease in EC proliferation.

These new data are reported on lines 198 to 205 of the manuscript.

Minor comments:

- Figure 3K data seems not to be necessary and even partially misleading after seeing Figure 3E. Fig. 3E represents the true strength of the phenotype in the different mutants.

Figure 3K has been removed from Figure 3.

- Typo Figure 3L (VII should be VI).

Thank you for spotting this typo. VII has been changed to VI.

- Line 242: The word "required" is too strong because there is vessel formation without Aqps in endothelial cells.

This has been changed to “ …Aqp1a.1 and Aqp8a.1 regulate sprouting angiogenesis…” (lines 238 - 239).

- From Figure S2, the doublets cluster should be removed.

We have performed a new analysis of 24 hpf, 34hpf and 3 dpf endothelial cells scRNAseq data (the previous analysis did not consist of 24 hpf endothelial cells). The doublets cluster is not included in the UMAP analysis.

- Better indicate the fluorescence markers/alleles/transgenes used for imaging in Figures 6A-D.

The transgenic lines used for this experiment are now indicated in the figure (this figure is now Figure 7).

Reviewer #3 (Public Review):

Summary:

Kondrychyn and colleagues describe the contribution of two Aquaporins Aqp1a.1 and Aqp8a.1 towards angiogenic sprouting in the zebrafish embryo. By whole-mount in situ hybridization, RNAscope, and scRNA-seq, they show that both genes are expressed in endothelial cells in partly overlapping spatiotemporal patterns. Pharmacological inhibition experiments indicate a requirement for VEGR2 signaling (but not Notch) in transcriptional activation.

To assess the role of both genes during vascular development the authors generate genetic mutations. While homozygous single mutants appear normal, aqp1a.1;aqp8a.1 double mutants exhibit defects in EC sprouting and ISV formation.

At the cellular level, the aquaporin mutants display a reduction of filopodia in number and length. Furthermore, a reduction in cell volume is observed indicating a defect in water uptake.

The authors conclude, that polarized water uptake mediated by aquaporins is required for the initiation of endothelial sprouting and (tip) cell migration during ISV formation. They further propose that water influx increases hydrostatic pressure within the cells which may facilitate actin polymerization and formation membrane protrusions.

Strengths:

The authors provide a detailed analysis of Aqp1a.1 and Aqp8a.1 during blood vessel formation in vivo, using zebrafish intersomitic vessels as a model. State-of-the-art imaging demonstrates an essential role in aquaporins in different aspects of endothelial cell activation and migration during angiogenesis.

Weaknesses:

With respect to the connection between Aqp1/8 and actin polymerization/filopodia formation, the evidence appears preliminary and the authors' interpretation is guided by evidence from other experimental systems.

Reviewer #3 (Recommendations For The Authors):

Figure 1 H, J:

The differential response of aqp1/-8 to ki8751 vs DBZ after 6h treatment is quite obvious. Why do the authors show the effect after 24h? The effect is more likely than not indirect.

We agree with the reviewer and we have now removed 24 hour Ki8751 treatment and all DBZ treatments from Figure 1.

Figure 2:

According to the authors' model anterior localization of Aqp1 protein is critical. The authors perform transient injections to mosaically express Aqp fusion proteins using an endothelial (fli1) promoter. For the interpretation, it would be helpful to also show the mCherry-CAAX channel in separate panels. From the images, it is not possible to discern how many cells we are looking at. In particular the movie in panel D may show two cells at the tip of the sprout. A marker labelling cell-cell junctions would help. Furthermore, the authors are using a strong exogenous promoter, thus potentially overexpressing the fusion protein, which may lead to mislocalization. For Aqp1a.1 an antibody has been published to work in zebrafish (e.g. Kwong et al., Plos1, 2013).

We would like to clarify that we generated transgenic lines - Tg(fli1ep:aqp1a.1-mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) - to visualize the localization of Aqp1a.1 and Aqp8a.1 in endothelial cells, and the images displayed in Fig. 2 are from the transgenic lines (not transient, mosaic expression).

To aid visualization and interpretation, we have now added mCherry-CAAX only channel to accompany the Aqp1a.1/Aqp8a.1-mEmerald channel in Fig. 2A and B. To discern how many cells there are in the ISVs at this stage, we have crossed Tg(fli1ep:aqp1a.1-mEmerald) and Tg(fli1ep:aqp8a.1-mEmerald) zebrafish to TgKI(tjp1a-tdTomato)^pd1224 (Levic et al., 2021) to visualize ZO1 at cell-cell junction. However, because tjp1-tdTomato is expressed in all cell types including the skin that lies just above the ISV and the signal in ECs in ISVs is very weak at 22 to 25 hpf, it was very difficult to obtain good quality images that can properly delineate cell boundaries to determine the number of cells in the ISVs at this early stage. Instead, we have annotated endothelial cell boundaries based on more intense mCherryCAAX fluorescence at cell-cell borders, and from the mosaic expression of mCherryCAAX that is intrinsic to the  Tg(kdrl:ras-mCherry)^s916 zebrafish line.

In Fig. 2D, there are two endothelial cells in the ISV during the period shown but there is only 1 cell occupying the tip cell position i.e. there is one tip cell in this ISV. Unlike the mouse retina where it has been demonstrated that two endothelial cells can occupy the tip cell position side-by-side (Pelton et al., 2014), this is usually not observed in zebrafish ISVs. This is demonstrated in Movie S3, where it is clear that one nucleus (belonging to the tip cell) occupies the tip of the growing ISV. The accumulation of intracellular membranes is often observed in tip cells that may serve as a reservoir of membranes for the generation of membrane protrusions at the leading edge of tip cells.

We agree that by generating transgenic Tg(fli1ep:aqp1a.1-mEmerald) and Tg(fli1ep:aqp8a.1mEmerald) zebrafish, Aqp1a.1 and Aqp8a.1 are overexpressed that may affect their localization. The eel anti-Aqp1a.1 antibody used in (Kwong et la., 2013) was a gift from Dr. Gordon Cramb, Univ. of St Andrews, Scotland and it was first published in 2001. This antibody is not available commercially. Instead, we have tried to several other antibodies against AQP1 (Alpha Diagnostic International , AQP11-A; ThermoFisher Scientific, MA120214; Alomone Labs, AQP-001) and AQP8 (Sigma Aldrich, SAB 1403559; Alpha Diagnostic International, AQP81-A; Almone Labs, AQP-008) but unfortunately none worked. As such, we cannot compare localization of Aqp1a.1-mEmerald and Aqp8a.1-mEmerald with the endogenous proteins.

Figure 3:

E: the quantification is difficult to read. Wouldn't it be better to set the y-axis in % of the DV axis? (see also Figure S6).

We would like to show the absolute length of the ISVs, and to illustrate that the ISV length decreases from anterior to posterior of the zebrafish trunk. We have increased the size of Fig. 3E to enable easier reading of the bars.

K: This quantification appears arbitrary.

We have removed this panel from Figure 3.

G-J: The magenta channel is difficult to see. Is the lifeact-mCherry mosaic? In panel J there appears to be a nucleus between the sprout and the DLAV. It would be helpful to crop the contralateral side of the image.

No, the Tg(fli1:Lifeact-mCherry) line is not mosaic. The “missing” vessels are not because of mosaicism in transgene but because of truncated ISVs that is a phenotype of loss Aquaporin function. We have changed the magenta channel to grey and hope that by doing so, the reviewer will be able to see the shape of the blood vessels more clearly. We would like to leave the contralateral side in the images, as it shows that the defective vessel is only on one side of body. Furthermore, when we tried to remove it (reducing the number of Z-stacks) neighbour ISV looks incomplete because the embryos were not mounted flat. To clarify what the nucleus between the sprout and the DLAV is, we have indicated that it is that of the contralateral ISV.

L: I do not quite understand the significance of the different classes of phenotypes. Do the authors propose different morphogenetic events or contexts of how these differences come about?

Here, we report the different types of ISV phenotypes that we observe in 3 dpf aqp1a.1^-/-; aqp8a.1^-/- zebrafish (Fig. 3 and Fig. S7). As demonstrated in Fig. 4, most of the phenotypes can be explained by the delayed emergence of tip cells from the dorsal aorta and slower tip cell migration. However, in some instances, we also observed retraction of tip cells (Movie S4) and failure of tip cells to emerge from the dorsal aorta or endothelial cell death (see attached figure on page 14), which can give rise to the Class II phenotype. In the dominant class I phenotype (in contrast to class II), secondary sprouting from the posterior cardinal vein is unaffected, and the secondary sprout migrates dorsally passing the level of horizontal myoseptum but cannot complete the formation of vISV (it stops beneath the spinal cord). The Class III phenotype appears to result from a failure of the secondary sprout to fuse with the regressed primary ISV. In the Class IV phenotype, the ventral EC does not maintain a connection to the dorsal aorta. We did not examine how Class III and IV phenotypes arise in detail in this current study.

Author response image 3.

Figure 4:

This figure nicely demonstrates the defects in cell behavior in aqp mutants.

In panel F it would be helpful to show the single channels as well as the merge.

We have now added single channels for PLCd1PH and Lifeact signal in panels F and G.

In Figure 1 the authors argue that the reduction of Aqp1/8 by VEGFR2 inhibition may account for part of that phenotype. In turn, the aqp phenotype seems to resemble incomplete VEGFR2 inhibition. The authors should check whether expression Aqp1Emerald can partially rescue ki8751 inhibition.

To address the reviewer’s comment, we have treated Tg(fli1ep:Aqp1-Emerald) embryos with ki8751 from 20 hpf for 6 hours but we were unable to observe a rescue in sprouting. It could be because VEGFR2 inhibition also affects other downstream signalling pathways that also control cell migration as well as proliferation.

Based on previous studies (Loitto et al.; Papadopoulus et al.) the authors propose that also in ISVs aquaporin-mediated water influx may promote actin polymerization and thereby filopodia formation. However, while the effect on filopodia number and length is well demonstrated, the underlying cause is less clear. For example, filopodia formation could be affected by reduced cell polarization. This can be tested by using a transgenic golgi marker (Kwon et al., 2016).

We have examined tip cell polarity of wildtype, aqp1a.1^-/- and  aqp8a. 1^-/- embryos at 24-26 hpf by analysing Golgi position relative to the nucleus. We were unable to analyze polarity in  aqp1a.1^rk28/rk28; aqp8a.1^rk29/rk29 embryos as they exist in an mCherry-containing transgenic zebrafish line (the Golgi marker is also tagged to mCherry). The results show that tip cell polarity is similar, if not more polarised, in aqp1a.1^-/- and  aqp8a. 1^-/- embryos when compared to wildtype embryos (Fig. S10D). This new data is discussed in lines 234 to 237.

Figure 5:

Panel D should be part of Figure 4.

Panel 5D is now in panel J of Figure 4 and described in lines 231 and 235.

Author Response

The following is the authors’ response to the original reviews.

Reviewer #1:

People can perform a wide variety of different tasks, and a long-standing question in cognitive neuroscience is how the properties of different tasks are represented in the brain. The authors develop an interesting task that mixes two different sources of difficulty, and find that the brain appears to represent this mixture on a continuum, in the prefrontal areas involved in resolving task difficulty. While these results are interesting and in several ways compelling, they overlap with previous findings and rely on novel statistical analyses that may require further validation.

Strengths

1) The authors present an interesting and novel task for combining the contributions of stimulus-stimulus and stimulus-response conflict. While this mixture has been measured in the multi-source interference task (MSIT), this task provides a more graded mixture between these two sources of difficulty

2) The authors do a good job triangulating regions that encoding conflict similarity, looking for the conjunction across several different measures of conflict encoding

3) The authors quantify several salient alternative hypothesis and systematically distinguish their core results from these alternatives

4) The question that the authors tackle is of central theoretical importance to cognitive control, and they make an interesting an interesting contribution to this question

We would like to thank the reviewer for the positive evaluation of our manuscript and the constructive comments and suggestions. Your feedback has been invaluable in our efforts to enhance the accessibility of our manuscript and strengthen our findings. In response to your suggestion, we reanalyzed our data using the approach proposed by Chen et al.’s (2017, NeuroImage) and applied stricter multiple comparison correction thresholds in our reporting. This reanalysis largely replicated our previous results, thereby reinforcing the robustness of our findings. We also have examined several alternative models and results supported the integration of the spatial Stroop and Simon conflicts within the cognitive space. In addition, we enriched the theoretical framework of our manuscript by connecting the cognitive space with other important theories such as the “Expected Value of Control” theory. We have incorporated your feedback, revisions and additional analyses into the manuscript. As a result, we firmly believe that these changes have significantly improved the quality of our work. We have provided detailed responses to your comments below.

1) It's not entirely clear what the current task can measure that is not known from the MSIT, such as the additive influence of conflict sources in Fu et al. (2022), Science. More could be done to distinguish the benefits of this task from MSIT.

We agree that the MSIT task incorporates Simon and Eriksen Flanker conflict tasks and can efficiently detect the additivity of conflict effects across orthogonal tasks. Like the MSIT, our task incorporates Simon with spatial Stroop conflicts and can test the same idea. For example, a previous study from our lab (Li et al., 2014) used the combined spatial Stroop-Simon condition with the arrows displayed on diagonal corners and found evidence for the additive hypothesis. However, the MSIT cannot be used to test whether/how different conflicts are parametrically represented in a low-dimensional space, a question that is important to address the debate of domain-general and domain-specific cognitive control.

To this end, our current study adopted the spatial Stroop-Simon task for the unique purpose of parametrically modulating conflict similarity. As far as we know, there is no way to define the similarity between the combined Simon_Flanker conflict condition and the Simon/Flanker conditions in the MSIT. In contrast, with the spatial Stroop-Simon paradigm, we can define the similarity with the cosine of the angle difference across the two conditions in question.

We have added the following texts in the discussion part to emphasize the 51 difference between our paradigm and other studies.

"The use of an experimental paradigm that permits parametric manipulation of conflict similarity provides a way to systematically investigate the organization of cognitive control, as well as its influence on adaptive behaviors. This approach extends traditional paradigms, such as the multi-source interference task (Fu et al., 2022), color Stroop-Simon task (Liu et al., 2010) and similar paradigms that do not afford a quantifiable metric of conflict source similarity."

References:

Li, Q., Nan, W., Wang, K., & Liu, X. (2014). Independent processing of stimulus-stimulus and stimulus-response conflicts. PloS One, 9(2), e89249.

2) The evidence from this previous work for mixtures between different conflict sources make the framing of 'infinite possible types of conflict' feel like a strawman. The authors cite classic work (e.g., Kornblum et al., 1990) that develops a typology for conflict which is far from infinite, and I think few people would argue that every possible source of difficulty will have to be learned separately. Such an issue is addressed in theories like 'Expected Value of Control', where optimization of control policies can address unique combinations of task demands.

The notion that there might be infinite conflicts arises when we consider the quantitative feature of cognitive control. If each combination of the Stroop-Simon combination is regarded as a conflict condition, there would be infinite combinations, and it is our major goal to investigate how these infinite conflict conditions are represented effectively in a space with finite dimensions. We agree that it is unnecessary to dissociate each of these conflict conditions into a unique conflict type, since they may not differ substantially. However, we argue that understanding variant conflicts within a purely categorical framework (e.g., Simon and Flanker conflict in MSIT) is insufficient, especially because it leads to dichotomic conclusions that do not capture how combinations of conflicts are organized in the brain, as our study addresses.

There could be different perspectives on how our cognitive control system flexibly encodes and resolves multiple conflicts. The cognitive space assumption we held provides a principle by which we can represent multiple conflicts in a lower dimensional space efficiently. While the “Expected Value of Control” theory addresses when and how much cognitive control to apply based on control demand, the “cognitive space” view seeks to explain how the conflict, which defines cognitive control demand, is encoded in the brain. Thus, we argue that these two lines of work are different yet complementary. The geometry of cognitive space of conflict can benefit the adjustment of cognitive control for upcoming conflicts. For example, our brain may evaluate the similarity/distance (and thus cost) between the consecutive conflict conditions, and selects the path with best cost-benefit tradeoff to switch from one state to another. This idea is conceptually similar to a recent study by Grahek et al. (2022) demonstrating that more frequently switching states were encoded as closer together than less frequently switching states in a “drift-threshold” space.

Nevertheless, Grahek et al (2022) investigated how cognitive control changes based on the expected value of control theory within the same conflict, whereas our study aims to examine organization of different conflict.

We have added the implications of cognitive space view in the discussion to indicate the potential values of our finding to understand the EVC account and the difference between the two theories.

“Previous researchers have proposed an “expected value of control (EVC)” theory, which posits that the brain can evaluate the cost and benefit associated with executing control for a demanding task, such as the conflict task, and specify the optimal control strength (Shenhav et al., 2013). For instance, Grahek et al. (2022) found that more frequently switching goals when doing a Stroop task were achieved by adjusting smaller control intensity. Our work complements the EVC theory by further investigating the neural representation of different conflict conditions and how these representations can be evaluated to facilitate conflict resolution. We found that different conflict conditions can be efficiently represented in a cognitive space encoded by the right dlPFC, and participants with stronger cognitive space representation have also adjusted their conflict control to a greater extent based on the conflict similarity (Fig 4C). The finding suggests that the cognitive space organization of conflicts guides cognitive control to adjust behavior. Previous studies have shown that participants may adopt different strategies to represent a task, with the model-based strategies benefitting goal-related behaviors more than the model-free strategies (Rmus et al., 2022). Similarly, we propose that cognitive space could serve as a mental model to assist fast learning and efficient organization of cognitive control settings. Specifically, the cognitive space representation may provide a principle for how our brain evaluates the expected cost of switching and the benefit of generalization between states and selects the path with the best cost-benefit tradeoff (Abrahamse et al., 2016; Shenhav et al., 2013). The proximity between two states in cognitive space could reflect both the expected cognitive demand required to transition and the useful mechanisms to adapt from. The closer the two conditions are in cognitive space, the lower the expected switching cost and the higher the generalizability when transitioning between them. With the organization of a cognitive space, a new conflict can be quickly assigned a location in the cognitive space, which will facilitate the development of cognitive control settings for this conflict by interpolating nearby conflicts and/or projecting the location to axes representing different cognitive control processes, thus leading to a stronger CSE when following a more similar conflict condition. On the other hand, without a cognitive space, there would be no measure of similarity between conflicts on different trials, hence limiting the ability of fast learning of cognitive control setting from similar trials.”

Reference:

Grahek, I., Leng, X., Fahey, M. P., Yee, D., & Shenhav, A. Empirical and Computational Evidence for Reconfiguration Costs During Within-Task Adjustments in Cognitive Control. CogSci.

3) Wouldn't a region that represented each conflict source separately still show the same pattern of results? The degree of Stroop vs Simon conflict is perfectly negatively correlated across conditions, so wouldn't a region that just tracks Stoop conflict show these RSA patterns? The authors show that overall congruency is not represented in DLPFC (which is surprising), but they don't break it down by whether this is due to Stroop or Simon congruency (I'm not sure their task allows for this).

To estimate the unique contributions of the spatial Stroop and Simon conflicts, we performed a model-comparison analysis. We constructed a Stroop-Only model and a Simon-Only model, with each conflict type projected onto the Stroop (vertical) axis or Simon (horizontal) axis, respectively. The similarity between any two conflict types was defined using the Jaccard similarity index (Jaccard, P., 1901), that is, their intersection divided by their union. By replacing the cognitive spacebased conflict similarity regressor with the Stroop-Only and Simon-Only regressors, we calculated their BICs. Results showed that the BIC was larger for Stroop-Only (5377122) and Simon-Only (5377096) than for the Cognitive-Space model (5377094). An additional Stroop+Simon model, including both Stroop-Only and Simon-Only regressors, also showed a poorer model fitting (BIC = 5377118) than the Cognitive-Space model. Considering that the pattern of conflict representations is more manifested when the conflict is present (i.e., on incongruent trials) than not (i.e., on congruent trials), we also conducted the model comparison using the incongruent trials only. Results showed that Stroop-Only (1344128), Simon-Only (1344120), and Stroop+Simon (1344157) models all showed higher BIC values than the CognitiveSpace model (1344104). These results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. Therefore, we believe the cognitive space has incorporated both dimensions. We added these additional analyses and results to the revised manuscript.

“To examine if the right 8C specifically encodes the cognitive space rather than the domain-general or domain-specific organizations, we tested several additional models (see Methods). Model comparison showed a lower BIC in the Cognitive-Space model (BIC = 5377094) than the Domain-General (BIC = 537127) or Domain-Specific (BIC = 537127) models. Further analysis showed the dimensionality of the representation in the right 8C was 1.19, suggesting the cognitive space was close to 1D. We also tested if the observed conflict similarity effect was driven solely by spatial Stroop or Simon conflicts, and found larger BICs for the models only including the Stroop similarity (i.e., the Stroop-Only model, BIC = 5377122) or Simon similarity (i.e., the Simon-Only model, BIC = 5377096). An additional Stroop+Simon model, including both StroopOnly and Simon-Only regressors, also showed a worse model fitting (BIC = 5377118). Moreover, we replicated the results with only incongruent trials, considering that the pattern of conflict representations is more manifested when the conflict is present (i.e., on incongruent trials) than not (i.e., on congruent trials). We found a poorer fitting in Domain-general (BIC = 1344129), Domain-Specific (BIC = 1344129), Stroop-Only (BIC = 1344128), Simon-Only (BIC = 1344120), and Stroop+Simon (BIC = 1344157) models than the Cognitive-Space model (BIC = 1344104). These results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. The more detailed model comparison results are listed in Table 2.”

We reason that we did not observe an overall congruency effect in the RSA results is because our definition of congruency here differed from traditional definitions (i.e., contrast between incongruent and congruent conditions). In the congruency regressor of our RSA model, we defined representational similarity as 1 if calculated between two incongruent, or two congruent trials, and 0 if between incongruent and congruent trials. Thus, our definition of the congruency regressor reflects whether multivariate patterns differ between incongruent and congruent trials, rather than whether activity strengths differ. Indeed, we did observe the latter form of congruency effects, with stronger univariate activities in pre-SMA for incongruent versus congruent conditions. We have added this in the Note S6 (“The multivariate representations of conflict type and orientation are different from the congruency effect”):

“Neither did we observe a multivariate congruency effect (i.e., the pattern difference between incongruent and congruent conditions compared to that within each condition) in the right 8C or any other regions. Note the definition of congruency here differed from traditional definitions (i.e., contrast between activity strength of incongruent and congruent conditions), with which we found stronger univariate activities in pre-SMA for incongruent versus congruent conditions.”

We could not determine whether the null effect of the congruency regressor was due to Stroop or Simon congruency alone, because congruency levels of the two types always covary. On all trials of the compound conditions (Conf 2-4), whenever the Stroop dimension was incongruent, the Simon dimension was also incongruent, and vice versa for the congruent condition. Thus, the contribution of spatial Stroop or Simon alone to the congruency effect could not be tested using compound conditions. Although we have pure spatial Stroop or Simon conditions, within-Stroop and withinSimon trial pairs constituted only 8% of cells in the representational similarity matrix. This was insufficient to determine whether the null congruency effect was due to solely Stroop or Simon.

Overall, with the added analysis we found that the data in the right 8C area supports conflict representations that are organized based on both Simon and spatial Stroop conflict. Although the current experimental design does not allow us to identify whether the null effect of the congruency regressor was driven by either conflict or both, we clarified that the congruency regressor did not test the 205 conventional congruency effect and the null finding does not contradict previous 206 research.

Reference:

Jaccard, P. (1901). Étude comparative de la distribution florale dans une portion des Alpes et des Jura. Bull Soc Vaudoise Sci Nat(37), 547-579.

4) The authors use a novel form of RSA that concatenates patterns across conditions, runs and subjects into a giant RSA matrix, which is then used for linear mixed effects analysis. This appears to be necessary because conflict type and visual orientation are perfectly confounded within the subject (although, if I understand, the conflict type x congruence interaction wouldn't have the same concern about visual confounds, which shouldn't depend on congruence). This is an interesting approach but should be better justified, preferably with simulations validating the sensitivity and specificity of this method and comparing it to more standard methods.

The confound exists for both the conflict type and the conflict type × congruence interaction in our design, since both incongruent and congruent conditions include stimuli from the full orientation space. For example, for the spatial Stroop type, the congruent condition could be either an up arrow at the top or a down arrow at the bottom. Similarly, the incongruent condition could be either an up arrow at the bottom or a down arrow at the top. Therefore, both the congruent and incongruent conditions are perfectly confounded with the orientation.

We reanalyzed the data using the well-documented approach by Chen et al. (2017, Neuroimage), as suggested by the reviewer. The new analysis replicated our previously reported results (Fig. 4-5, S4-S7). As Chen et al (2017) has provided abundant simulations to validate this approach, we did not run any further simulations.

5) A chief concern is that the same pattern contributes to many entries in the DV, which has been addressed in previous work using row-wise and column-wise random effects (Chen et al., 2017, Neuroimage). It would also be informative to know whether the results hold up to removing within-run similarity, which can bias similarity measures (Walther et al., 2016, Neuroimage).

Thank you for the comment. In our revised manuscript, we followed your suggestion and adopted the approach proposed by Chen et al. (2017). Specifically, we included both the upper and lower triangle of the representational similarity matrix (excluding the diagonal). Moreover, we also removed all the within-subject similarity (thus also excluding the within-run similarity as suggested by Walther et al. (2016)) to minimize the bias of the potentially strong within-subject similarity. In addition, we added both the row-wise and column-wise random effects to capture the dependence of cells within each column and each row, respectively (Chen et al., 2017).

Results from this approach largely replicated our previous results. The right 8C again showed significant conflict similarity representation, with greater representational strength in incongruent than congruent condition, and positively correlated to behavioral performance. The orientation effect was also identified in the visual (e.g., right V1) and oculomotor (e.g., left FEF) regions.

We have revised the methodology and the results in the revised manuscript:

"Representational similarity analysis (RSA).

For each cortical region, we calculated the Pearson’s correlations between fMRI activity patterns for each run and each subject, yielding a 1400 (20 conditions × 2 runs × 35 participants) × 1400 RSM. The correlations were calculated in a cross297 voxel manner using the fMRI activation maps obtained from GLM3 described in the previous section. We excluded within-subject cells from the RSM (thus also excluding the within-run similarity as suggested by Walther et al., (2016)), and the remaining cells were converted into a vector, which was then z-transformed and submitted to a linear mixed effect model as the dependent variable. The linear mixed effect model also included regressors of conflict similarity and orientation similarity. Importantly, conflict similarity was based on how Simon and spatial Stroop conflict are combined and hence was calculated by first rotating all subject’s stimulus location to the top right and bottom-left quadrants, whereas orientation was calculated using original stimulus locations. As a result, the regressors representing conflict similarity and orientation similarity were de-correlated. Similarity between two conditions was measured as the cosine value of the angular difference. Other regressors included a target similarity regressor (i.e., whether the arrow directions were identical), a response similarity regressor (i.e., whether the correct responses were identical); a spatial Stroop distractor regressor (i.e., vertical distance between two stimulus locations); a Simon distractor regressor (i.e., horizontal distance between two stimulus locations). Additionally, we also included a regressor denoting the similarity of Group (i.e., whether two conditions are within the same subject group, according to the stimulus-response mapping). We also added two regressors including ROI316 mean fMRI activations for each condition of the pair to remove the possible uni-voxel influence on the RSM. A last term was the intercept. To control the artefact due to dependence of the correlation pairs sharing the same subject, we included crossed random effects (i.e., row-wise and column-wise random effects) for the intercept, conflict similarity, orientation and the group factors (G. Chen et al., 2017)."

Reference:

Walther, A., Nili, H., Ejaz, N., Alink, A., Kriegeskorte, N., & Diedrichsen, J. (2016). Reliability of dissimilarity measures for multi-voxel pattern analysis. Neuroimage, 137, 188-200. doi:10.1016/j.neuroimage.2015.12.012

6) Another concern is the extent to which across-subject similarity will only capture consistent patterns across people, making this analysis very similar to a traditional univariate analysis (and unlike the traditional use of RSA to capture subject-specific patterns).

With proper normalization, we assume voxels across different subjects should show some consistent localizations, although individual differences can be high. J. Chen et al. (2017) has demonstrated that consistent multi-voxel activation patterns exist across individuals. Previous studies have also successfully applied cross-subject RSA (see review by Freund et al, 2021) and cross-subject decoding approaches (e.g., Jiang et al., 2016; Tusche et al., 2016), so we believe cross-subject RSA should be feasible to capture distributed activation patterns shared at the group level. We added this argument in the revised manuscript:

"Previous studies (e.g., J. Chen et al., 2017) have demonstrated that consistent multivoxel activation patterns exist across individuals, and successful applications of cross-subject RSA (see review by Freund, Etzel, et al., 2021) and cross-subject decoding approaches (Jiang et al., 2016; Tusche et al., 2016) have also been reported."

In the revised manuscript, we also tested whether the representation in right 8C held for within-subject data. We reasoned that the conflict similarity effects identified by cross-subject RSA should be replicable in within-subject data, although the latter is not able to dissociate the conflict similarity effect from the orientation effect. We performed similar RSA for within-subject RSMs, excluding the within-run cells. We replaced the perfectly confounded factors of conflict similarity and orientation with a common factor called similarity_orientation. Other confounding factor pairs were addressed similarly. Results showed a significant effect of similarity_orientation, t(13993) = 3.270, p = .0005, 1-tailed. Given the specific representation of conflict similarity identified by the cross-subject RSA, we believe that the within-subject data of right 8C probably showed similar conflict similarity modulation effects as the cross-subject data, although future research that orthogonalizes conflict type and orientation is needed to fully answer this question. We added this result in the revised section Note S7.

"Note S7. The cross-subject RSA captures similar effects with the within-subject RSA Considering the variability in voxel-level functional localizations among individuals, one may question whether the cross-subject RSA results were biased by the consistent multi-voxel patterns across subjects, distinct from the more commonly utilized withinsubject RSA. We reasoned that the cross-subject RSA should have captured similar effects as the within-subject RSA if we observe the conflict similarity effect in right 8C with the latter analysis. Therefore, we tested whether the representation in right 8C held for within-subject data. Specifically, we performed similar RSA for withinsubject RSMs, excluding the within-run cells. We replaced the perfectly confounded factors of conflict similarity and orientation with a common factor called similarity_orientation. Other confounding factor pairs (i.e., target versus response, and Stroop distractor versus Simon distractor) were addressed similarly. Results showed a significant effect of similarity_orientation, t(13993) = 3.270, p = .0005, 1tailed. Given the specific representation of conflict similarity identified by the crosssubject RSA, the within-subject data of right 8C may show similar conflict similarity modulation effects as the cross-subject data. Further research is needed to fully dissociate the representation of conflict and the representation of visual features such as orientation."

Reference:

Chen, J., Leong, Y. C., Honey, C. J., Yong, C. H., Norman, K. A., & Hasson, U. (2017). Shared memories reveal shared structure in neural activity across individuals. Nature Neuroscience, 20(1), 115-125.

Freund, M. C., Etzel, J. A., & Braver, T. S. (2021). Neural Coding of Cognitive Control: The Representational Similarity Analysis Approach. Trends in Cognitive Sciences, 25(7), 622-638.

Jiang, J., Summerfield, C., & Egner, T. (2016). Visual Prediction Error Spreads Across Object Features in Human Visual Cortex. J Neurosci, 36(50), 12746-12763.

Tusche, A., Bockler, A., Kanske, P., Trautwein, F. M., & Singer, T. (2016). Decoding the Charitable Brain: Empathy, Perspective Taking, and Attention Shifts Differentially Predict Altruistic Giving. Journal of Neuroscience, 36(17), 4719-4732.

7) Finally, the authors should confirm all their results are robust to less liberal methods of multiplicity correction. For univariate analysis, they should report the effects from the standard p < .001 cluster forming threshold for univariate analysis (or TFCE). For multivariate analyses, FDR can be quite liberal. The authors should consider whether their mixed-effects analyses allow for group-level randomization, and consider (relatively powerful) Max-Stat randomization tests (Nichols & Holmes, 2002, Hum Brain Mapp).

In our revised manuscript, we have corrected the univariate results using the probabilistic TFCE (pTFCE) approach by Spisak et al. (2019). This approach estimates the conditional probability of cluster extent based on Bayes’ rule. Specifically, we applied pTFCE on our univariate results (i.e., the z-maps of our contrasts). This returned enhanced Z-score maps, which were then thresholded based on simulated cluster size thresholds using 3dClustSim. A cluster-forming threshold of p < .001 was employed. Results showed only the pre-SMA was activated in the incongruent > congruent contrast, and right IPS and right dmPFC were activated in the linear Simon modulation effect. Further tests also showed these regions were not correlated with the behavioral performance, uncorrected ps >.28. These results largely replicated our previous results. We have revised the method and results accordingly.

Methods:

"Results were corrected with the probabilistic threshold-free cluster enhancement(pTFCE) and then thresholded by 3dClustSim function in AFNI (Cox & Hyde, 1997) with voxel-wise p < .001 and cluster-wize p < .05, both 1-tailed."

Results:

"In the fMRI analysis, we first replicated the classic congruency effect by searching for brain regions showing higher univariate activation in incongruent than congruent conditions (GLM1, see Methods). Consistent with the literature (Botvinick et al., 2004; Fu et al., 2022), this effect was observed in the pre-supplementary motor area (preSMA) (Fig. 3, Table S1). We then tested the encoding of conflict type as a cognitive space by identifying brain regions with activation levels parametrically covarying with the coordinates (i.e., axial angle relative to the horizontal axis) in the hypothesized cognitive space. As shown in Fig. 1B, change in the angle corresponds to change in spatial Stroop and Simon conflicts in opposite directions. Accordingly, we found the right inferior parietal sulcus (IPS) and the right dorsomedial prefrontal cortex (dmPFC) displayed positive correlation between fMRI activation and the Simon conflict (Fig. 3, Fig. S3, Table S1)."

We appreciate the reviewer’s suggestion to apply the Max-Stat randomization tests (Nichols & Holmes, 2002) for the multivariate analyses. However, the representational similarity matrix was too large (1400×1400) to be tested with a balanced randomization approach (i.e., the Max-Stat), due to (1) running even 1000 times for all ROIs cost very long time; (2) the distribution generated from normal times of randomization (e.g., 5000 iterations) would probably be unbalanced, since the full range of possible samples that could be generated by a complete randomization is not adequately represented. Instead, we adopted a very strict Bonferroni correction p < 0.0001/360 when reporting the regression results from RSA. Notebally, Chen et al (2017) has shown that their approach could control the FDR at an acceptable level.

Reference:

Spisák, T., Spisák, Z., Zunhammer, M., Bingel, U., Smith, S., Nichols, T., & Kincses,T. (2019). Probabilistic TFCE: A generalized combination of cluster size and voxel intensity to increase statistical power. NeuroImage, 185, 12-26.

Chen, G., Taylor, P. A., Shin, Y.-W., Reynolds, R. C., & Cox, R. W. J. N. (2017). Untangling the relatedness among correlations, Part II: Inter-subject correlation group analysis through linear mixed-effects modeling. 147, 825-840.

Minor concerns:

8) I appreciate the authors wanting to present the conditions in a theory-agnostic way, but the framing of 5 conflict types was confusing. I think framing the conditions as a mixture of 2 conflict types (Stroop and Simon) makes more sense, especially given the previous work on MSIT.

We have renamed the Type1-5 as spatial Stroop, StHSmL, StMSmM, StLSmH, and Simon conditions, respectively. H, L, and M indicate high, low andmedium similarity with the corresponding conflict, respectively. This is alsoconsistent with the naming of our previous work (Yang et al., 2021).

Reference:

Yang, G., Xu, H., Li, Z., Nan, W., Wu, H., Li, Q., & Liu, X. (2021). The congruency sequence effect is modulated by the similarity of conflicts. Journal of Experimental Psychology: Learning, Memory, and Cognition, 47(10), 1705-1719.

9) It would be helpful to have more scaffolding for the key conflict & orientation analyses. A schematic in the main text that outlines these contrasts would be very helpful (e.g. similar to S4).

We have inserted Figure 7 in the revised manuscript. In this figure, we plotted the schematic of the difference between the conflict similarity 467 and orientation regressors according to their cross-group representational similarity 468 matrices.

10) Figure 4D could be clearer, both in labeling and figure caption. 'Modeled similarity' could be relabelled to something more informative, like 'conflict type (or mixture) similarity'. Alternatively, it would be helpful to show a summary RDM for region r-8C. For example, breaking it down by just conflict type and congruence.

We have relabeled the x-axis to “Conflict type similarity” and y-axis to “Neural similarity” for Figure 4D in the revised manuscript.

We have also added a summary RSM figure in Fig. S5 to show the different similarity patterns between incongruent and congruent conditions.

11) It may be helpful to connect your work to how people have discussed multiple forms of conflict monitoring and control with respect to target and distractor features e.g., Lindsay & Jacoby, 1994, JEP:HPP; Mante, Sussillo et al., 2013, Nature; Soutschek et al., 2015, JoCN; Jackson et al., 2021, Comm Bio; Ritz & Shenhav, 2022, bioRxiv

We have added an analysis to examine how cognitive control modulates target and distractor representation. To this end, we selected the left V4, a visual region showing joint representation of target, Stroop distractor and Simon distractor, as the region of interest. We tested whether these representation strengths differed between incongruent and congruent conditions, finding the representation of target was stronger and representations of both distractors were weaker in the incongruent condition. This suggests that cognitive control modulates the stimuli in both directions. We added the results in Note S10 and Fig. S8, and also added discussion of it in “Methodological implications”.

“Note S10. Cognitive control enhances target representation and suppresses distractor representation Using the separability of confounding factors afforded by the cross-subject RSA, we examined how representations of targets and distractors are modulated by cognitive control. The key assumption is that exerting cognitive control may enhance target representation and suppress distractor representation. We hypothesized that stimuli are represented in visual areas, so we chose a visual ROI from the main RSA results showing joint representation of target, spatial Stroop distractor and Simon distractor (p < .005, 1-tail, uncorrected). Only the left V4 met this criterion. We then tested representations with models similar to the main text for incongruent only trials, congruent only trials, and the incongruent – congruent contrast. The contrast model additionally used interaction between the congruency and target, Stroop distractor and Simon distractor terms. Results showed that in the incongruent condition, when we employ more cognitive control, the target representation was enhanced (t(237990) = 2.59, p = .029, Bonferroni corrected) and both spatial Stroop (t(237990) = –4.18, p < .001, Bonferroni corrected) and Simon (t(237990) = –3.14, p = .005, Bonferroni corrected) distractor representations were suppressed (Fig. S8). These are consistent with the idea that the top-down control modulates the stimuli in both directions (Polk et al., 2008; Ritz & Shenhav, 2022).”

Discussion:

“Moreover, the cross-subject RSA provides high sensitivity to the variables of interest and the ability to separate confounding factors. For instance, in addition to dissociating conflict type from orientation, we dissociated target from response, and spatial Stroop distractor from Simon distractor. We further showed cognitive control can both enhance the target representation and suppress the distractor representation (Note S10, Fig. S8), which is in line with previous studies (Polk et al., 2008; Ritz & Shenhav, 2022)."

12) For future work, I would recommend placing stimuli along the whole circumference, to orthogonalize Stroop and Simon conflict within-subject.

We thank the reviewer for this highly helpful suggestion. Expanding the 547 conflict conditions to a full conflict space and replicating our current results could 548 provide stronger evidence for the cognitive space view.

In the revised manuscript, we added this as a possible future design:

“A possible improvement to our current design would be to include left, right, up, and down arrows presented in a grid formation across four spatially separate quadrants, with each arrow mapped to its own response button. However, one potential confounding factor would be that these conditions have different levels of difficulty (i.e., different magnitude of conflict), which may affect the CSE results and their representational similarity."

Reviewer #2:

Summary, general appraisal

This study examines the construct of "cognitive spaces" as they relate to neural coding schemes present in response conflict tasks. The authors utilize a novel paradigm, in which subjects must map the direction of a vertically oriented arrow to either a left or right response. Different types of conflict (spatial Stroop, Simon) are parametrically manipulated by varying the spatial location of the arrow (a taskirrelevant feature). The vertical eccentricity of the arrow either agrees or conflicts with the arrow's direction (spatial Stroop), while the horizontal eccentricity of the arrow agrees or conflicts with the side of the response (Simon). A neural coding model is postulated in which the stimuli are embedded in a cognitive space, organized by distances that depend only on the similarity of congruency types (i.e., where conditions with similar relative proportions of spatial-Stroop versus Simon congruency are represented with similar activity patterns). The authors conduct a behavioral and fMRI study to provide evidence for such a representational coding scheme. The behavioral findings replicate the authors' prior work in demonstrating that conflict-related cognitive control adjustments (the congruency sequence effect) shows strong modulation as a function of the similarity between conflict types. With the fMRI neural activity data, the authors report univariate analyses that identified activation in left prefrontal and dorsomedial frontal cortex modulated by the amount of Stroop or Simon conflict present, and multivariate representational similarity analyses (RSA) that identified right lateral prefrontal activity encoding conflict similarity and correlated with the behavioral effects of conflict similarity.

This study tackles an important question regarding how distinct types of conflict, which have been previously shown to elicit independent forms of cognitive control adjustments, might be encoded in the brain within a computationally efficient representational format. The ideas postulated by the authors are interesting ones and the utilized methods are rigorous.

We would like to express our sincere appreciation for the reviewer’s positive evaluation of our manuscript and the constructive comments and suggestions. Through careful consideration of your feedback, we have endeavored to make our manuscript more accessible to readers and further strengthened our findings. In response to your suggestion, we reanalyzed our data with the approach proposed by Chen et al.’s (2017, NeuroImage). This reanalysis largely replicated our previous results, reinforcing the validity of our findings. Additionally, we conducted tests with several alternative models and found that the cognitive space hypothesis best aligns with our observed data. We have incorporated these revisions and additional analyses into the manuscript based on your valuable feedback. As a result, we believe that these changes and additional analyses have significantly enhanced the quality of our manuscript. We have provided detailed responses to your comments below.

However, the study has critical limitations that are due to a lack of clarity regarding theoretical hypotheses, serious confounds in the experimental design, and a highly non-standard (and problematic) approach to RSA. Without addressing these issues it is hard to evaluate the contribution of the authors findings to the computational cognitive neuroscience literature.

1) The primary theoretical question and its implications are unclear. The paper would greatly benefit from more clearly specifying potential alternative hypotheses and discussing their implications. Consider, for example, the case of parallel conflict monitors. Say that these conflict monitors are separately tuned for Stroop and Simon conflict, and are located within adjacent patches of cortex that are both contained within a single cortical parcel (e.g., as defined by the Glasser atlas used by the authors for analyses). If RSA was conducted on the responses of such a parcel to this task, it seems highly likely that an activation similarity matrix would be observed that is quite similar (if not identical) to the hypothesized one displayed in Figure 1. Yet it would seem like the authors are arguing that the "cognitive space" representation is qualitatively and conceptually distinct from the "parallel monitor" coding scheme. Thus, it seems that the task and analytic approach is not sufficient to disambiguate these different types of coding schemes or neural architectures.

The authors also discuss a fully domain-general conflict monitor, in which different forms of conflict are encoded within a single dimension. Yet this alternative hypothesis is also not explicitly tested nor discussed in detail. It seems that the experiment was designed to orthogonalize the "domain-general" model from the "cognitive space" model, by attempting to keep the overall conflict uniform across the different stimuli (i.e., in the design, the level of Stroop congruency parametrically trades off with the level of Simon congruency). But in the behavioral results (Fig. S1), the interference effects were found to peak when both Stroop and Simon congruency are present (i.e., Conf 3 and 4), suggesting that the "domain-general" model may not be orthogonal to the "cognitive space" model. One of the key advantages of RSA is that it provides the ability to explicitly formulate, test and compare different coding models to determine which best accounts for the pattern of data. Thus, it would seem critical for the authors to set up the design and analyses so that an explicit model comparison analysis could be conducted, contrasting the domain-general, domain-specific, and cognitive space accounts.

We appreciate the reviewer pointing out the need to formally test alternative models. In the revised manuscript, we have added and compared a few alternative models, finding the Cognitive-Space model (the one with graded conflict similarity levels as we reported) provided the best fit to our data. Specifically, we tested the following five models against the Cognitive-Space model:

(1) Domain-General model. This model treats each conflict type as equivalent, so each two conflict types only differ in the magnitude of their conflict. Therefore, we defined the domain-general matrix as the difference in their effects indexed by the group-averaged RT in Experiment 2. Then the z-scored model vector was sign-flipped to reflect similarity instead of distance. This model showed non-significant conflict type effects (t(951989) = 0.92, p = .179) and poorer fit (BIC = 5377126) than the Cognitive-Space model (BIC = 5377094).

(2) Domain-Specific model. This model treats each conflict type differently, so we used a diagonal matrix, with within-conflict type similarities being 1 and all crossconflict type similarities being 0. This model also showed non-significant effects (t(951989) = 0.84, p = .201) and poorer fit (BIC = 5377127) than the Cognitive-Space model.

(3) Stroop-Only model. This model assumes that the right 8C only encodes the spatial Stroop conflict. We projected each conflict type to the Stroop (vertical) axis and calculated the similarity between any two conflict types as the Jaccard similarity index (Jaccard, 1901), that is, their intersection divided by their union. This model also showed non-significant effects (t(951989) = 0.20, p = .423) and poorer fit (BIC = 5377122) than the Cognitive-Space model.

(4) Simon-Only model. This model assumes that the right 8C only encodes the Simon conflict. We projected each conflict type to the Simon (horizontal) axis and calculated the similarity like the Stroop-Only model. This model showed significant effects (t(951989) = 4.19, p < .001) but still quantitatively poorer fit (BIC = 5377096) than the Cognitive-Space model.

(5) Stroop+Simon model. This model assumes the spatial Stroop and Simon conflicts are parallelly encoded in the brain, similar to the "parallel monitor" hypothesis suggested by the reviewer. It includes both Stroop-Only and Simon-Only regressors. This model showed nonsignificant effect for the Stroop regressor (t(951988) = 0.06, p = .478) and significant effect for the Simon regressor (t(951988) = 3.30, p < .001), but poorer fit (BIC = 5377118) than the Cognitive-Space model.

“Moreover, we replicated these results with only incongruent trials (i.e., when conflict is present), considering that the pattern of conflict representations is more manifested when the conflict is present (i.e., on incongruent trials) than not (i.e., on congruent trials). We found a poorer fitting in Domain-general (BIC = 1344129), Domain-Specific (BIC = 1344129), Stroop-Only (BIC = 1344128), Simon-Only (BIC = 1344120), and Stroop+Simon (BIC = 1344157) models than the Cognitive-Space model (BIC = 1344104).”

In summary, these results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. We added the above results to the revised manuscript.

The above analysis approach was added to the method “Model comparison and representational dimensionality”, and the results were added to the “Multivariate patterns of the right dlPFC encodes the conflict similarity” in the revised manuscript.

Methods:

“Model comparison and representational dimensionality To estimate if the right 8C specifically encodes the cognitive space, rather than the domain-general or domain-specific structures, we conducted two more RSAs. We replaced the cognitive space-based conflict similarity matrix in the RSA we reported above (hereafter referred to as the Cognitive-Space model) with one of the alternative model matrices, with all other regressors equal. The domain-general model treats each conflict type as equivalent, so each two conflict types only differ in the magnitude of their conflict. Therefore, we defined the domain-general matrix as the difference in their congruency effects indexed by the group-averaged RT in Experiment 2. Then the zscored model vector was sign-flipped to reflect similarity instead of distance. The domain-specific model treats each conflict type differently, so we used a diagonal matrix, with within-conflict type similarities being 1 and all cross-conflict type similarities being 0.

Moreover, to examine if the cognitive space is driven solely by the Stroop or Simon conflicts, we tested a spatial Stroop-Only (hereafter referred to as “Stroop-Only”) and a Simon-Only model, with each conflict type projected onto the spatial Stroop (vertical) axis or Simon (horizontal) axis, respectively. The similarity between any two conflict types was defined using the Jaccard similarity index (Jaccard, 1901), that is, their intersection divided by their union. We also included a model assuming the Stroop and Simon dimensions are independently represented in the brain, adding up the StroopOnly and Simon-Only regressors (hereafter referred to as the Stroop+Simon model). We conducted similar RSAs as reported above, replacing the original conflict similarity regressor with the Strrop-Only, Simon-Only, or both regressors (for the Stroop+Simon model), and then calculated their Bayesian information criterions (BICs).”

Results:

“To examine if the right 8C specifically encodes the cognitive space rather than the domain-general or domain-specific organizations, we tested several additional models (see Methods). Model comparison showed a lower BIC in the Cognitive-Space model (BIC = 5377094) than the Domain-General (BIC = 537127) or Domain-Specific (BIC = 537127) models. Further analysis showed the dimensionality of the representation in the right 8C was 1.19, suggesting the cognitive space was close to 1D. We also tested if the observed conflict similarity effect was driven solely by spatial Stroop or Simon conflicts, and found larger BICs for the models only including the Stroop similarity (i.e., the Stroop-Only model, BIC = 5377122) or Simon similarity (i.e., the Simon-Only model, BIC = 5377096). An additional Stroop+Simon model, including both StroopOnly and Simon-Only regressors, also showed a worse model fitting (BIC = 5377118). Moreover, we replicated the results with only incongruent trials, considering that the pattern of conflict representations is more manifested when the conflict is present (i.e., on incongruent trials) than not (i.e., on congruent trials). We found a poorer fitting in Domain-general (BIC = 1344129), Domain-Specific (BIC = 1344129), Stroop-Only (BIC = 1344128), Simon-Only (BIC = 1344120), and Stroop+Simon (BIC = 1344157) models than the Cognitive-Space model (BIC = 1344104). These results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. The more detailed model comparison results are listed in Table 2.”

Reference:

Jaccard, P. (1901). Étude comparative de la distribution florale dans une portion des Alpes et des Jura. Bull Soc Vaudoise Sci Nat(37), 547-579.

2a) Relatedly, the reasoning for the use of the term "cognitive space" is unclear. The mere presence of graded coding for two types of conflict seems to be a low bar for referring to neural activity patterns as encoding a "cognitive space". It is discussed that cognitive spaces/maps allow for flexibility through inference and generalization. But no links were made between these cognitive abilities and the observed representational structure.

In the revised manuscript, we have clarified that we tested a specific prediction of the cognitive space hypothesis: the geometry of the cognitive space predicts that more similar conflict types will have more similar neural representations,leading to the CSE and RSA patterns tested in this study. These results add to the literature by providing empirical evidence on how different conflict types are encoded in the brain. We agree that this study is not a comprehensive test of the cognitive space hypothesis. Thus, in the revised manuscript we explicitly clarified that this study is a test of the geometry of the cognitive space hypothesis.

Critically, the cognitive space view holds that the representations of different abstract information are organized continuously and the representational geometry in the cognitive space are determined by the similarity among the represented information (Bellmund et al., 2018).

"The present study aimed to test the geometry of cognitive space in conflict representation. Specifically, we hypothesize that different types of conflict are represented as points in a cognitive space. Importantly, the distance between the points, which reflects the geometry of the cognitive space, scales with the difference in the sources of the conflicts being represented by the points."

We have also discussed the limitation of the results and stressed the need for more research to fully test the cognitive space hypothesis.

“Additionally, our study is not a comprehensive test of the cognitive space hypothesis but aimed primarily to provide original evidence for the geometry of cognitive space in representing conflict information in cognitive control. Future research should examine other aspects of the cognitive space such as its dimensionality, its applicability to other conflict tasks such as Eriksen Flanker task, and its relevance to other cognitive abilities, such as cognitive flexibility and learning.

2b) Additionally, no explicit tests of generality (e.g., via cross-condition generalization) were provided.

To examine the generality of cognitive space across conditions, we conducted a leave-one-out prediction analysis. We used the behavioral data from Experiment 1 for this test, due to its larger amount of data than Experiment 2. Specifically, we removed data from one of the five similarity levels (as illustrated by the θs in Fig. 1C) and used the remaining data to perform the same mixed-effect model as reported in the main text (i.e., the two-stage analysis). This yielded one pair of beta coefficients including the similarity regressor and the intercept for each subject, with which we predicted the CSE for the removed similarity level for each subject. We repeated this process for each similarity level once. The predicted results were highly correlated with the original data, with r = .87 for the RT and r = .84 for the ER, ps < .001. We have added this analysis and result to the “Conflict type 706 similarity modulated behavioral congruency sequence effect (CSE)” section.

“Moreover, to test the continuity and generalizability of the similarity modulation, we conducted a leave-one-out prediction analysis. Specifically, we removed data from one of the five similarity levels (as illustrated by the θs in Fig. 1C) and used the remaining data to perform the same mixed-effect model (i.e., the two-stage analysis). This yielded one pair of beta coefficients including the similarity regressor and the intercept for each subject, with which we predicted the CSE for the removed similarity level for each subject. We repeated this process for each similarity level once. The predicted results were highly correlated with the original data, with r = .87 for the RT and r = .84 for the ER, ps < .001."

2c) Finally, although the design elicits strong CSE effects, it seems somewhat awkward to consider CSE behavioral patterns as a reflection of the kind of abilities supported by a cognitive map (if this is indeed the implication that was intended). In fact, CSE effects are well-modeled by simpler "model-free" associative learning processes, that do not require elaborate representations of abstract structures.

We argue the conflict similarity modulation of CSEs we observed cannot be explained by the “model-free” stimulus-driven associative learning process. This mainly refers to the feature integration account proposed by Hommel et al. (2004), which explains poorer performance in CI and IC trials (compared with CC and II trials) with the partial repetition cost caused by the breaking of stimulus-response binding. Although we cannot remove its influence on the within-type trials (similarity level 5, θ = 0), it should not affect the cross-type trials (similarity level 1-4, θ = 90°, 67.5°, 45° and 22.5°, respectively), because the CC, CI, IC, II trials had equal probabilities of partially repeated and fully switched trials (see the Author response image 1 for an example of trials across Conf 1 and Conf 3 conditions). Thus, feature integration cannot explain the gradual CSE decrease from similarity level 1 to 4, which sufficiently reproduce the full effect, as suggested by the leave-one-out prediction analysis mentioned above. We thus conclude that the similarity modulation of CSE cannot be explained by the stimulus-driven associative learning.

Author response image 1.

Notably, however, our findings are aligned with an associative learning account of cognitive control (Abrahamse et al., 2016), which extends association learning from stimulus/response level to cognitive control. In other words, abstract cognitive control state can be learned and generalized like other sensorimotor features. This view explicitly proposes that “transfer occurs to the extent that two tasks overlap”, a hypothesis directly supported by our CSE results (see also Yang et al., 2021). Extending this, our fMRI results provide the neural basis of how cognitive control can generalize through a representation of cognitive space. The cognitive space view complements associative learning account by providing a fundamental principle for the learning and generalization of control states. Given the widespread application of CSE as indicator of cognitive control generalization (Braem et al., 2014), we believe that it can be recognized as a kind of ability supported by the cognitive space. This was further supported by the brain-behavioral correlation: stronger encoding of cognitive space was associated with greater bias of trial-wise behavioral adjustment by the consecutive conflict similarity.

We have incorporated these ideas into the discussion:

“Similarly, we propose that cognitive space could serve as a mental model to assist fast learning and efficient organization of cognitive control settings. Specifically, the cognitive space representation may provide a principle for how our brain evaluates the expected cost of switching and the benefit of generalization between states and selects the path with the best cost-benefit tradeoff (Abrahamse et al., 2016; Shenhav et al., 2013). The proximity between two states in cognitive space could reflect both the expected cognitive demand required to transition and the useful mechanisms to adapt from. The closer the two conditions are in cognitive space, the lower the expected switching cost and the higher the generalizability when transitioning between them. With the organization of a cognitive space, a new conflict can be quickly assigned a location in the cognitive space, which will facilitate the development of cognitive control settings for this conflict by interpolating nearby conflicts and/or projecting the location to axes representing different cognitive control processes, thus leading to a stronger CSE when following a more similar conflict condition.”

References:

Hommel, B., Proctor, R. W., & Vu, K. P. (2004). A feature-integration account of sequential effects in the Simon task. Psychological Research, 68(1), 1-17. Abrahamse, E., Braem, S., Notebaert, W., & Verguts, T. (2016). Grounding cognitive control in associative learning. Psychological Bulletin, 142(7), 693-728.

Yang, G., Xu, H., Li, Z., Nan, W., Wu, H., Li, Q., & Liu, X. (2021). The congruency sequence effect is modulated by the similarity of conflicts. Journal of 770 Experimental Psychology: Learning, Memory, and Cognition, 47(10), 1705-1719.

Braem, S., Abrahamse, E. L., Duthoo, W., & Notebaert, W. (2014). What determines the specificity of conflict adaptation? A review, critical analysis, and proposed synthesis. Frontiers in Psychology, 5, 1134.

3) More generally, it seems problematic that Stroop and Simon conflict in the paradigm parametrically trade-off against each other. A more powerful design would have de-confounded Stroop and Simon conflict so that each could be separately estimation via (potentially orthogonal) conflict axes. Additionally, incorporating more varied stimulus sets, locations, or responses might have enabled various tests of generality, as implied by a cognitive space account.

We thank the reviewer for these valuable suggestions. We argue that the current design is adequate to test the prediction that more similar conflict types have more similar neural representations. That said, we agree that further examination using more powerful experimental designs are needed to fully test the cognitive space account of cognitive control. We also agree that employing more varied stimulus sets,locations and responses would further extend our findings. We have included this as a future research direction in the revised manuscript.

We have revised our discussion about the limitation as:

“A few limitations of this study need to be noted. To parametrically manipulate the conflict similarity levels, we adopted the spatial Stroop-Simon paradigm that enables parametrical combinations of spatial Stroop and Simon conflicts. However, since this paradigm is a two-alternative forced choice design, the behavioral CSE is not a pure measure of adjusted control but could be partly confounded by bottom-up factors such as feature integration (Hommel et al., 2004). Future studies may replicate our findings with a multiple-choice design (including more varied stimulus sets, locations and responses) with confound-free trial sequences (Braem et al., 2019). Another limitation is that in our design, the spatial Stroop and Simon effects are highly anticorrelated. This constraint may make the five conflict types represented in a unidimensional space (e.g., a circle) embedded in a 2D space. Future studies may test the 2D cognitive space with fully independent conditions. A possible improvement to our current design would be to include left, right, up, and down arrows presented in a grid formation across four spatially separate quadrants, with each arrow mapped to its own response button. However, one potential confounding factor would be that these conditions have different levels of difficulty (i.e., different magnitude of conflict), which may affect the CSE results and their representational similarity.”

4) Serious confounds in the design render the results difficult to interpret. As much prior neuroimaging and behavioral work has established, "conflict" per se is perniciously correlated with many conceptually different variables. Consequently, it is very difficult to distinguish these confounding variables within aggregate measures of neural activity like fMRI. For example, conflict is confounded with increased time-on-task with longer RT, as well as conflict-driven increases in coding of other task variables (e.g., task-set related coding; e.g., Ebitz et al. 2020 bioRxiv). Even when using much higher resolution invasive measures than fMRI (i.e., eCoG), researchers have rightly been wary of making strong conclusions about explicit encoding of conflict (Tang et al, 2019; eLife). As such, the researchers would do well to be quite cautious and conservative in their analytic approach and interpretation of results.

We acknowledge the findings showing that encoding of conflicts may not be easily detected in the brain. However, recent studies have shown that the representational similarity analysis can effectively detect representations of conflict tasks (e.g., the color Stroop) using factorial designs (Freund et al., 2021a; 2021b).

In our analysis, we are aware of the potential impact of time-on-task (e.g., RT) on univariate activation levels and subsequent RSA patterns. To address this issue, we added univariate fMRI activation levels as nuisance regressors to the RSA. To de confound conflict from other factors such as orientation of stimuli related to the center of the screen, we also applied the cross-subject RSA approach. Furthermore, we were cautious about determining regions that encoded conflict control. We set three strict criteria: (1) Regions must show a conflict similarity modulation effect; (2) regions must show higher representational strength in the incongruent condition compared with the congruent condition; and (3) regions must correlate with behavioral performance. With these criteria, we believe that the results we reported are already conservative. We would be happy to implement any additional criteria the reviewer recommends.

Reference:

Freund, M. C., Etzel, J. A., & Braver, T. S. (2021a). Neural Coding of Cognitive Control: The Representational Similarity Analysis Approach. Trends in Cognitive Sciences, 25(7), 622-638.

Freund, M. C., Bugg, J. M., & Braver, T. S. (2021b). A Representational Similarity 823 Analysis of Cognitive Control during Color-Word Stroop. Journal of 824 Neuroscience, 41(35), 7388-7402.

5) This issue is most critical in the interpretation of the fMRI results as reflecting encoding of conflict types. A key limitation of the design, that is acknowledged by the authors is that conflict is fully confounded within-subject by spatial orientation. Indeed, the limited set of stimulus-response mappings also cast doubt on the underlying factors that give rise to the CSE modulations observed by the authors in their behavioral results. The CSE modulations are so strong - going from a complete absence of current x previous trial-type interaction in the cos(90) case all the way to a complete elimination of any current trial conflict when the prior trial was incongruent in the cos(0) case - that they cause suspicion that they are actually driven by conflict-related control adjustments rather than sequential dependencies in the stimulus-response mappings that can be associatively learned.

Unlike the fMRI data, we cannot tease apart the effects of conflict similarity and orientation in a similar manner as the cross-subject RSA for behavioral CSEs. However, we have a few reasons that the orientation and other bottom-up factors should not be the factors driving the similarity modulation effect.

First, we did not find any correlation between the regions showing orientation effects and behavioral CSEs. This suggests that orientation does not directly contribute to the CSE modulation.

Second, if the CSE modulation is purely driven by the association learning of the stimulus-response mapping, we should observe a stronger modulation effect after more extensive training. However, our results do not support this prediction. Using data from Experiment 1, we found that the modulation effect remained constant across the three sessions (see Note S3).

“Note S3. Modulation of conflict similarity on behavioral CSEs does not change across time We tested if the conflict similarity modulation on the CSE is susceptible to training. We collected the data of Experiment 1 across three sessions, thus it is possible to examine if the conflict similarity modulation effect changes across time. To this end, we added conflict similarity, session and their interaction into a mixed-effect linear model, in which the session was set as a categorical variable. With a post-hoc analysis of variance (ANOVA), we calculated the statistical significance of the interaction term. This approach was applied to both the RT and ER. Results showed no interaction effect in either RT, F(2,1479) = 1.025, p = .359, or ER, F(2,1479) = 0.789, p = .455. This result suggests that the modulation effect does not change across time. “

Third, the observed similarity modulation on the CSE, particularly for similarity levels 1-4, should not be attributed to the stimulus-response associations, such as feature integration, as have been addressed in response to comment 2.c.

Finally, other bottom-up factors, such as the spatial location proximity did not drive the CSE modulation results, which we have addressed in the original manuscript in Note S2.

"Note S2. Modulation of conflict similarity on behavioral CSEs cannot be explained by the physical proximity

In our design, the conflict similarity might be confounded by the physical proximity between stimulus (i.e., the arrow) of two consecutive trials. That is, when arrows of the two trials appear at the same quadrant, a higher conflict similarity also indicates a higher physical proximity (Fig. 1A). Although the opposite is true if arrows of the two trials appear at different quadrants, it is possible the behavioral effects can be biased by the within quadrant trials. To examine if the physical distance has confounded the conflict similarity modulation effect, we conducted an additional analysis.

We defined the physical angular difference across two trials as the difference of their polar angles relative to the origin. Therefore, the physical angular difference could vary from 0 to 180°. For each CSE conditions (i.e., CC, CI, IC and II), we grouped the trials based on their physical angular distances, and then averaged trials with the same previous by current conflict type transition but different orders (e.g., StHSmL−StLSmH and StLSmH−StHSmL) within each subject. The data were submitted to a mixed-effect model with the conflict similarity, physical proximity (i.e., the opposite of the physical angular difference) as fixed-effect predictors, and subject and CSE condition as random effects. Results showed significant conflict similarity modulation effects in both Experiment 1 (RT: β = 0.09 ± 0.01, t(7812) = 13.74, p < .001, ηp2 = .025; 875 ER: β = 0.09 ± 0.01, t(7812) = 7.66, p < .001, ηp2 = .018) and Experiment 2 (RT: β = 876 0.21 ± 0.02, t(3956) = 9.88, p < .001, ηp2 = .043; ER: β = 0.20 ± 0.03, t(4201) = 6.11, 877 p < .001, ηp2 = .038). Thus, the observed modulation of conflict similarity on behavioral 878 CSEs cannot be explained by physical proximity."

6) To their credit, the authors recognize this confound, and attempt to address it analytically through the use of a between-subject RSA approach. Yet the solution is itself problematic, because it doesn't actually deconfound conflict from orientation. In particular, the RSA model assumes that whatever components of neural activity encode orientation produce this encoding within the same voxellevel patterns of activity in each subject. If they are not (which is of course likely), then orthogonalization of these variables will be incomplete. Similar issues underlie the interpretation target/response and distractor coding. Given these issues, perhaps zooming out to a larger spatial scale for the between-subject RSA might be warranted. Perhaps whole-brain at the voxel level with a high degree of smoothing, or even whole-brain at the parcel level (averaging per parcel). For this purpose, Schaefer atlas parcels might be more useful than Glasser, as they more strongly reflect functional divisions (e.g., motor strip is split into mouth/hand divisions; visual cortex is split into central/peripheral visual field divisions). Similarly, given the lateralization of stimuli, if a within-parcel RSA is going to be used, it seems quite sensible to pool voxels across hemispheres (so effectively using 180 parcels instead of 360).

Doing RSA at the whole-brain level is an interesting idea. However, it does not allow the identification of specific brain regions representing the cognitive space. Additionally, increasing the spatial scale would include more voxels that are not involved in representing the information of interest and may increase the noise level of data. Given these concerns, we did not conduct the whole-brain level RSA.

We agree that smoothing data can decrease cross-subject variance in voxel distribution and may increase the signal-noise ratio. We reanalyzed the results for the right 8C region using RSA on smoothed beta maps (6-mm FWHM Gaussian kernel). This yielded a significant conflict similarity effect, t(951989) = 5.55, p < .0001, replicating the results on unsmoothed data (t(951989) = 5.60, p < .0001). Therefore, we retained the results from unsmoothed data in the main text, and added the results based on smoothed data to the supplementary material (Note S9).

“Note S9. The cross-subject pattern similarity is robust against individual differences Due to individual differences, the multivoxel patterns extracted from the same brain mask may not reflect exactly the same brain region for each subject. To reduce the influence of individual difference, we conducted the same cross-subject RSA using data smoothed with a 6-mm FWHM Gaussian kernel. Results showed a significant conflict similarity effect, t(951989) = 5.55, p < .0001, replicating the results on unsmoothed data (t(951989) = 5.60, p < .0001). “

We also used the bilateral 8C area as a single mask and conducted the same RSA. We found a significant conflict type similarity effect, t(951989) = 4.36, p < .0001. However, the left 8C alone showed no such representation, t(951989) = 0.38, p = .351, consistent with the right lateralized representation of cognitive space we reported in Note S8. Therefore, we used ROIs from each hemisphere separately.

“Note S8. The lateralization of conflict type representation

We observed the right 8C but not the left 8C represented the conflict type similarity. A further test is to show if there is a lateralization. We tested several regions of the left dlPFC, including the i6-8, 8Av, 8C, p9-46v, 46, 9-46d, a9-46v (Freund, Bugg, et al., 2021). We found that none of these regions show the representation of conflict type, all uncorrected ps > .35. These results indicate that the conflict type is specifically represented in the right dlPFC. “

We have also discussed the lateralization in the manuscript:

“In addition, we found no such representation in the left dlPFC (Note S8), indicating a possible lateralization. Previous studies showed that the left dlPFC was related to the expectancy-related attentional set up-regulation, while the right dlPFC was related to the online adjustment of control (Friehs et al., 2020; Vanderhasselt et al., 2009), which is consistent with our findings. Moreover, the right PFC also represents a composition of single rules (Reverberi et al., 2012), which may explain how the spatial Stroop and Simon types can be jointly encoded in a single space.”

7) The strength of the results is difficult to interpret due to the non-standard analysis method. The use of a mixed-level modeling approach to summarize the empirical similarity matrix is an interesting idea, but nevertheless is highly non-standard within RSA neuroimaging methods. More importantly, the way in which it was implemented makes it potentially vulnerable to a high degree of inaccuracy or bias. In this case, this bias is likely to be overly optimistic (high false positive rate). No numerical or formal defense was provided for this mixed-level model approach. As a result, the use of this method seems quite problematic, as it renders the strength of the observed results difficult to interpret. Instead, the authors are encouraged using a previously published method of conducting inference with between-subject RSA, such as the bootstrapping methods illustrated in Kragel et al. (2018; Nat Neurosci), or in potentially adopting one of the Chen et al. methods mentioned above, that have been extensively explored in terms of statistical properties.

No numerical or formal defense was provided for this mixed-level model approach. As a result, the use of this method seems quite problematic, as it renders the strength of the observed results difficult to interpret. Instead, the authors are encouraged using a previously published method of conducting inference with between-subject RSA, such as the bootstrapping methods illustrated in Kragel et al. (2018; Nat Neurosci), or in potentially adopting one of the Chen et al. methods mentioned above, that have been extensively explored in terms of statistical properties.

In our revised manuscript, we have adopted the approach proposed by Chen et al. (2017). Specifically, we included both the upper and lower triangle of the representational similarity matrix (excluding the diagonal). Moreover, we also removed all the within-subject similarity (thus also excluding the within-run similarity) to minimize the bias of the potentially strong within-subject similarity (note we also analyzed the within-subject data and found significant effects for the similarity modulation, though this effect cannot be attributed to the conflict similarity or orientation alone. We added this part in Note S7, see below). In addition, we added both the row-wise and column-wise random effects to capture the dependence of cells within each column/row (Chen et al., 2017). We have revised the method part as:

“We excluded within-subject cells from the RSM (thus also excluding the withinrun similarity as suggested by Walther et al., (2016)), and the remaining cells were converted into a vector, which was then z-transformed and submitted to a linear mixed effect model as the dependent variable. The linear mixed effect model also included regressors of conflict similarity and orientation similarity. Importantly, conflict similarity was based on how Simon and spatial Stroop conflicts are combined and hence was calculated by first rotating all subject’s stimulus location to the topright and bottom-left quadrants, whereas orientation was calculated using original stimulus locations. As a result, the regressors representing conflict similarity and orientation similarity were de-correlated. Similarity between two conditions was measured as the cosine value of the angular difference. Other regressors included a target similarity regressor (i.e., whether the arrow directions were identical), a response similarity regressor (i.e., whether the correct responses were identical); a spatial Stroop distractor regressor (i.e., vertical distance between two stimulus locations); a Simon distractor regressor (i.e., horizontal distance between two stimulus locations). Additionally, we also included a regressor denoting the similarity of Group (i.e., whether two conditions are within the same subject group, according to the stimulus-response mapping). We also added two regressors including ROImean fMRI activations for each condition of the pair to remove the possible uni-voxel influence on the RSM. A last term was the intercept. To control the artefact due to dependence of the correlation pairs sharing the same subject, we included crossed random effects (i.e., row-wise and column-wise random effects) for the intercept, conflict similarity, orientation and the group factors (G. Chen et al., 2017).”

Results from this approach highly replicated our original results. Specifically, we found the right 8C again showed a strong conflict similarity effect, a higher representational strength in the incongruent condition compared to the congruent condition, and a significant correlation with the behavioral CSE. The orientation effect was also identified in the visual (e.g., right V1) and oculomotor (e.g., left FEF) regions.

We revised the results accordingly:

For the conflict type effect:

“The first criterion revealed several cortical regions encoding the conflict similarity, including the Brodmann 8C area (a subregion of dlPFC(Glasser et al., 2016)) and a47r in the right hemisphere, and the superior frontal language (SFL) area, 6r, 7Am, 24dd, and ventromedial visual area 1 (VMV1) areas in the left hemisphere (Bonferroni corrected ps < 0.0001, one-tailed, Fig. 4A). We next tested whether these regions were related to cognitive control by comparing the strength of conflict similarity effect between incongruent and congruent conditions (criterion 2). Results revealed that the left SFL, left VMV1, and right 8C met this criterion, Bonferroni corrected ps < .05, one-tailed, suggesting that the representation of conflict type was strengthened when conflict was present (e.g., Fig. 4D). The intersubject brain-behavioral correlation analysis (criterion 3) showed that the strength of conflict similarity effect on RSM scaled with the modulation of conflict similarity on the CSE (slope in Fig. S2C) in right 8C (r = .52, Bonferroni corrected p = .002, onetailed, Fig. 4C, Table 1) but not in the left SFL and VMV1 (all Bonferroni corrected ps > .05, one-tailed). “

For the orientation effect:

“We observed increasing fMRI representational similarity between trials with more similar orientations of stimulus location in the occipital cortex, such as right V1, right V2, right V4, and right lateral occipital 2 (LO2) areas (Bonferroni corrected ps < 0.0001). We also found the same effect in the oculomotor related region, i.e., the left 997 frontal eye field (FEF), and other regions including the right 5m, left 31pv and right parietal area F (PF) (Fig. 5A). Then we tested if any of these brain regions were related to the conflict representation by comparing their encoding strength between incongruent and congruent conditions. Results showed that the right V1, right V2, left FEF, and right PF encoded stronger orientation effect in the incongruent than the congruent condition, Bonferroni corrected ps < .05, one-tailed (Table1, Fig. 5B). We then tested if any of these regions was related to the behavioral performance, and results showed that none of them positively correlated with the behavioral conflict similarity modulation effect, all uncorrected ps > .45, one-tailed. Thus all regions are consistent with the criterion 3.”

“Note S7. The cross-subject RSA captures similar effects with the within-subject RSA Considering the variability in voxel-level functional localizations among individuals, one may question whether the cross-subject RSA results were biased by the consistent multi-voxel patterns across subjects, distinct from the more commonly utilized withinsubject RSA. We reasoned that the cross-subject RSA should have captured similar effects as the within-subject RSA if we observe the conflict similarity effect in right 8C with the latter analysis. Therefore, we tested whether the representation in right 8C held for within-subject data. Specifically, we performed similar RSA for withinsubject RSMs, excluding the within-run cells. We replaced the perfectly confounded factors of conflict similarity and orientation with a common factor called similarity_orientation. Other confounding factor pairs (i.e., target versus response, and Stroop distractor versus Simon distractor) were addressed similarly. Results showed a significant effect of similarity_orientation, t(13993) = 3.270, p = .0005, 1tailed. Given the specific representation of conflict similarity identified by the crosssubject RSA, the within-subject data of right 8C may show similar conflict similarity modulation effects as the cross-subject data. Further research is needed to fully dissociate the representation of conflict and the representation of visual features such as orientation.”

8) Another potential source of bias is in treating the subject-level random effect coefficients (as predicted by the mixed-level model) as independent samples from a random variable (in the t-tests). The more standard method for inference would be to use test statistics derived from the mixed-model fixed effects, as those have degrees of freedom calculations that are calibrated based on statistical theory.

In our revised manuscript, we reported the statistical p values calculated from the mixed-effect models. Note that because we used the Chen et al. (2017) method, which includes data from the symmetric matrix, we corrected the degrees of freedom and estimated the true p values based on the t statistics of model results. For the I versus C comparison results, we calculated the p values by combining I and C RSMs into a larger model and then adding the condition type, as well as the interaction between the regressors of interest (conflict similarity and orientation) and the condition type. We made the statistical inference based on the interaction effect.

We have revised the corresponding methods as:

“The statistical significance of these beta estimates was based on the outputs of the mixed-effect model estimated with the “fitlme” function in Matlab 2022a. Since symmetric cells from the RSM matrix were included in the mixed-effect model, we adjusted the t and p values with the true degree of freedom, which is half of the cells included minus the number of fixed regressors. Multiple comparison correction was applied with the Bonferroni approach across all cortical regions at the p < 0.0001 level. To test if the representation strengths are different between congruent and incongruent conditions, we also conducted the RSA using only congruent (RDM_C) and incongruent (RDM_I) trials separately. The contrast analysis was achieved by an additional model with both RDM_C and RDM_I included, adding the congruency and the interaction between conflict type (and orientation) and congruency as both fixed and random factors. The difference between incongruent and congruent representations was indicated by a significant interaction effect.”

Reviewer #3:

Yang and colleagues investigated whether information on two task-irrelevant features that induce response conflict is represented in a common cognitive space. To test this, the authors used a task that combines the spatial Stroop conflict and the Simon effect. This task reliably produces a beautiful graded congruency sequence effect (CSE), where the cost of congruency is reduced after incongruent trials. The authors measured fMRI to identify brain regions that represent the graded similarity of conflict types, the congruency of responses, and the visual features that induce conflicts.

Using several theory-driven exclusion criteria, the authors identified the right dlPFC (right 8C), which shows 1) stronger encoding of graded similarity of conflicts in incongruent trials and 2) a positive correlation between the strength of conflict similarity type and the CSE on behavior. The dlPFC has been shown to be important for cognitive control tasks. As the dlPFC did not show a univariate parametric modulation based on the higher or lower component of one type of conflict (e.g., having more spatial Stroop conflict or less Simon conflict), it implies that dissimilarity of conflicts is represented by a linear increase or decrease of neural responses. Therefore, the similarity of conflict is represented in multivariate neural responses that combine two sources of conflict.

The strength of the current approach lies in the clear effect of parametric modulation of conflict similarity across different conflict types. The authors employed a clever cross-subject RSA that counterbalanced and isolated the targeted effect of conflict similarity, decorrelating orientation similarity of stimulus positions that would otherwise be correlated with conflict similarity. A pattern of neural response seems to exist that maps different types of conflict, where each type is defined by the parametric gradation of the yoked spatial Stroop conflict and the Simon conflict on a similarity scale. The similarity of patterns increases in incongruent trials and is correlated with CSE modulation of behavior.

We would like to thank the reviewer for the positive evaluation of our manuscript and for providing constructive comments. By addressing these comments, we believe that we have made our manuscript more accessible for the readers while also strengthening our findings. In particular, we have tested a few alternative models and confirmed that the cognitive space hypothesis best fits the data. We have also demonstrated the geometric properties of the cognitive space by examining the continuity and dimensionality of the space, further supporting our main arguments. We have incorporated revisions and additional analyses to the manuscript based on your feedback. Overall, we believe that these changes and additional analyses have significantly improved the manuscript. Please find our detailed responses below.

However, several potential caveats need to be considered.

1) One caveat to consider is that the main claim of recruitment of an organized "cognitive space" for conflict representation is solely supported by the exclusion criteria mentioned earlier. To further support the involvement of organized space in conflict representation, other pieces of evidence need to be considered. One approach could be to test the accuracy of out-of-sample predictions to examine the continuity of the space, as commonly done in studies on representational spaces of sensory information. Another possible approach could involve rigorously testing the geometric properties of space, rather than fitting RSM to all conflict types. For instance, in Fig 6, both the organized and domain-specific cognitive maps would similarly represent the similarity of conflict types expressed in Fig1c (as evident from the preserved order of conflict types). The RSM suggests a low-dimensional embedding of conflict similarity, but the underlying dimension remains unclear.

Following the reviewer’s first suggestion, we conducted a leave-one-out prediction approach to examine the continuity of the cognitive space. We used the behavioral data from Experiment 1 for this test, due to its larger amount of data than Experiment 2. Specifically, we removed data from one of the five similarity levels (as illustrated by the θs in Fig. 1C) and used the remaining data to perform the same mixed-effect model as reported in the main text (i.e., the two-stage analysis). This yielded one pair of beta coefficients including the similarity regressor and the intercept for each subject, with which we predicted the CSE for the removed similarity level at subject level. We repeated this process for each similarity level once. The predicted results were highly correlated with the original data, with r = .87 for the RT and r = .84 for the ER, ps < .001. We have added this analysis and result to the “Conflict type similarity modulated behavioral congruency sequence effect (CSE)” 1079 section:

“Moreover, to test the continuity and generalizability of the similarity modulation, we conducted a leave-one-out prediction analysis. We used the behavioral data from Experiment 1 for this test, due to its larger amount of data than Experiment 2. Specifically, we removed data from one of the five similarity levels (as illustrated by the θs in Fig. 1C) and used the remaining data to perform the same mixed-effect model (i.e., the two-stage analysis). This yielded one pair of beta coefficients including the similarity regressor and the intercept for each subject, with which we predicted the CSE for the removed similarity level for each subject. We repeated this process for each similarity level once. The predicted results were highly correlated with the original data, with r = .87 for the RT and r = .84 for the ER, ps < .001.”

To estimate if the domain-specific model could explain the results we observed in right 8C, we conducted a model-comparison analysis. The domain-specific model treats each conflict type differently, so we used a diagonal matrix, with within-conflict type similarities being 1 and all cross-conflict type similarities being 0. This model showed non-significant effects (t(951989) = 0.84, p = .201) and poorer fit (BIC = 5377127) than the cognitive space model (t(951989) = 5.60, p = 1.1×10−8, BIC = 5377094). We also compared other alternative models and found the cognitive space model best fitted the data. We have included these results in the revised manuscript:

“To examine if the right 8C specifically encodes the cognitive space rather than the domain-general or domain-specific organizations, we tested several additional models (see Methods). Model comparison showed a lower BIC in the Cognitive-Space model (BIC = 5377094) than the Domain-General (BIC = 537127) or Domain-Specific (BIC = 537127) models. Further analysis showed the dimensionality of the representation in the right 8C was 1.19, suggesting the cognitive space was close to 1D. We also tested if the observed conflict similarity effect was driven solely by spatial Stroop or Simon conflicts, and found larger BICs for the models only including the Stroop similarity (i.e., the Stroop-Only model, BIC = 5377122) or Simon similarity (i.e., the Simon-Only model, BIC = 5377096). An additional Stroop+Simon model, including both StroopOnly and Simon-Only regressors, also showed a worse model fitting (BIC = 5377118). Moreover, we replicated the results with only incongruent trials, considering that the pattern of conflict representations is more manifested when the conflict is present (i.e., on incongruent trials) than not (i.e., on congruent trials). We found a poorer fitting in Domain-general (BIC = 1344129), Domain-Specific (BIC = 1344129), Stroop-Only (BIC = 1344128), Simon-Only (BIC = 1344120), and Stroop+Simon (BIC = 1344157) models than the Cognitive-Space model (BIC = 1344104). These results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. The more detailed model comparison results are listed in Table 2.”

We also estimated the dimensionality of the right 8C with the averaged RSM and found the dimensionality of the cognitive space was ~ 1.19, very close to a 1D space. This result is consistent with our experimental design, as the only manipulated variable is the angular distance between conflict types. We have added these results and the methods to the revised manuscript.

Results:

“Further analysis showed the dimensionality of the representation in the right 8C was 1.19, suggesting the cognitive space was close to 1D.”

Methods:

“To better capture the dimensionality of the representational space, we estimated its dimensionality using the participation ratio (Ito & Murray, 2023). Since we excluded the within-subject cells from the whole RSM, the whole RSM is an incomplete matrix and could not be used. To resolve this issue, we averaged the cells corresponding to each pair of conflict types to obtain an averaged 5×5 RSM matrix, similar to the matrix shown in Fig. 1C. We then estimated the participation ratio using the formula:

where λi is the eigenvalue of the RSM and m is the number of eigenvalues.

2) Another important factor to consider is how learning within the confined task space, which always negatively correlates the two types of conflicts within each subject, may have influenced the current results. Is statistical dependence of conflict information necessary to use the organized cognitive space to represent conflicts from multiple sources? Answering this question would require a paradigm that can adjust multiple sources of conflicts parametrically and independently. Investigating such dependencies is crucial in order to better understand the adaptive utility of the observed cognitive space of conflict similarity.

As the central goal of our design was to test the geometry of neural representations of conflict, we manipulated the conflict similarity. The anticorrelated Simon and spatial Stroop conflict aimed to make the overall magnitude of conflict similar among different conflict types. We agree that with the current design the likely cognitive space is not a full 2D space with Simon and spatial Stroop being two dimensions. Instead, the likely cognitive space is a subspace (e.g., a circle) embedded in the 2D space, due to the constraint of anticorrelated Simon and spatial Stroop conflict across conflict types. Nevertheless, the subspace can also be used to test the geometry that similar conflict types share similar neural representations.

To test the full 2D cognitive space, a possible revision of our current design is to have multiple hybrid conditions (like Type 2-4) that cover the whole space. For instance, imagine arrow locations in the first quadrant space. We could have a 3×3 design with 9 conflict conditions, where their horizontal/vertical coordinates could be one of the combinations of 0, 0.5 and 1. This way, the spatial Stroop and Simon conditions would be independent of each other. Notably, however, one potential confounding factor would be that these conditions have different levels of difficulty (i.e., different magnitude of conflict), which may affect the CSE results and their representational similarity.<br /> We have added the above limitations and future designs to the revised 1156 manuscript.

“Another limitation is that in our design, the spatial Stroop and Simon effects are highly anticorrelated. This constraint may make the five conflict types represented in a unidimensional space (e.g., a circle) embedded in a 2D space. Future studies may test the 2D cognitive space with fully independent conditions. A possible improvement to our current design would be to include left, right, up, and down arrows presented in a grid formation across four spatially separate quadrants, with each arrow mapped to its own response button. However, one potential confounding factor would be that these conditions have different levels of difficulty (i.e., different magnitude of conflict), which may affect the CSE results and their representational similarity.”

Major comments:

3) The RSM result (and the absence of univariate effect) seem to be a good first step to claim the use of cognitive space of conflict. Yet, the presence of an organized (unidimensional; Fig. 6) and continuous cognitive space should be further tested and backed up.

We thank the reviewer for recognizing the methods and results of our current work. Indeed, the utilization of a parametric design and RSA to examine organization of neural representations is a widely embraced methodology in the field of cognitive neuroscience (e.g., Freund et al., 2021; Ritz et al., 2022). Our current study aimed primarily to provide original evidence for whether similar conflicts are represented similarly in the brain, which reflects the geometry of conflict representations (i.e., the structure of differences between conflict representations). We have used multiple criteria to back up the findings by showing the representation is sensitive to the presence of conflict and has behavioral relevance.

We agree that the cognitive space account of cognitive control requires further validation. Therefore, in the revised manuscript, we have added several additional tests to strengthen the evidence supporting the organized cognitive space representation. Firstly, we tested five alternative models (Domain-General, Domain Specific, Stroop-Only, Simon-Only and Stroop+Simon models), and found that the Cognitive-Space model best fitted our data. Secondly, we explicitly calculated the dimensionality of the representation and observed a low dimensionality (1.19D). We have added these results to the “Multivariate patterns of the right dlPFC encodes the conflict similarity” section in the revised manuscript (see also the response to Comment 1).

Furthermore, we utilized data from Experiment 1 to demonstrate the continuity of the cognitive space by showing its ability to predict out-of-sample data. We have included this result to the “Conflict type similarity modulated behavioral congruency sequence effect (CSE)” section in the revised manuscript:

“Moreover, to test the continuity and generalizability of the similarity modulation, we conducted a leave-one-out prediction analysis. We used the behavioral data from Experiment 1 for this test, due to its larger amount of data than Experiment 2. Specifically, we removed data from one of the five similarity levels (as illustrated by the θs in Fig. 1C) and used the remaining data to perform the same mixed-effect model (i.e., the two-stage analysis). This yielded one pair of beta coefficients including the similarity regressor and the intercept for each subject, with which we predicted the CSE for the removed similarity level for each subject. We repeated this process for each similarity level once. The predicted results were highly correlated with the original data, with r = .87 for the RT and r = .84 for the ER, ps < .001.”

References:

Freund, M. C., Bugg, J. M., & Braver, T. S. (2021). A Representational Similarity Analysis of Cognitive Control during Color-Word Stroop. Journal of Neuroscience, 41(35), 7388-7402.

Ritz, H., & Shenhav, A. (2022). Humans reconfigure target and distractor processing to address distinct task demands. bioRxiv. doi:10.1101/2021.09.08.459546

4) Is the conflict similarity effect not driven by either coding of the weak to strong gradient of the spatial Stroop conflict or the Simon conflict? For example, would simply identifying brain regions that selectively tuned to the Simon conflict continuously enough to create a graded similarity in Fig. C.

We recognize that our current design and analyzing approach cannot fully exclude the possibility that the current results are driven solely by either Stroop or Simon conflicts, since their gradients are correlated to the conflict similarity gradient we defined. To estimate their unique contributions, we performed a model-comparison analysis. We constructed a Stroop-Only model and a Simon-Only model, with each conflict type projected onto the Stroop (vertical) axis or Simon (horizontal) axis, respectively. The similarity between any two conflict types was defined using the Jaccard similarity index (Jaccard, P., 1901), that is, their intersection divided by their union. By replacing the cognitive space-based conflict similarity regressor with the Stroop-Only and Simon-Only regressors, we calculated their BICs. Results showed that the BIC was larger for Stroop-Only (5377122) and Simon-Only (5377096) than for the cognitive space model (5377094). An additional Stroop+Simon model, including both Stroop-Only and Simon-Only regressors, also 1220 showed a poorer model fitting (BIC = 5377118) than the cognitive space model.

Moreover, we replicated the results with only incongruent trials. We found a poorer fitting in Stroop-Only (BIC = 1344128), Simon-Only (BIC = 1344120), and Stroop+Simon (BIC = 1344157) models than the Cognitive-Space model (BIC = 1344104). These results indicate that the right 8C encodes an integrated cognitive space for resolving Stroop and Simon conflicts. Therefore, we believe the cognitive space has incorporated both dimensions. We added these additional analyses and results to the revised manuscript (see also the response to the above Comment 1).

5) Is encoding of conflict similarity in the unidimensional organized space driven by specific requirements of the task or is this a general control strategy? Specifically, is the recruitment of organized space something specific to the task that people are trained to work with stimuli that negatively correlate the spatial Stroop conflict and the Simon conflict?

We argue that this encoding is a general control strategy. In our task design, we asked the participants to respond to the target arrow and ignore the location that appeared randomly for them. So, they were not trained to deal with the stimuli in any certain way. We also found the conflict similarity modulation on CSE did not change with more training (We added this result in Note S3), indicating that the cognitive space did not depend on strategies that could be learned through training.

“Note S3. Modulation of conflict similarity on behavioral CSEs does not change across time We tested if the conflict similarity modulation on the CSE is susceptible to training. We collected the data of Experiment 1 across three sessions, thus it is possible to examine if the conflict similarity modulation effect changes across time. To this end, we added conflict similarity, session and their interaction into a mixed-effect linear model, in which the session was set as a categorical variable. With a post-hoc analysis of variance (ANOVA), we calculated the statistical significance of the interaction term.

This approach was applied to both the RT and ER. Results showed no interaction effect in either RT, F(2,1479) = 1.025, p = .359, or ER, F(2,1479) = 0.789, p = .455. This result suggests that the modulation effect does not change across time."

Instead, the cognitive space should be determined by the intrinsic similarity structure of the task design. A previous study (Freitas et al., 2015) has found that the CSE across different versions of spatial Stroop and flanker tasks was stronger than that across either of the two conflicts and Simon. In their designs, the stimulus similarity was controlled at the same level, so the difference in CSE was only attributable to the similar dimensional overlap between Stroop and flanker tasks, in contrast to the Simon task. Furthermore, recent studies showed that the cognitive space generally exists to represent structured latent states (e.g., Vaidya et al., 2022), mental strategy cost (Grahek et al., 2022), and social hierarchies (Park et al., 2020). Therefore, we argue that cognitive space is likely a universal strategy that can be applied to different scenarios.

We added this argument in the discussion:

“Although the spatial orientation information in our design could be helpful to the construction of cognitive space, the cognitive space itself was independent of the stimulus-level representation of the task. We found the conflict similarity modulation on CSE did not change with more training (see Note S3), indicating that the cognitive space did not depend on strategies that could be learned through training. Instead, the cognitive space should be determined by the intrinsic similarity structure of the task design. For example, a previous study (Freitas et al, 2015) has found that the CSE across different versions of spatial Stroop and flanker tasks was stronger than that across either of the two conflicts and Simon. In their designs, the stimulus similarity was controlled at the same level, so the difference in CSE was only attributable to the similar dimensional overlap between Stroop and flanker tasks, in contrast to the Simon task. Furthermore, recent studies showed that the cognitive space generally exists to represent structured latent states (e.g., Vaidya et al., 2022), mental strategy cost (Grahek et al., 2022), and social hierarchies (Park et al., 2020). Therefore, cognitive space is likely a universal strategy that can be applied to different scenarios."

Reference:

Freitas, A. L., & Clark, S. L. (2015). Generality and specificity in cognitive control: conflict adaptation within and across selective-attention tasks but not across selective-attention and Simon tasks. Psychological Research, 79(1), 143-162.

Vaidya, A. R., Jones, H. M., Castillo, J., & Badre, D. (2021). Neural representation of 1280 abstract task structure during generalization. Elife, 10, 1-26.

Grahek, I., Leng, X., Fahey, M. P., Yee, D., & Shenhav, A. Empirical and 1282 Computational Evidence for Reconfiguration Costs During Within-Task 1283 Adjustments in Cognitive Control. CogSci.

Park, S. A., Miller, D. S., Nili, H., Ranganath, C., & Boorman, E. D. (2020). Map 1285 Making: Constructing, Combining, and Inferring on Abstract Cognitive Maps. 1286 Neuron, 107(6), 1226-1238 e1228. doi:10.1016/j.neuron.2020.06.030

6) The observed pattern seems to suggest that there is conflict similarity space that is defined by the combination of the conflict similarity (i.e., the strength of conflicts) and the sources of conflict (i.e., the Simon vs the spatial Stroop). What are the rational reasons to separate conflicts of different sources (beyond detecting incongruence)? And how are they used for better conflict resolutions?

The necessity of separating conflicts of different sources lies in that the spatial Stroop and the Simon effects are resolved with different mechanisms. The behavioral congruency effects of a combined conflict from two different sources were shown to be the summation of the two conflict sources (Liu et al., 2010), suggesting that the conflicts are resolved independently. Moreover, previous studies have shown that different sources of conflict are resolved with different brain regions (Egner, 2008; Li et al., 2017), and at different processing stages (Wang et al., 2013). Therefore, when multiple sources of conflict occur simultaneously or sequentially, it should be more efficient to resolve the conflict by identifying the sources.

We have added this argument to the revised manuscript:

“The rationale behind defining conflict similarity based on combinations of different conflict sources, such as spatial-Stroop and Simon, stems from the evidence that these sources undergo independent processing (Egner, 2008; Li et al., 2014; Liu et al., 2010; Wang et al., 2014). Identifying these distinct sources is critical in efficiently resolving potentially infinite conflicts."

Reference:

Egner, T. (2008). Multiple conflict-driven control mechanisms in the human brain. Trends in Cognitive Sciences, 12(10), 374-380.

Li, Q., Yang, G., Li, Z., Qi, Y., Cole, M. W., & Liu, X. (2017). Conflict detection and 1307 resolution rely on a combination of common and distinct cognitive control networks. Neuroscience and Biobehavioral Reviews, 83, 123-131.

Wang, K., Li, Q., Zheng, Y., Wang, H., & Liu, X. (2014). Temporal and spectral 1310 profiles of stimulus-stimulus and stimulus-response conflict processing. NeuroImage, 89, 280-288.

Liu, X., Park, Y., Gu, X., & Fan, J. (2010). Dimensional overlap accounts for independence and integration of stimulus-response compatibility effects. Attention, Perception, & Psychophysics, 72(6), 1710-1720.

7) The congruency effect is larger in conflict type 2, 3, 4 consistently compared to conflict 1 and 5. Are these expected under the hypothesis of unified cognitive space of conflict similarity? Is the pattern of similarity modeled in RSA?

Yes, this is expected. The spatial Stroop and Simon effects have been shown to be additive and independent (Li et al., 2014). Therefore, the congruency effects of conflict type 2, 3 and 4 would be the weighted sum of the spatial Stroop and Simon effects. The weights can be defined by the sine and cosine of the polar angle.

For instance, in Type 2, wy = sin(67.5°) and wx = cos(67.5°). The sum of the two 1321 weight values (i.e., 1.31) is larger than 1, leading to a larger congruency effect than 1322 the pure spatial Stroop (Conf 1) and Simon (Conf 5) conditions.

Note that this hypothesis underlies the Stroop+Simon model, which assumes the Stroop and Simon dimensions are independently represented in the brain and drive the behavior in an additive fashion. Moreover, the observed difference of behavioral congruency effects may have reflected the variance in the Domain-General model, which treats all conflict types as equivalent, with the only difference between each two conflict types in the magnitude of their conflict. Therefore, we did not model the behavioral congruency effects as a covariance regressor in the major RSA. Instead, we conducted a model comparison analysis by comparing these models and the Cognitive-Space model. Results showed worse model fitting of both the Domain-general and Stroop+Simon models. Specially, the regressor of congruency effect difference in the Domain-General model was not significant (p = .575), which also suggests that the higher congruency effect in conflict type 2, 3 and 4 should not influence the Cognitive-Space model results. We have added these methods and results to the revised manuscript (see also our response to Comment 1):

Methods:

“Model comparison and representational dimensionality

To estimate if the right 8C specifically encodes the cognitive space, rather than the domain-general or domain-specific structures, we conducted two more RSAs. We replaced the cognitive space-based conflict similarity matrix in the RSA we reported above (hereafter referred to as the Cognitive-Space model) with one of the alternative model matrices, with all other regressors equal. The domain-general model treats each conflict type as equivalent, so each two conflict types only differ in the magnitude of their conflict. Therefore, we defined the domain-general matrix as the difference in their congruency effects indexed by the group-averaged RT in Experiment 2. Then the z scored model vector was sign-flipped to reflect similarity instead of distance. The domain-specific model treats each conflict type differently, so we used a diagonal matrix, with within-conflict type similarities being 1 and all cross-conflict type similarities being 0.

Moreover, to examine if the cognitive space is driven solely by the Stroop or Simon conflicts, we tested a spatial Stroop-Only (hereafter referred to as “Stroop-Only”) and a Simon-Only model, with each conflict type projected onto the spatial Stroop (vertical) axis or Simon (horizontal) axis, respectively. The similarity between any two conflict types was defined using the Jaccard similarity index (Jaccard, 1901), that is, their intersection divided by their union. We also included a model assuming the Stroop and Simon dimensions are independently represented in the brain, adding up the Stroop Only and Simon-Only regressors. We conducted similar RSAs as reported above, replacing the original conflict similarity regressor with the Strrop-Only, Simon-Only, or both regressors, and then calculated their Bayesian information criterions (BICs)."

Reference:

Li, Q., Nan, W., Wang, K., & Liu, X. (2014). Independent processing of stimulus stimulus and stimulus-response conflicts. PloS One, 9(2), e89249.

8) Please clarify the observed patterns of CSE effects in relation to the hypothesis of common cognitive space of conflict. In particular, right 8C shows that the patterns become dissimilar in incongruent trials compared to congruent trials. How does this direction of the effect fit to the common unidimensional cognitive space account? And how does such a representation contribute to the CES effects?

The behavioral CSE patterns provide initial evidence for the cognitive space hypothesis. Previous studies have debated whether cognitive control relies on domain-general or domain-specific representations, with much evidence gathered from behavioral CSE patterns. A significant CSE across two conflict conditions typically suggests domain-general representations of cognitive control, while an absence of CSE suggests domain-specific representations. The cognitive space view proposes that conflict representations are neither purely domain-general nor purely domain-specific, but rather exist on a continuum. This view predicts that the CSE across two conflict conditions should depend on the representational distance between them within this cognitive space. Our finding that CSE values systematically vary with conflict similarity level support this hypothesis. We have added this point in the discussion of the revised manuscript:

“Previous research on this topic often adopts a binary manipulation of conflict(Braem et al., 2014) (i.e., each domain only has one conflict type) and gathered evidence for the domain-general/specific view with presence/absence of CSE, respectively. Here, we parametrically manipulated the similarity of conflict types and found the CSE systematically vary with conflict similarity level, demonstrating that cognitive control is neither purely domain-general nor purely domain-specific, but can be reconciled as a cognitive space(Bellmund et al., 2018) (Fig. 6, middle).

Fig. 4D was plotted to show the steeper slope of the conflict similarity effect for incongruent versus congruent conditions. Note the y-aixs displays z-scored Pearson correlation values, so the grand mean of each condition was 0. The values for the first two similarity levels (level 1 and 2) were lower for incongruent than congruent conditions, seemingly indicating lower average similarity. However, this was not the case. The five similarity levels contained different numbers of data points (see Fig. 1C), so levels 4 and 5 should be weighted more heavily than levels 1 and 2. When comparing the grand mean of raw Pearson correlation values, the incongruent condition (0.0053) showed a tendency toward higher similarity than the congruent condition (0.0040), t(475998) = 1.41, p = .079. We have also plotted another version of Fig. 4D in Fig. S5, in which the raw Pearson correlation values were used.

The greater representation of conflict type in incongruent condition compared to congruent condition (as evidenced by a steeper slope) suggests that the conflict representation was driven by the incongruent condition. This is probably due to the stronger involvement of cognitive control in incongruent condition (than congruent condition), which in turn leads to more distinct patterns across different conflict types. This is consistent with the fact that the congruent condition is typically a baseline, where any conflict related effects should be weaker.

The representation of cognitive space may contribute to the CSE as a mental model. This model allows our brain to evaluate the cost and benefit associated with transitioning between different conflict conditions. When two consecutive trials are characterized by more similar conflict types, their representations in the cognitive space will be closer, resulting in a less costly transition. As a consequence, stronger CSEs are observed. We revised the corresponding discussion part as:

“Similarly, we propose that cognitive space could serve as a mental model to assist fast learning and efficient organization of cognitive control settings. Specifically, the cognitive space representation may provide a principle for how our brain evaluates the expected cost of switching and the benefit of generalization between states and selects the path with the best cost-benefit tradeoff (Abrahamse et al., 2016; Shenhav et al., 2013). The proximity between two states in cognitive space could reflect both the expected cognitive demand required to transition and the useful mechanisms to adapt from. The closer the two conditions are in cognitive space, the lower the expected switching cost and the higher the generalizability when transitioning between them. With the organization of a cognitive space, a new conflict can be quickly assigned a location in the cognitive space, which will facilitate the development of cognitive control settings for this conflict by interpolating nearby conflicts and/or projecting the location to axes representing different cognitive control processes, thus leading to a stronger CSE when following a more similar conflict condition.”

Minor comments:

9) Some of the labels of figure axes are unclear (e.g., Fig4C) about what they represent.

In Fig. 4C, the x-axis label is “neural representational strength”, which refers to the beta coefficient of the conflict type effect computed from the main RSA, denoting the strength of the conflict type representation in neural patterns. The y-axis label is “behavioral representational strength”, which refers to the beta coefficient obtained from the behavioral linear model using conflict similarity to predict the CSE in Experiment 2; it reflects how strong the conflict similarity modulates the behavioral 1440 CSE. We apologize for any confusion from the brief axis labels. We have added expanded descriptions to the figure caption of Fig. 4C.

Author response:

The following is the authors’ response to the original reviews.

Reviewer 1:

One concern is regarding the experimental task design. Currently, only subjective reports of interoceptive intensity are taken into account, the addition of objective behavioural measures would have given additional value to the study and its impact. 

To address this comment, we calculated interoceptive accuracy during the cardiorespiratory perturbation (isoproterenol) task according to our previous methods (e.g., Khalsa et al 2009 Int J Psychophys, Khalsa et al, 2015 IJED, Khalsa et al 2020 Psychophys, Hassanpour et al, 2018 NPP, Teed et al 2022 JAMA Psych). Thus, we quantified interoceptive accuracy as the cross-correlation between heart rate and real-time cardiorespiratory perception; specifically, the zero-lag cross-correlation between the heart rate and dial rating time series, and the maximum cross-correlation between these time series while allowing for different temporal delays (or lags). As expected, we found a dose-related increase in interoceptive accuracy from the 0.5mcg moderate perturbation dose (for which neuroimaging maps were not included in the current study) to the 2.0mcg high perturbation dose: zero-lag cross-correlations of 0.25 and 0.61, maximum cross-correlations of 0.41 and 0.73, for 0.5mcg and 2.0mcg doses, respectively, when averaged across all participants in the current study. Taking a closer examination at just the 2.0mcg dose, there were no group differences in zero-lag cross-correlation (t89\=-0.68, p=0.50) or maximum cross-correlation (t87\=-1.0, p=0.32) (depicted below, panel A). Furthermore, there were no associations between either of these interoceptive accuracy measures and the magnitude of activation within bilateral dysgranular convergent regions (F1\= 0.27 and 0.01, p=0.61 and 0.91, for the main effect of percent signal change on max and zero-lag cross-correlations, respectively; depicted below, panel B). When considering the significant correlation between the right insula signal intensity and subjective dial ratings, this lack of association with interoceptive accuracy suggests that the right dysgranular convergent insula was preferentially tracking the magnitude estimation rather than accuracy facet of interoceptive awareness during cardiorespiratory perturbation. Notably, during the saline placebo infusion, there were no systematic changes in heart rate and thus no systematic change in dial rating, precluding the calculation of the cross-correlation as a measure of interoceptive accuracy.

In reviewing these findings, we did not feel that the results add meaningful information to our interpretation of convergence, and accordingly we have chosen not to include it in the manuscript.

Author response image 1.

(A) Interoceptive accuracy during 2.0mcg isoproterenol perturbation, as measured by the maximum (left panel) and zero-lag (right panel) cross-correlation between the time series of heart rate and perceptual dial rating. There were no differences between groups. (B) There were no associations between interoceptive accuracy ratings and signal intensity within the convergence dysgranular insula during the Peak period of 2.0mcg perturbation. 

This brings me to my second concern. The authors mostly refer to their own previous work, without highlighting other methods used in the field. Some tasks measure interoceptive accuracy or other behavioural outcomes, instead of merely subjective intensity. Expanding the scientific context would aid the understanding and integration of this study with the rest of the field. 

Given our focus on the neural basis of bottom-up perturbations of interoception, we found it relevant to reference previous studies from our lab, as we built directly upon these previous findings to inform the hypotheses and design of the current experiment, but we can appreciate to provide a broader view of the literature. To expand the contextual frame, we have cited two fMRI meta-analyses of cardiac and gastrointestinal interoception (line 101). There are few studies that have used comparable perturbation approaches during neuroimaging in clinical populations, although we have referenced an exemplar study from the respiratory domain by Harrison et al (2021) in the discussion (line 612). In considering this comment more carefully, we felt that expanding the context further to other task-based methods or behavioral outcomes would shift the focus beyond our emphasis on the insular cortex and top-down/bottom-up convergence, though we have previously discussed and integrated such approaches (e.g., Khalsa & Lapidus, 2016 Front Psych, Khalsa et al, 2018 Biol Psychiatry CNNI, Khalsa et al 2022, Curr Psych Rep).

Lastly, the suggestions for future research lack substance compared to the richness of the discussion. I recommend a slight revision of the introduction/discussion. There is text in the discussion (explanatory or illuminating) which is better suited to the introduction. 

When discussing our study limitations (beginning line 732), we offer numerous areas for future research including different preprocessing pipelines, more sophisticated analysis techniques (such as multivariate pattern analysis) that would allow for individual-level inferences regarding convergent patterns of activation within the insula. However, we have revised the last sentence of our limitations paragraph (line 757), and have added more specificity regarding future approaches examining insular and whole-brain interoceptive signal flow.

Reviewer 2:

(1) The interpretation of the resting-state data is not quite as clear-cut as the task-based data - as presented currently, changes could potentially represent fluctuations over time rather than following interoception specifically. In contrast, much stronger conclusions can be drawn from the authors' task-based data. …I was also unsure about the interpretation of the resting state analysis (Figure 5), as there was no control condition without interoceptive tasks, meaning any change could represent a change over time that differed between groups and not necessarily a change from pre- to post-interoception. Relatedly I wondered if the authors had calculated the test-retest reliability of the resting state data (e.g. intraclass correlation coefficients for the whole-brain functional connective of convergent dysgranular insula subregions and left middle frontal gyrus before vs. after the tasks), as it would be generally useful for the field to know its stability. 

We have acknowledged the lack of a control condition in the isoproterenol task (note that the VIA task contained an exteroceptive trial that was included in the brain image contrast analysis). We have also provided further justification for our approach in both the Methods (see the first paragraph “fMRI resting state analysis” subsection) and Results (see the last paragraph of the “Convergence analysis” subsection). We cannot estimate test-retest reliability from the current dataset, given that we do not have resting state scans separated by a similar time frame without the performance of the interoceptive tasks in between (this is now clarified in line 346).

(2) The transdiagnostic sample could be better characterised in terms of diagnostic information, and was almost entirely female; it is also unclear what the effect of psychotropic medications may have been on the results given the effects of (e.g.) serotonergic medication on the BOLD signal. …Table 1 would be substantially improved by a fuller clinical characterisation of the specific sample included in the analysis - the diagnostic acronyms included in the table caption are not used in the table itself at present and would be an excellent addition, describing, for example, the demographics and symptom scores of patients meeting criteria for MDD, GAD, and AN (and perhaps those meeting criteria for more than 1). Similarly, additional information about the specific medications patients (or controls?) were taking in this study would be welcome (given the potential influences of common medications (e.g. antidepressants) on neurovascular coupling). 

We have expanded Table 1 to include more specific diagnostic information for the transdiagnostic ADE group (GAD, MDD, and/or AN, as well as other psychiatric diagnoses). We have also included medication use.  

Finally, Figures 7c and 7d would be greatly improved by showing individual data points if possible, and there may be a typo in the caption 'The cardiac group reported higher cardiac intensity ratings in the ADE group'.

We have adjusted Figure 7c and 7d to include individual data points, as we agree that this provides greater transparency to the data itself. We have also fixed the typo in the figure caption.

(3) As the authors point out, there may have been task-specific preprocessing/analysis differences that influenced results, for example, due to physiological correction in one but not both tasks. Although I note this is mentioned in the limitations, it was not clear to me why physiological noise was removed from the ISO task and whether it would be possible to do the same in the VIA task, which could be important for the most robust comparison of the two. 

In this study, we intentionally chose different task-specific preprocessing pipelines so we could ensure that our results were not simply due to new ways of handling the data. This would allow us to evaluate evidence of replicating the previous group-level findings of insular activation that informed the current approach and hypotheses. We agree that a harmonized approach is also merited, and in a subsequent project using this dataset, we have matched preprocessing pipelines for a connectivity-based analysis, to best facilitate comparison across tasks. We look forward to sharing those results with the scientific community in due time.

Reviewer 3:

Maybe I missed it (and my apologies in case I did), but there were a few instances where it was not entirely clear whether differential effects (say between groups or conditions) were compared directly, as would be required. One example is l. 459 ff: The authors report the interesting lateralisation effect for the two interception tasks and say it was absent in the exteroceptive VIA task. As a reader, it would be great to know whether that finding (effect in one condition but not in the other) is meaningful, i.e. whether the direct comparison becomes statistically significant. … The same applies to later comparisons, for example, the correlations reported in l. 465 ff (do these differ from one another?) as well as the FC patterns reported in l. 476 ff - again, there is a specific increase in the ADE group (but not in the HC), but is this between-group difference statistically meaningful? 

Thank you for these questions. We have added greater detail in the Results section in order to increase clarity regarding which statistical comparisons support which conclusions. Generally, we limited our comparisons to the effect of group, as comparing ADE vs. HC individuals was of primary interest, and in some cases also the effect of hemisphere and epoch. However, we did not perform exhaustive comparisons for all measures, in the interest of keeping the focus of our multi-level multi-task analysis on the hypothesis-driven questions specifically related to convergence of top-down and bottom-up processing.

Regarding the comment asking if we could compare the lateralization effect directly across task conditions (i.e., is there a greater difference between hemispheres in the ISO task compared to VIA?): unfortunately, directly comparing signal intensity across tasks is not possible because the isoproterenol infusion induces physiological changes that can cause some dose-related signal reduction (we have attempted to address this in the past, e.g., Hassanpour et al, 2018 HumBrMapp). Consequently, our conclusions about spatial localization of top-down and bottom-up convergence are limited to group-level comparisons based on binary activation.

(2) A second 'major' relates to the intensity ratings (l. 530 ff). I found it very interesting that the ADE group reported higher cardiac, but lower exteroceptive intensity ratings during the VIA task. I understand the authors' approach to collapse within the ADE group, but it would be great to know which subgroup of patients drives this differential effect. It could be the case that the cardiac effect is predominantly present in the anxiety group, while the lower exteroceptive ratings are driven by the depression patients. Even if that were not the case, it would be highly instructive to understand the rating pattern within the anxiety group in greater detail. Do these patients 'just' selectively upregulate interoception, or is there even a perceived downregulation of exteroceptive signalling? 

We have depicted these data below for reviewers’ reference, showing individual responses for each group (HC and ADE; panel A), as well as the ADE individuals separated by primary diagnosis (GAD = generalized anxiety disorder, n=24; AN = anorexia nervosa, n=16; MDD = major depressive disorder, n=6; panel B). When tested via linear regression, we found no differences in ratings across ADE subgroups (rating ~ subgroup * condition, F3\=1.71, p=0.16 for main effect of subgroup). However, several factors should be considered in interpreting this result: first, all subgroups are small, particularly the MDD sample. Second, while these diagnostic labels refer to the most prominent symptom expression of each patient, every clinical participant in the study had a co-morbid disorder. Therefore, it is not possible to isolate disorder-specific pathology from our multi-diagnostic sample, and for this reason we refrained from including the subgroup-specific data in the manuscript.

Author response image 2.

(A) Post-trial ratings during the Visceral Interoceptive attention task, for reference. This is also shown in Figure 7D. (B) The same post-trial ratings in (A), but with the ADE group separated by primary diagnoses. Importantly, although assigned to one diagnostic category on the basis of most prominent symptom expression, most patients had one or more comorbidities across disorders. GAD = Generalized Anxiety Disorder. MDD = major depressive disorder. AN = anorexia nervosa. HC = healthy comparison.

l. 86: 'Conscious experience' of what, precisely? During the first round of reading, I was wondering about the extent to which consciousness as a general concept will play a role, which could be misleading. 

We have changed it to “conscious experience of the inner body” in the text. The current study is limited in scope to the neurobiology of conscious perceptions of the inner body, not consciousness as a general phenomenon. We hope this distinction is now clear.

l.115: Particularly given the focus on predictive processing, I was wondering whether the (slightly outdated) spotlight metaphor is really needed here. 

While not perfect, we believe it is still valid to metaphorically reference goal-directed attention towards the body as an “attentional spotlight”. Given the concern, we have minimized the focus on this metaphor, and the sentence now reads as follows:

“Extending beyond these model-based influences are goal-directed activities (also described previously as the ‘attentional spotlight’ effect ((Brefczynski and DeYoe 1999)), whereby focusing voluntary attention towards certain environmental signals not only alters their conscious experience but selectively enhances neural activity in the responsive area of cortex.”

l. 129 ff: The sentence has three instances of 'and' in it, most likely a typo. 

We have fixed this in the text.

l. 245: What do these ratings correspond to, i.e. what was the precise question/instruction? 

The instructions for subjective ratings in each task are mentioned in the Methods (line 223 for ISO task, line 249 for the VIA task), and we have added more detail regarding the scale used to collect subjective intensity ratings.

l. 322: Could you provide the equation of the LMEM in the main text? It would be interesting to know e.g. whether participants/patients were included as a random effect. 

We have provided this equation in the Methods (line 326).

l. 418 ff: I was confused about the statistical approach here. Why use separate t-tests instead of e.g. another LMEM which would adequately model task and condition factors? 

We did not use t-tests, but instead used linear regression to look at differences in agranular PSC across groups, hemispheres, and epochs, as well as potential associations between PSC and trait measures. We have adjusted the wording in this Methods paragraph (line 418) to help clarity.

l. 425: As a general comment, it would be great to provide the underlying scripts openly through GitHub, OSF, ... 

We agree with this comment, and our main analysis scripts have been posted on our OSF as an addition to the original preregistration of this work (https://osf.io/6nxa3/).

l. 443: For consistency, please report the degrees of freedom for the X² test.

l. 454: ... and the F statistic would require two degrees of freedom (only the second is reported).

l. 523: The t value is reported without degrees of freedom here (but has them in other instances).

l. 540: Typo ('were showed').

We have reported degrees of freedom for all statistics.

Author response:

The following is the authors’ response to the original reviews

Reviewer 1:

(1) In general, the representation of target and distractor processing is a bit of a reach. Target processing is represented by SSVEP amplitude, which is most likely going to be related to the contrast of the dots, as opposed to representing coherent motion energy, which is the actual target. These may well be linked (e.g., greater attention to the coherent motion task might increase SSVEP amplitude), but I would call it a limitation of the interpretation. Decoding accuracy of emotional content makes sense as a measure of distractor processing, and the supplementary analysis comparing target SSVEP amplitude to distractor decoding accuracy is duly noted.

We agree with the reviewer. The SSVEP amplitude of the target at the whole trial level indeed reflected the combined effect of the stimulus parameters (e.g., contrast of the moving dots) as well as attention. However, the time course of the target SSVEP amplitude within a trial, derived from the moving window analysis, reflected the temporal fluctuations of target processing, since the stimulus parameters remained the same during the trial. We now make this clearer in the revised manuscript.

(2) Comparing SSVEP amplitude to emotional category decoding accuracy feels a bit like comparing apples with oranges. They have different units and scales and probably reflect different neural processes. Is the result the authors find not a little surprising in this context? This relationship does predict performance and is thus intriguing, but I think this methodological aspect needs to be discussed further. For example, is the phase relationship with behaviour a result of a complex interaction between different levels of processing (fundamental contrast vs higher order emotional processing)?

Traditionally, the SSVEP amplitude at the distractor frequency is used to quantify distractor processing. Given that the target SSVEP amplitude is stronger than that of the distractor, it is possible that the distractor SSVEP amplitude is contaminated by the target SSVEP amplitude due to spectral power leakage; see Figure S4 for a demonstration of this. Because of this issue we therefore introduced the use of decoding accuracy as an index of distractor processing. The lack of correlation between the distractor SSVEP amplitude and the distractor decoding accuracy, although it is kind of like comparing apples with oranges as pointed out by the reviewer, serves the purpose of showing that these two measures are not co-varying, and the use of decoding accuracy is free from the influence of the distractor SSVEP amplitude which is influenced by the target SSVEP amplitude. Also, to address the apples-vs-oranges issue, the correlation was computed on normalized time series, in which a z-score time series replaced the original time series so that the correlated variables are dimensionless. Regarding the question of assessing the relation between behavior and different levels of processing, we do not have means to address it, given that we are not able to empirically separate the effects of stimulus parameters versus attention.

Reviewer 2:

(1) Incomplete Evidence for Rhythmicity at 1 Hz: The central claim of 1 Hz rhythmic sampling is insufficiently validated. The windowing procedure (0.5s windows with 0.25s step) inherently restricts frequency resolution, potentially biasing toward low-frequency components like 1 Hz. Testing different window durations or providing controls would significantly strengthen this claim.

We appreciate the reviewer’s insightful suggestion. In response, we tested different windowing parameters, e.g., 0.1s sliding window with a 0.05s step size. Figure S5 demonstrates that the strength of both target and distractor processing fluctuates around ~1 Hz, both at the individual and group levels. Additionally, Figures S6(A) and S6(B) show that the relative phase between target and distractor processing time series exhibits a uniform distribution across subjects. In terms of the relation between relative phase and behavior, Figure S6(C) illustrates two representative cases: a high-performing subject with 84.34% task accuracy exhibited a relative phase of 0.9483π (closer to π), while a low-performing subject with 30.95% accuracy showed a phase of 0.29π close to 0). At the group level, a significant positive correlation between relative phase and task performance was found (r = 0.6343, p = 0.0004), as shown in Figure S6(D). All these results, aligning closely with our original findings (0.5s window length and 0.25s step size), suggest that the conclusions are not dependent on windowing parameters. We discuss these results in the revised manuscript.

To further validate our findings, we also employed the Hilbert transform to extract amplitude envelopes of the target and distractor signals on a time-point-by-time-point basis, providing a window-free estimate of signal strength (Figures R3 and R4). The results remain consistent with both the original findings and the new sliding window analyses (Figure S6). Specifically, Figure S7 reveals ~1 Hz fluctuations in target and distractor processing at both individual and group levels. Figures S8(A) and S8(B) confirm a uniform distribution of the relative phase across subjects. In Figure S8(C), the relative phase was 0.9567π for a high-performing subject (84.34% accuracy) and 0.2247π for a low-performing subject (28.57% accuracy). At the group level, a significant positive correlation was again observed between relative phase and task performance (r = 0.4020, p = 0.0376), as shown in Figure S8(D).

(2) No-Distractor Control Condition: The study lacks a baseline or control condition without distractors. This makes it difficult to determine whether the distractor-related decoding signals or the 1 Hz effect reflect genuine distractor processing or more general task dynamics.

The lack of a no-distractor control condition is certainly a limitation and will be acknowledged as such in the revised manuscript. However, given that our decoding results are between two different classes of distractors, we are confident that they reflect distractor processing.

(3) Decoding Near Chance Levels: The pairwise decoding accuracies for distractor categories hover close to chance (~55%), raising concerns about robustness. While statistically above chance, the small effect sizes need careful interpretation, particularly when linked to behavior.

This is an important point. To test robustness, we have implemented a random permutation procedure in which trial labels were randomly shuffled to construct a nullhypothesis distribution for decoding accuracy. We then compared the decoding accuracy from the actual data to this distribution. Figure S9 shows the results based on 1,000 permutations. For each of the three pairwise classifications—pleasant vs. neutral, unpleasant vs. neutral, and pleasant vs. unpleasant—as well as the three-way classification, the actual decoding accuracies fall far outside the null-hypothesis distribution (p < 0.001), and the effect size in all four cases is extremely large. These findings indicate that the observed decoding accuracies are statistically significant and robust in terms of both statistical inference and effect size.

(4) No Clear Correlation Between SSVEP and Behavior: Neither target nor distractor signal strength (SSVEP amplitude) correlates with behavioral accuracy. The study instead relies heavily on relative phase, which - while interesting - may benefit from additional converging evidence.

We felt that what the reviewer pointed out is actually the main point of our study, namely, it is not the target or distractor strength over the whole trial that matters for behavior, it is their temporal relationship within the trial that matters for behavior. This reveals a novel neuroscience principle that has not been reported in the past. We have stressed this point further in the revised manuscript.

(5) Phase-analysis: phase analysis is performed between different types of signals hindering their interpretability (time-resolved SSVEP amplitude and time-resolved decoding accuracy).

The time-resolved SSVEP amplitude is used to index the temporal dynamics of target processing whereas the time-resolved decoding accuracy is used to index the temporal dynamics of distractor processing. As such, they can be compared, using relative phase for example, to examine how temporal relations between the two types of processes impact behavior. This said, we do recognize the reviewer’s concern that these two processes are indexed by two different types of signals. We thus normalized each time course using zscoring, making them dimensionless, and then computed the temporal relations between them.

Appraisal of Aims and Conclusions:

The authors largely achieved their stated goal of assessing rhythmic sampling of distractors. However, the conclusions drawn - particularly regarding the presence of 1 Hz rhythmicity - rest on analytical choices that should be scrutinized further. While the observed phaseperformance relationship is interesting and potentially impactful, the lack of stronger and convergent evidence on the frequency component itself reduces confidence in the broader conclusions.

Impact and Utility to the Field:

If validated, the findings will advance our understanding of attentional dynamics and competition in complex visual environments. Demonstrating that ignored distractors can be rhythmically sampled at similar frequencies to targets has implications for models of attention and cognitive control. However, the methodological limitations currently constrain the paper's impact.

Thanks for these comments and positive assessment of our work’s potential implications and impact. As indicated above, in the revision process, we have carried out a number of additional analyses, some suggested by the reviewers, and the results of the additional analyses, now included in the Supplementary Materials, served to further validate the main findings and strengthen our conclusions.

Additional Context and Considerations:

(1) The use of EEG-fMRI is mentioned but not leveraged. If BOLD data were collected, even exploratory fMRI analyses (e.g., distractor modulation in visual cortex) could provide valuable converging evidence.

Indeed, leveraging fMRI data in EEG studies would be very beneficial, as has been demonstrated in our previous work. However, given that this study concerns the temporal relationship between target and distractor processing, it is felt that fMRI data, which is known to possess low temporal resolution, has limited potential to contribute. We will be exploring this rich dataset in other ways in the future, where we will be integrating the two modalities for more insights that are not possible with either modality used alone.

Author response image 1.

Appyling moving window analysis (0.02s window duration and 0.01 step size) to a different EEG-fMRI dataset. (A) The amplitude time series of the 4.29 Hz component and the Fourier spectrum. (B) The group level Fourier spectrum. At both individual and group level, no 1 Hz modulation is observed, suggesting that the 1 Hz modulation observed in our data is not introduced by the artifact removal procedure.

(2) In turn, removal of fMRI artifacts might introduce biases or alter the data. For instance, the authors might consider investigating potential fMRI artifact harmonics around 1 Hz to address concerns regarding induced spectral components.

We have done extensive work in the area of simultaneous EEG-fMRI and have not encountered artifacts with a 1Hz rhythmicity. Our scanner artifact removal procedure is very standardized. As such, it stands to reason that if the 1Hz rhythmicity observed here results from the artifact removal process, it should also be present in other datasets where the same preprocessing steps were implemented. We tested this using another EEG-fMRI dataset (Rajan et al., 2019) . Author response image 1 shows that the EEG power time series of the new dataset doesn't have 1 Hz rhythmicity, whether at the individual level or at the group level, suggesting that the 1 Hz rhythmicity reported in the manuscript is not coming from the removal of the scanner artifacts, but instead reflects true rhythmic sampling of stimulus information. Also, the fact that the temporal relations between target processing and distractor processing at 1Hz impact behavior is another indication that the 1Hz rhythmicity is a neuroscientific effect, not an artifact.

References

Rajan, A., Siegel, S. N., Liu, Y., Bengson, J., Mangun, G. R., & Ding, M. (2019). Theta Oscillations Index Frontal Decision-Making and Mediate Reciprocal Frontal–Parietal Interactions in Willed Attention. Cerebral Cortex, 29(7), 2832–2843. https://doi.org/10.1093/cercor/bhy149

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

This report contains two parts. In the first part, several experiments were carried out to show that CsoR binds to CheA, inhibits CheA phosphorylation, and impairs P. putida chemotaxis. The second part provides some evidence that CsoR is a copper-binding protein, binds to CheA in a copper-dependent manner, and regulates P. putida response to copper, a chemorepellent. Based on these results, a working model is proposed to describe how CsoR coordinates chemotaxis and resistance to copper in P. putida. While the second part of the study is relatively solid, there are some major concerns about the first part.

Critiques:

(1) The rigor from prior research is not clear. In addition to talking about other bacterial chemotaxis, the Introduction should briefly summarize previous work on P. putida chemotaxis and copper resistance.

We summarized previous results on P. putida copper resistance and added those results to the introduction section of the revised manuscript. As for chemotaxis, most studies in P. putida focused on the sensing/responding of the bacteria to different chemical compounds and the methyl-accepting chemotaxis proteins (MCPs) involved in the sensing, which is not relevant to the main content of this study. The component of the chemotaxis system in P. putida is similar to that in E. coli, and the signaling mechanism is presumably similar.

(2) The rationale for identifying those CheA-binding proteins is vague. CheA has been extensively studied and its functional domains (P1 to P5) have been well characterized. Compared to its counterparts from other bacteria, does P. putida CheA contain a unique motif or domain? Does CsoR bind to other bacterial CheAs or only to P. putida CheA?

The original purpose of the pull-down assay was to detect the interaction between CheA and c-di-GMP metabolizing enzymes, which was another project. However, we ignored that most c-di-GMP metabolizing enzymes were membrane proteins, and we made a mistake by using whole-cell lysate in the pull-down experiment. Thus, we failed to identify c-di-GMP metabolizing enzymes in “target” proteins of the pull-down assay. However, we found several novel “target” proteins in the pull-down assay. We wondered about the function of these proteins and the physiological roles of the interaction between CheA and these proteins, which was the primary purpose of this study. Although the function of CheA has been well characterized, most previous results focused on the role of CheA in chemotaxis, and its role in other bacterial processes was poorly studied. To extend our knowledge about CheA, we analyzed the results of the pull-down assay and decided to test the interaction between CheA and identified proteins, as well as the physiological roles of the interaction.

BLAST results showed that the CheA of P. putida shared 41.12% sequence similarity with the CheA of E. coli, and the CheA of P. putida had a similar domain pattern to those CheAs from other bacteria. To test whether  CsoR_{P. putida} interacted with CheA from other bacteria, we performed a BTH assay to investigate the interaction between  CsoR_{P. putida} and eight CheAs, including CheA from E. coli, CheA from A. caldus, CheA from B. diazoefficiens, CheA from B. subtilis, CheA from L. monocytogenes, CheA from P. fluorescens, CheA from P. syringae, and CheA from P. stutzeri. As shown in the following Fig. 1,  CsoR_{P. putida} could interact with CheA from A. caldus, B. subtilis, L. monocytogenes, P. fluorescens, P. syringae, and P. stutzeri. Besides, among these strains, cheA and csoR coexist in A. caldus, B. diazoefficiens, B. subtilis, L. monocytogenes, P. fluorescens, P. syringae, and P. stutzeri. We previously tested the interaction of the two proteins from these bacterial species. The results showed that the CheA-CsoR interaction existed between proteins from A. caldus, B. subtilis, P. syringae, and P. stutzeri (Fig. 7 in the manuscript). However, CheA and CsoR from B. diazoefficiens, L. monocytogenes, and P. fluorescens showed no apparent interaction (Fig. 7 in the manuscript). These results suggested that unique amino acid sequences in the two proteins might be required to achieve interaction.

(3) Line 133-136, "Collectively, using pull-down, BTH, and BiFC assays, we identified 16 new CheA-interacting proteins in P. putida." It is surprising that so many proteins were identified but none of them were chemotaxis proteins, in particular those known to interact with CheA, such as CheW, CheY and CheZ, which raises a concern about the specificity of these methods. BTH and BiFC often give false-positive results and thus should be substantiated by other approaches such as co-IP, surface plasmon resonance (SPR), or isothermal titration calorimetry (ITC) along with mutagenesis studies.

The response regulator CheY and the phosphatase CheZ (two proteins known to be associated with CheA) were identified in the pull-down assay (Table S1), and the two proteins showed high Log₂(fold change) values, indicating that they were obtained in the pull-down assay with high amount in the experimental group and low amount in the control group. Our study aimed to identify new CheA-interacting proteins; thus, the two proteins (CheY and CheZ) were not included in subsequent investigations. The CheA-interacting proteins were initially obtained through an in vitro assay (pull-down), followed by an in vivo assay (BTH and BiFC) to test the interaction further. Only proteins that showed positive results in all three assays were considered trustworthy CheA-interacting proteins and kept for further study.

(4) Line 147-149, "Fig. 2a, five strains (WT+pcsoR, WT+pispG, WT+pnfuA, WT+pphaD, and WT+pPP_1644) displayed smaller colony than the control strain (WT+pVec), indicating a weaker chemotaxis ability in these five strains." If copper is a chemorepellent, these strains should swim away from high concentrations of copper; thus, the sizes of colonies couldn't be used to measure this response. In the cited reference (reference 29), bacterial response to phenol was measured using a response index (RI).

Except for CsoR, the rest of the CheA-interacting proteins had no direct connection with copper and were involved in different processes (Table S1). A reasonable speculation is that these proteins involved in different processes can integrate signals from specific processes into chemotaxis by regulating CheA autophosphorylation, leading to better regulation of chemotaxis according to intracellular physiological state. We used semisolid nutrient agar plates to test and compare bacterial chemotaxis ability. In a fixed attractant/repellent gradient, chemokine, such as copper, can lead to two subpopulations traveling at different speeds, with the slower one being held back by the chemokinetic drift. In the case of semisolid plate migration, bacteria with chemotaxis ability formed large colonies by generating their gradient by consuming nutrients/producing toxic metabolic waste and following attractant/repellent gradients leading outward from the colony origin (Cremer et al., 2019. Nature 575:658–663). The observation of successive sharp circular bands (rings) progressing outward from the inoculation point was taken to confirm the chemotaxis genotype, and mutants without chemotaxis spread out uniformly and formed a small colony (Wolfe and Berg, PNAS. 1989, 86:6973-6977). In our experiment, we were unsure about the signals/chemokines of each target protein, so we could not design a fixed attractant/repellent gradient. Besides, all target proteins interacted with CheA, which is a crucial factor in chemotaxis, and we assume that these proteins would affect chemotaxis under overexpression conditions. Thus, we used semisolid nutrient plates to test and compare bacterial chemotaxis ability.

(5) Figures 2 and 3 show both CsoR and PhaD bind to CheA and inhibit CheA autophosphorylation. Do these two proteins share any sequence or structural similarity? Does PhaD also bind to copper? Otherwise, it is difficult to understand these results.

Thanks a lot. This is an enlightening comment. CsoR is a protein with a size of 10.8 kDa, and PhaD is 23.1 kDa. Because of the difference in size, we took it for granted that the two proteins were not similar. We recently compared their sequence on NCBI BLAST. Although both CsoR and PhaD are transcriptional regulators and interact with CheA, they have no significant sequence similarity. In terms of protein structure, we predicted their structures using AlphaFold. The results showed that CsoR consisted of three α-helixes and PhaD consisted of nine α-helixes (new Fig. S5a and S5b in the manuscript). We further compared their structure using Pymol but found no significant similarity between the two proteins (new Fig. S5c in the manuscript).

PhaD is a TetR family transcriptional regulator located adjacent to the genes involved in PHA biosynthesis, and it behaves as a carbon source-dependent activator of the pha cluster related to polyhydroxyalkanoates (PHAs) biosynthesis (de Eugenio et al., Environ Microbiol. 2010, 12:1591-1603; Tarazona et al., Environ Microbiol. 2020, 22:3922-3936). Bacterial PHAs are isotactic polymers synthesized under unfavorable growth conditions in the presence of excess carbon sources. PHAs are critical in central metabolism, acting as dynamic carbon reservoirs and reducing equivalents (Gregory et al., Trends Mol Med. 2022, 28:331-342). The interaction between PhaD and CheA leads us to speculate that there might be some connection between PHA synthesis and bacterial chemotaxis. For example, chemotaxis helps bacteria move towards specific carbon sources that favor PHA synthesis, and the interaction between PhaD and CheA weakens chemotaxis, causing bacteria to linger in areas rich in these carbon sources. This is an interesting hypothesis worth testing in the future.

(6) Line 195-196, "CsoR/PhaD had no apparent influence on the phosphate transfer between CheA and CheY". CheA controls bacterial chemotaxis through CheY phosphorylation. If this is true, how do CsoR and PhaD affect chemotaxis?

During the autophosphorylation assay, CheA was mixed with CsoR/PhaD and incubated for about 10 min before adding [³²P]ATP[γP]. Thus, the effect of CsoR/PhaD on CheA autophosphorylation happened through the assay, and a significant inhibition effect was observed in the final result. Regarding transphosphorylation, CheA was mixed with ATP and incubated for about 30 min, at which time the autophosphorylation of CheA happened. Then, CsoR/PhaD and CheY were added to the phosphorylated CheA to investigate transphosphorylation. CsoR and PhaD affected chemotaxis via inhibiting CheA autophosphorylation, which was a crucial step in chemotaxis signaling, and the decrease in CheA autophosphorylation caused decreased chemotaxis.

(7) Figure 3 shows that CsoR/PhaD bind to CheA through P1, P3, and P4. This result is intriguing. All CheA proteins contain these three domains. If this is true, CsoR/PhaD should bind to other bacterial CheAs too. That said, this experiment is premature and needs to be confirmed by other approaches.

As replied to comment (2) above, we performed a BTH assay to investigate whether  CsoR_{P. putida} interacts with CheA from other bacterial species. The results revealed that  CsoR_{P. putida} interacted with CheA from A. caldus, B. subtilis, L. monocytogenes, P. fluorescens, P. syringae, and P. stutzeri, but not with CheA from E. coli and B. diazoefficiens. This result suggested that CheA-CsoR interaction required specific/unique amino acid sequence patterns in the two proteins, and similar domain composition alone was insufficient.

(8) Figure 5, does PhaD contain these three residues (C40, H65, and C69)? If not, how does PhaD inhibit CheA autophosphorylation and chemotactic response to copper?

No, there is no significant sequence similarity between PhaD and CsoR, and PhaD contains none of the three residues of CsoR (C40, H65, and C69). The size of the two proteins is also quite different (CsoR 10.8 kDa, PhaD 23.1 kDa). The structure alignment also revealed no apparent similarity between the predicted structures of PhaD and CsoR (new Fig. S5c in the manuscript). Nevertheless, CsoR and PhaD interacted with CheA through its P1, P3, and P4 domains. It is interesting how the two proteins interacted with CheA, but we currently have no answer.

(9) Does deletion of cosR or cheA have any impact on P. putida resistance to high concentrations of copper?

No, deletion of cosR/cheA had no noticeable impact on P. putida's resistance to high concentrations of copper. We performed a growth assay to test the effect of CsoR and CheA on copper resistance under both liquid and solid medium conditions. The copper concentration was set at 0, 200, 500, 1000 μM. With the increase of copper concentration, the growth of bacteria was gradually inhibited, but the growth trends of csoR mutant, cheA mutant, and complementary strains were similar to that of the wild-type strain (new Fig. S6b and S6c in the manuscript). We speculated that this might be attributed to CsoR being a repressor and inhibiting gene expression in the absence of copper. When copper existed, the inhibitory effect of CsoR was relieved, which is the same as that in the csoR mutant. Besides, although deletion of cosR led to a slight increase (about 1.3-fold) in the expression of copper resistance genes (Fig. 4b in the manuscript), its effect on gene expression was much weaker than its homologous protein in other bacterial species. In M. tuberculosis, B. subtilis, C. glutamicum, L. monocytogenes, and S. aureus, deletion of csoR resulted in an about 10-fold increase in the expression of target genes in the absence of copper. This difference might be attributed to several vital regulators that activated the expression of copper-resistance genes in response to copper in P. putida, such as CueR and CopR (Adaikkalam and Swarup, Microbiology. 2002, 148:2857-2867; Hofmann et al., Int J Mol Sci, 2021, 22:2050; Quintana et al., J Biol Chem, 2017, 292:15691-15704). CueR positively regulated the expression of cueA, encoding a copper-transporting P1-type ATPase that played a crucial role in copper resistance. CopR was essential for expressing several genes implicated in cytoplasmic copper homeostasis, such as copA-II, copB-II, and cusA. The existence of these positive regulators makes the function of CosR a secondary or even dispensable insurance in the expression of copper-resistance genes. Consistent with this, there is no CosR homolog in P. aeruginosa, and copper homeostasis is mainly controlled by CueR and CopR.

Reviewer #2 (Public Review):

This manuscript focuses on the apparent involvement of a proposed copper-responsive regulator in the chemotactic response of Pseudomonas putida to Cu(II), a chemorepellent. Broadly, this area is of interest because it could provide insight into how soil microbes mitigate metal stress. Additionally, copper has some historical agricultural use as an antimicrobial, thus can accumulate in soil. The manuscript bases its conclusions on an in vitro screen to identify interacting partners of CheA, an essential kinase in the P. putida chemotaxis-signaling pathway. Much of the subsequent analysis focuses on a regulator of the CsoR/RcnR family (PP_2969).

Weaknesses:

The data presented in this work does not support the model (Figure 8). In particular, PP_2969 is linked to Ni/Co resistance, not Cu resistance. Further, it is not clear how the putative new interactions with CheA would be integrated into diverse responses to various chemoattract/repellents. These two comments are justified below.

Thanks a lot for all these comments. Before designing experiments to explore the function of PP_2969, we found three clues: (i) its sequence showed 38% similarity to the copper-responsive regulator CsoR of M. tuberculosis, and the three conserved amino acids essential for copper-binding were conserved in PP_2969; (ii) it located next to a Ni²⁺/Co²⁺ transporter (PP_2968) on the genome; (iii) a previous report revealed that PP_2969 (also named MreA) expression increased during metal stress, and overexpression of PP_2969 in P. putida and E. coli led to metal accumulation (Zn, Cd, and Cr) (Lunavat et al., Curr Microbiol. 2022, 79:142). These clues indicate that the function of PP_2969 is related to metal-binding, but it remains to be explored which metal(s) PP_2969 binds to. Thus, we played MST assay to test the interaction between PP_2969 and metals, including copper (Cu²⁺), zinc (Zn²⁺), nickel (Ni²⁺), cobalt (Co²⁺), cadmium (Cd²⁺), and magnesium (Mg²⁺). The result showed that PP_2969 was bound to three metal ions (Cu²⁺, Zn²⁺, Ni²⁺), and the binding to Cu²⁺ was the strongest. Besides, the EMSA assay revealed that Cu²⁺/Ni²⁺/Zn²⁺ inhibited the interaction between PP_2969 and promoter DNA, and Cu²⁺ showed the most substantial inhibitory effect at the same concentration. These results suggested that PP_2969 was mainly bound to Cu²⁺, followed by Zn²⁺ and Ni²⁺. To further test whether PP_2969 functioned as a metal-responsive repressor and which metal resistance was related to its target gene, we constructed a PP_2969 deletion mutant and complementary strain and performed a qPCR assay to compare the expression of metal resistance-related genes. 14 metal-resistant-related genes were chosen as targets. The results showed that PP_2969 deletion led to a weak but significant increase (about 1.3-fold) in expression of 10 genes, including three copper-resistance genes (copA-I, copA-II, and copB-II), one nickel-resistance gene (nikB), two cadmium-resistance genes (cadA-I and cadA-III), one cobalt-resistance gene (cbtA), and three multiple metal-resistance genes (czcC-I, czcB-II, and PP_0026) (Fig. 4b, Fig. S5a in the manuscript). Meanwhile, complementation with a multicopy plasmid containing the PP_2969 gene decreased the gene expression in Δ_PP_2969_. Although PP_2969 regulated the expression of multiple metal resistance genes, it showed the most robust binding to Cu²⁺. Thus, we considered its primary function as a Cu²⁺-responsive regulator.

As for the second comment, “How would the putative new interactions with CheA be integrated into diverse responses to various chemoattract/repellents?”, We have some speculations based on our results and previous reports. For example, PP_2969 interacted with CheA and decreased its autophosphorylation activity, and copper inhibited the interaction between CheA and PP_2969. In the absence of copper, PP_2969 binds to promoters to inhibit the expression of copper resistance genes, and it also binds to CheA to inhibit its autophosphorylation, resulting in lower chemotaxis. When the bacteria move to an area of high copper concentration, PP_2969 binds to copper and falls off the DNA promoter, leading to higher expression of copper resistance genes. Meanwhile, copper-binding of PP_2969 decreases its interaction with CheA, increasing CheA autophosphorylation promoting chemotaxis, and bacteria swim away from the high copper concentration. Another attractive target protein is PhaD, a TetR family transcriptional regulator located adjacent to the genes involved in PHA biosynthesis, and it behaves as a carbon source-dependent activator of the pha cluster related to polyhydroxyalkanoates (PHAs) biosynthesis (de Eugenio et al., Environ Microbiol. 2010, 12:1591-1603; Tarazona et al., Environ Microbiol. 2020, 22:3922-3936). Bacterial PHAs are isotactic polymers synthesized under unfavorable growth conditions in the presence of excess carbon sources. PHAs are critical in central metabolism, acting as dynamic carbon reservoirs and reducing equivalents (Gregory et al., Trends Mol Med. 2022, 28:331-342). The interaction between PhaD and CheA leads us to speculate that there might be some connection between PHA synthesis and bacterial chemotaxis. For example, chemotaxis helps bacteria move towards particular carbon sources that favor PHA synthesis; the regulator PhaD activates the genes related to PHA synthesis. Meanwhile, the interaction between PhaD and CheA weakens chemotaxis, causing bacteria to linger in areas rich in these carbon sources. Collectively, we speculate that by interacting with CheA and modulating its autophosphorylation, target proteins such as CsoR/PhaD integrate signals from their original process pathway into chemotaxis signaling.

PP_2969

(1) The authors present a sequence alignment (Figure S5) that is the sole basis for their initial assignment of this ORF as a CsoR protein. There is a conservation of the primary coordinating ligands (highlighted with asterisks) known to be involved in Cu(I) binding to CsoR (ref 31). There are some key differences, though, in residues immediately adjacent to the conserved Cys (the preceding Ala, which is Tyr in the other sequences). The effect of these changes may be significant in a physiological context.

We constructed a point mutation in PP_2969 by replacing the Ala residue before the conserved Cys with a Tyr (CsoR_A39Y) and then analyzed the effect of this mutation on CsoR. As shown in Author response image 1a, CsoR_A39Y showed similar promoter-binding ability as the wild-type CsoR and the presence of Cu²⁺ abolished the interaction between CsoR_A39Y and DNA, suggesting that the A39 residue in PP_2969 was not essential for the DNA-binding and Cu²⁺-binding abilities. Besides, CsoR_A39Y interacted with CheA as the wild-type CsoR did (Author response image 1b), indicating that the Ala39 residue was not required to interact with CheA.

The CsoR from B. subtilis has a Tyr before the conserved Cys, which is the same as other sequences, and the BTH result showed that interaction existed between CsoR and CheA from B. subtilis (Fig. 7 in the manuscript).

Author response image 1.

The effect of CsoR point mutation (CsoR_A39Y) on the DNA-binding and Cu²⁺-binding abilities of CsoR. (a) Analysis for interactions between CsoR/CsoR_A39Y and copA-I promoter DNA using EMSA. The concentrations of CsoR/CsoR_A39Y and Cu²⁺ added in each lane are shown above the gel. Free DNA and protein-DNA complexes are indicated. (b) The interaction between CsoR/CsoR_A39Y and CheA was tested by BTH. Blue indicates protein-protein interaction in the colony after 60 h of incubation, while white indicates no protein-protein interaction. CK+ represents positive control, and CK- represents negative control.

(2) The gene immediately downstream of PP_2969 is homologous to E. coli RcnA, a demonstrated Ni/Co efflux protein, suggesting that P2969 may be Ni or Co responsive. Indeed PP_2970 has previously been reported as Ni/Co responsive (J. Bact 2009 doi:10.1128/JB.00465-09). The host cytosol plays a critical role in determining metal response, in addition to the protein, which can explain the divergence from the metal response expected from the alignment.

Correction: The gene immediately upstream (not downstream) of PP_2969 (the ID is PP_2968, not PP_2970) is homologous to E. coli RcnA, a demonstrated Ni/Co efflux protein. The previous JBact study (J. Bact 2009 doi:10.1128/JB.00465-09) named PP_2968 as MrdH, and mrdH disruption led to sensitivity to cadmium, zinc, nickel, and cobalt, but not copper. Their results also revealed that MrdH was a broad-spectrum metal efflux transporter with a substrate range including Cd²⁺, Zn²⁺, and Ni²⁺. However, the role of MrdH in Cu²⁺ efflux was not tested. Commonly, metal efflux transporter has a broad substrate spectrum, allowing transporters to influence bacterial resistance to a variety of metals (Munkelt et al., J Bacteriol. 2004, 186:8036-8043; Grass et al., J Bacteriol. 2005, 187:1604-1611; Nies et al., J Ind Microbiol. 1995, 14:186-199; Kelley et al., Metallomics. 2021, 13:mfaa002). Our results showed that PP_2969 bound to Cu²⁺, Zn²⁺, and Ni²⁺ under our experimental conditions, and CsoR regulated the expression of genes related to Cu²⁺, Zn²⁺, and Ni²⁺ resistance, indicating that CsoR was involved in resistance to these metals. But the binding of CsoR to Cu²⁺ was the strongest, and Cu²⁺ showed the most substantial inhibitory effect on CsoR-DNA interaction. Thus, we considered its primary function as a Cu²⁺-responsive regulator.

(3) The previous JBact study also explains the lack of an effect (Figure 5b) of deleting PP_2969 on copper-efflux gene expression (copA-I, copA-II, and copB-II) as these are regulated by CueR not PP_2969 consistent with the previous report. Deletion of CsoR/RcnR family regulator will result in constitutive expression of the relevant efflux/detoxification gene, at a level generally equivalent to the de-repression observed in the presence of the signal.

We performed qPCR to test the effect of PP_2969 on gene expression, and we chose 14 target genes, including copper-resistance genes, nickel-resistance genes, zinc-resistance genes, cadmium-resistance genes, and cobalt-resistance genes. The results showed that PP_2969 deletion led to a weak but significant increase (about 1.3-fold) in the expression of 10 genes (Fig. 4b, new Fig. S5a in the manuscript), and complementation with a multicopy plasmid containing PP_2969 gene decreased the gene expression in Δ_PP_2969_. We were confused about these results. Why was the effect of PP_2969 on gene expression so weak? Did we pick the wrong target genes? In other bacteria, deletion of csoR led to an about ten-fold increase in gene expression, generally equivalent to the de-repression observed in the presence of metal. Thus, to further identify target genes, we performed RNA-seq to compare the gene expression in WT and Δ_PP_2969_ without copper. The result surprised us because no gene expression levels changed more than two-fold (data not shown). This result might be attributed to several vital regulators that activated the expression of metal-resistance genes in response to metal in P. putida, such as CueR and CopR (Adaikkalam and Swarup, Microbiology. 2002, 148:2857-2867; Hofmann et al., Int J Mol Sci, 2021, 22:2050; Quintana et al., J Biol Chem, 2017, 292:15691-15704). CueR positively regulated the expression of cueA, encoding a copper-transporting P1-type ATPase that played a crucial role in copper resistance. CopR was essential for expressing several genes implicated in cytoplasmic copper homeostasis, such as copA-II, copB-II, and cusA. The existence of these positive regulators might make the function of CosR a secondary or even dispensable insurance in the expression of copper-resistance genes. Consistent with this, there is no CosR homolog in P. aeruginosa, and copper homeostasis is mainly controlled by CueR and CopR.

(4) Further, CsoR proteins are Cu(I) responsive so measuring Cu(II) binding affinity is not physiologically relevant (Figures 5a and S5b). The affinities of demonstrated CsoR proteins are 10-18 M and these values are determined by competition assay. The MTS assay and resulting affinities are not physiologically relevant.

Thank you for this enlightening comment. This question also confused us during our experiment. The first study on CsoR from Mycobacterium tuberculosis showed that CsoR bound a single-monomer mole equivalent of Cu(I) to form a trigonally coordinated complex, and that was a convincing result from protein structure analysis (Liu et al., Nat Chem Biol. 2007, 3:60-68). They further revealed that the presence of Cu(I) in the EMSA assay abolished the DNA-binding ability of CsoR, but the impact of Cu(II) was not tested. Besides, their results also showed that adding CuCl₂ in the medium induced the expression of the cso operon involved in copper resistance. Perhaps Cu(II) converted to Cu(I) and then bound to CsoR in bacterial cells. Later studies in diverse bacterial species (including Listeria monocytogenes, Corynebacterium glutamicum, Deinococcus radiodurans, and Thermus thermophilus) showed that in vitro assays with Cu(II) abolished the DNA-binding ability of CsoR, indicating that CsoR bound to both Cu (I) and Cu(II) (Corbett et al., Mol Microbiol. 2011, 81:457-472; Teramoto et al., Biosci Biotechnol Biochem. 2012, 76:1952-1958; Zhao et al., Mol Biosyst. 2014, 10:2607-2616; Sakamoto et al., Microbiology. 2010, 156:1993-2005). Here, our results from in vitro assays (MST and EMSA) showed that CsoR bound to Cu(II) and Cu(II) affected the interaction between CsoR and promoter DNA. Compounds containing Cu(I) are poorly soluble in water and easily oxidized by Cu(II). DTT can reduce Cu(II) to Cu(I) (Krzel et al., J Inorg Biochem. 2001, 84:77-88). To test whether Cu(I) bound to CsoR and affected its DNA-binding ability, we recently performed an EMSA assay with the addition of CuCl₂/DTT/CuCl₂+DTT. As shown in Fig. 4d, the addition of DTT (0.1 and 1 mM) decreased CsoR-DNA interaction in the presence of 0.2 mM CuCl₂, while the addition of DTT alone had no apparent influence on CsoR-DNA interaction, indicating that DTT enhanced the inhibition of CuCl₂ on CsoR-DNA interaction, and the Cu(I) converted from Cu(II) by DTT had stronger inhibitory effect than Cu(II) on CsoR-DNA interaction. Together, these results suggested that CsoR bound to Cu(I) more strongly than it bound to Cu(II). We have added these results to the new version of manuscript.

(5) The DNA-binding assays are carried out at protein concentrations well above physiological ranges (Figures 5c and d, and S5c, d). The weak binding will in part result from using DNA sequences upstream of the copA genes and not from PP_2970.

We performed the vitro DNA-binding assay several times, and the lowest CsoR concentration used to obtain a shifted band was about 3 μM, and a higher concentration (15 μM) caused total DNA binding. Thus, we used the concentration of 15 and 20 μM to test the effect of metal on protein-DNA interaction in the assay. We also realized that these concentrations were above physiological ranges. We considered that the in vitro DNA-binding assay was only a mimic of the in vivo process, and the extracellular physiological conditions in EMSA might restrict the activity of CsoR. Besides, we recently performed EMSA to investigate the interaction between CsoR and its own promoter (csoRpro). As shown in Author response image 2, CsoR bound to csoRpro with a similar intensity to that it bound to copA-Ipro. Thus, the weak binding was not caused by the promoter used in the assay. 

Author response image 2.

The binding of CsoR to its own promoter (csoRpro) and copA-I promoter (copA-1pro) in EMSA. The concentrations of CsoR added in each lane are shown above the gel. Free DNA and CsoR-DNA complex are indicated.

CheA interactions

(1) There is no consideration given to the likely physiological relevance of the new interacting partners for CheA.

Thank you for this comment. The initial purpose of this research was to identify new CheA-interacting proteins to broaden our knowledge of CheA and bacterial chemotaxis. Thus, we are currently focusing on the effect of the interaction on CheA and chemotaxis and trying to find the link between different processes and bacterial chemotaxis. We infer that the interaction between these new interacting partners and CheA can integrate signals from different pathways into the chemotaxis signaling pathway so that bacteria can better sense and adapt to different environments. Besides, the other role of the interaction, which is the influence of CheA on these new interacting partners, is also an exciting question that remains to be answered. Among the 16 new CheA-interacting proteins, five showed significant influence on chemotaxis, and the remaining 11 proteins had no obvious impact on chemotaxis (Fig. 2a in the manuscript). CsoR and PhaD inhibited CheA autophosphorylation, and here we focused on the effect of CsoR on chemotaxis. We also investigated the impact of CheA on CsoR, such as gene regulation and copper resistance. However, the results showed that CheA had no obvious influence on these functions of CsoR. The interactions between CheA and these proteins may be physiologically biased, with some interactions affecting the function of CheA and others mainly affecting the function of partners. Future studies on the function of these new CheA-interacting proteins and the role of CheA in regulating their functions would further expand our knowledge of CheA.

(2) How much CheA is present in the cell (copies) and how many copies of other proteins are present? How would specific responses involving individual interacting partners be possible with such a heterogenous pool of putative CheA-complexes in a cell? For PP_2969, the affinity reported (Figure 5A) may lay at the upper end of the CsoR concentration range (for example, CueR in Salmonella is present at ~40 nM).

Thank you for this insightful comment. We don’t know the copy number of CheA and other proteins in the cell. We were also initially surprised and felt skeptical about the reliability of CheA interaction with so many proteins. CheA interacts with CheY, CheW, and CheB in the classical chemotaxis pathway. This study found 16 new CheA-interacting proteins using pull-down assay and subsequent analysis. Moreover, in another unpublished result, we found that CheA interacted with eight c-di-GMP-metabolizing proteins, and CheA transferred the phosphate group to one of them. Together, it seemed that CheA could interact with at least 27 proteins. With such a heterogeneous pool of CheA-complexes, performing a specific response seemed difficult. However, several previous studies have reported the example of one protein interacting with dozens of proteins. For example, the c-di-GMP effector LapD in Pseudomonas fluorescens and Pseudomonas putida can interact with a dozen different c-di-GMP-metabolizing proteins (Giacalone et al., mBio. 2018, 9:e01254-18; Nie et al., Mol Microbiol. 2024, 121:1-17.) In Escherichia coli, a subset of DGCs and PDEs operated as central interaction hubs in a larger “supermodule” by interacting with dozens of proteins (Sarenko et al., mBio. 2017, 8:e01639-17). We infer that the expression of different CheA-interacting proteins might happen at different growth stages or under different conditions, and their interaction with CheA under that stage/condition changed bacterial chemotaxis or the process in which the target protein was involved.

(3) The two-hybrid system experiment uses a long growth time (60 h) before analysis. Even low LacZ activity levels will generate a blue color, depending upon growth medium (see doi: 10.1016/0076-6879(91)04011-c). It is also not clear how Miller units can be accurately or precisely determined from a solid plate assay (the reference cited describes a protocol for liquid culture).

We didn’t observe a blue color on the colony after 60 h growth on a plate under our experimental conditions. The BTH experiment was described as follows: After transforming the two plasmids into E. coli BTH101 cells, the plates containing transformants were placed at 28° for 48 h, at which time the colonies of the transformants were big enough to be picked up and incubated in a liquid medium for 24 h at 28°. Then, 5 μL of the culture was spotted onto an LB agar plate supplemented with antibiotics, X-gal, and IPTG and incubated for 60 h at 28° before taking the photos. After the photos were taken, the bacteria on the plate were scraped off and resuspended with buffer, and then the LacZ activity of the bacteria was tested. According to our experience, culture at 28°(lower than 30°) is a critical condition, and we have not observed false positives in BTH assays under this condition.

Reviewer #1 (Recommendations For The Authors):

In addition to genetic and biochemical approaches, structural studies should be conducted to elucidate the molecular interaction between CheA and CsoR with/without copper.

It would be more logical to first establish the role of CsoR in copper regulation and chemotaxis (the second part of this report) and then investigate its underpinning mechanism (the first part).

Thank you for these recommendations. Structural analysis can reveal more details about the molecular mechanism of CheA-CsoR interaction, but we currently don’t have sufficient experimental conditions for such structural analysis.

As for the presentation logic of the results, we wrote the manuscript following the sequence of experiments. Firstly, screening of CheA interacting proteins (pull-down assay) was conducted, and then the influence of interacting proteins on the chemotaxis of strains and CheA autophosphorylation activity was detected. Based on these results, we obtained two proteins, CsoR and PhaD, and decided to go deeper into the function of CsoR and its effect on chemotaxis. We considered that this writing logic reflected our research design better and could also lay a foundation for future exploration of the functions of other interacting proteins and the physiological significance of interactions.

Reviewer #2 (Recommendations For The Authors):

A huge amount of effort has gone into this work.

It would be good to see at least one of the newly identified interactions turn out to be physiologically relevant.

The experimental tools appear to be available to do this, but it is critical to consider how these tools can lead to attempts to prove rather than test and possibly refute a model or hypothesis. In particular, please consider some of the comments about the physiological relevance of affinities when generating models.

Thank you for these recommendations. Our study aimed to screen new interacting proteins of CheA and explore how new interacting proteins affect CheA activity and bacterial chemotaxis, thereby broadening our understanding of chemotaxis. However, the impact of each protein-protein interaction has two sides: the influence of A to B and B to A. During experimental design, we focused more on the influence of identified interacting proteins on CheA function and chemotaxis but paid less attention to the function of interacting proteins and the influence of the interaction on their function. Moreover, our study found that the influence of protein-protein interaction was biased. In the interaction between CsoR and CheA, CsoR mainly affected the function of CheA and then affected the chemotaxis, while CheA had no significant effect on the function of CsoR. This might be attributed to the weak effect of CsoR in regulating metal resistance in P. putida, and we speculated that this interaction was more about favoring the sensing and avoiding metal stress. In addition, we planned to explore the interaction between CheA and another interacting protein (PhaD) in the future, reveal the effect of the interaction on PhaD function (regulation of PHAS synthesis in bacteria), and explore the effect of the interaction on CheA function and chemotaxis, to find out whether the association existed between PHAS anabolism and bacterial chemotaxis. Besides, for those proteins that did not have significant effects on CheA autophosphorylation and bacterial chemotaxis, we speculated that CheA might affect their function/activity through interactions, which meant that the physiological effects of the interaction mainly reflected through the interacting protein rather than CheA. These are speculations that need to be tested by experiments.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

Rossi et al. asked whether gait adaptation is solely a matter of slow perceptual realignment or if it also involves fast/flexible stimulus-response mapping mechanisms. To test this, they conducted a series of split-belt treadmill experiments with ramped perturbations, revealing behavior indicative of a flexible, automatic stimulus-response mapping mechanism.

Strengths:

(1) The study includes a perceptual test of leg speed, which correlates with the perceptual realignment component of motor aftereffects. This indicates that there are motor performances that are not accounted for by perceptual re-alignment.

(2) They study incorporates qualitatively distinct, hypothesis-driven models of adaptation and proposes a new framework that integrates these various mechanisms.

Weaknesses:

(1) The study could benefit from considering other alternative models. As the authors noted in their discussion, while the descriptive models explain some patterns of behaviour/aftereffects, they don't currently account for how these mechanisms influence the initial learning process itself.

(1a) For example, the pattern of gait asymmetric might differ for perceptual realignment (a smooth, gradual process), structural learning (more erratic, involving hypothesis testing/reasoning to understand the perturbation, see (Tsay et al. 2024) for a recent review on Reasoning), and stimulus-response mapping (possibly through a reinforcement based trial-and-error approach). If not formally doing a model comparison, the manuscript might benefit from clearly laying out the behavioural predictions for how these different processes shape initial learning.

(1b) Related to the above, the authors noted that the absence of difference during initial learning suggests that the differences in Experiment 2 in the ramp-up phase are driven by two distinct processes: structural learning and memory-based processes. If the assumptions about initial learning are not clear, this logic of this conclusion is hard to follow.

Thank you for this insightful comment. We agree that considering alternative models and clarifying their potential contributions to the initial learning process would enhance the manuscript. We performed additional analyses and revised the text to outline how the mechanisms of adaptation in our study align with the framework described by Tsay et al. (2024) regarding the initial learning process and other features of adaptation.

First, we referenced the Tsay et al. framework in the Introduction and Discussion to highlight parallels between their description of implicit adaptation and our forward model recalibration mechanism (producing motor changes and perceptual realignment). Specifically, the features defining recalibration in our study – gradual, trial-by-trial adjustments, rigid learning that leads to aftereffects, and limited contribution to generalization – align with those described by Tsay et al.

Second, we used the description provided by Tsay et al. to test the presence of explicit strategies in our study. We specifically test for the criteria of reportability and intentionality, corroborating the finding that our stimulus response mapping mechanism differs from explicit strategies.

“A recent framework for motor learning by Tsay et al. defines explicit strategies as motor plans that are both intentional and reportable (Tsay et al., 2024). Within this framework, Tsay et al. clarify that "intentional" means participants deliberately perform the motor plan, while "reportable" means they are able to clearly articulate it.” (Experiment 2 Results, lines 515-518).

“…the motor adjustments reported by participants consistently fail to meet the criteria for explicit strategies as outlined by Tsay et al.: reportability and intentionality (Tsay et al., 2024).” (Discussion, lines 657-660).

Third, we interpreted the operation of stimulus-response mapping within the Tsay theoretical framework for the three stages of motor learning: 1) “reasoning” to acquire new action–outcome relationships, 2) “refinement” of the motor action parameters, and 3) “retrieval” of learnt motor actions based on contextual cues. We note that the definition of these stages closely aligns with our definition for stimulus response mapping mechanisms. Moreover, according to Tsay’s definition, both implicit and explicit learning mechanisms can involve similar reasoning and retrieval processes. This shared operational basis may explain why our stimulus-response mapping mechanism exhibits some characteristics associated with explicit strategies, such as flexibility and generalizability.

We performed a new analysis to evaluate Tsay’s framework predictions that, if walking adaptation includes a stimulus-response mapping mechanism following these three stages of motor learning, the learning process would initially be erratic and would then stabilize as learning progresses. We assessed within-participant residual variance in step length asymmetry around a double exponential model fit during adaptation, testing the prediction that this variability would decrease between the start and end of adaptation. Experiment 1 results confirmed this prediction, showing that a significant reduction in variability as adaptation progressed.

“We finally tested whether the pattern of motor variability during adaptation aligns with predictions for learning new  stimulus response maps. In contrast to recalibration, mapping mechanisms are predicted to be highly  variable  and  erratic  during  early learning, and stabilize as learning progresses (Tsay et al., 2024). Consistent with these predictions,  the  step  length  asymmetry residual  variance  (around  a  double exponential  fit)  decreased  significantly between the start and end of adaptation (residual variance at start minus end of adaptation = 0.005 [0.004, 0.007], mean [CI]; SI Appendix, Fig. S3). These control analyses corroborate the hypothesis that the “no aftereffects” region of the Ramp Down reflects the operation of a mapping mechanism.”

(Experiment 1 Results, lines 187-194; Methods, lines 1040-1050).

Moreover, Experiment 2 results demonstrated that the pattern of variability (its magnitude and decay in adaptation) did not differ between participants using memory-based versus structure-based stimulus-response mapping mechanisms. These findings suggest that both types of mapping operate accordingly to Tsay’s stages of motor learning.

“Furthermore, the pattern of step length asymmetry variability was similar between the subgroups (structure – memory difference in residual variance relative to double exponential during initial adaptation = -0.0052 [0.0161, 0.0044], adaptation plateau = -0.0007 [-0.0021, 0.0003], difference in variance decay = -0.0045 [-0.0155, 0.0052], mean [CI]; SI Appendix, Fig. S16). This confirms that the distinct performance clusters in the Ramp Up & Down task are not driven by natural variations in learning ability, such as differences in learning speed or variability. Rather, these findings indicate that the subgroups employ different types of mapping mechanisms, which perform similarly during initial learning but differ fundamentally in how they encode, retrieve, and generalize relationships between perturbations and Δ motor outputs.” (Experiment 2 Results, lines 503-511).

“Both memory- and structure-based operations of mapping align with Tsay et al.’s framework for motor learning: first, action–outcome relationships are learned through exploration; second, motor control policies are refined to optimize rewards or costs, such as reducing error; and finally, learned mappings or policies are retrieved based on contextual cues (Tsay et al., 2024). Consistent with the proposed stages of exploration followed by refinement, we found that motor behavior during adaptation was initially erratic but became less variable at later stages of learning. Similarly, consistent with the retrieval stage, the generalization observed in the ramp tasks indicates that learned motor outputs are flexibly retrieved based on belt speed cues.” (Discussion, lines 701-708).

Finally, we addressed the prediction outlined by Tsay et al. that repeated exposure to perturbations attenuates the magnitude of forward model recalibration, with savings being driven by stimulus-response mapping mechanisms. While we could not directly test savings for the primary perturbation used during adaptation, we were able to indirectly evaluate savings for a different perturbation through analyses of our control experiments combined with previous results from Leech et al. (Leech et al., 2018). Specifically, we examined how motor aftereffects and perceptual realignment evolved across repeated iterations of the speed-matching task post-adaptation in Ascending groups. Each task began with the right leg stationary and the left leg moving at 0.5 m/s – a configuration corresponding to a perturbation of -0.5 m/s, which is opposite in direction to the adaptation perturbation. By analyzing repeated exposures to this -0.5 m/s perturbation across iterations, we gained insights into the learning dynamics associated with this perturbation and the effect of repeated exposures on motor aftereffects and perceptual realignment. Consistent with predictions from Tsay et al., our results combined with Leech et al. demonstrate that, with repeated exposures to the same perturbation, perceptual realignment decays while the contribution of stimulus-response mapping to aftereffect savings is enhanced. We present this analysis and interpretation in Control Experiments Results, lines 429-442; Figure 8B; Table S7; and Discussion lines 709-753.

(1c) The authors could also test a variant of the dual-rate state-space model with two perceptual realignment processes where the constraints on retention and learning rate are relaxed. This model would be a stronger test for two perceptual re-alignment processes: one that is flexible and another that is rigid, without mandating that one be fast learning and fast forgetting, and the other be slow learning and slow forgetting.

We tested multiple variants of the suggested models, and confirmed that they cannot capture the motor behavior observed in our Ramp Down task. We include Author response image 1 with the models fits, Author response table 1 with the BIC statistics, and the models equations below. Only the recalibration + mapping model captures the matching-then-divergent behavior of the Δ motor output, corroborating our interpretation that state-space based models cannot capture the mapping mechanism (see Discussion, “Implications for models of adaptation”). Furthermore, all models fit the data significantly worse than the recalibration+mapping model according to the BIC statistic.

Model fits:

Author response image 1.

Statistical results:

Author response table 1.

Model definitions:

• DualStateRelaxed: same equations as the original Dual State, but no constraints dictating the relative relationship between the parameters

• DualStateRelaxedV2: same equations as the original Dual State, but no constraints dictating the relative relationship between the parameters, and “loose” parameter bounds (parameters can take values between -10 to 10).

• PremoOriginalRelaxed: PReMo with two states (see below), no constraints dictating the relative relationship between the parameters

• PremoOriginalRelaxed: PReMo with two states (see below), no constraints dictating the relative relationship between the parameters, and “loose” parameter bounds (parameters can take values between -10 to 10).

PReMo with two states – the remaining equations are the same as the original PReMo (see Methods):

(2) The authors claim that stimulus-response mapping operates outside of explicit/deliberate control. While this could be true, the survey questions may have limitations that could be more clearly acknowledged.

(2a) Specifically, asking participants at the end of the experiments to recall their strategies may suffer from memory biases (e.g., participants may be biased by recent events, and forget about the explicit strategies early in the experiment), be susceptible to the framing of the questions (e.g., participants not being sure what the experimenter is asking and how to verbalize their own strategy), and moreover, not clear what is the category of explicit strategies one might enact here which dictates what might be considered "relevant" and "accurate".

(2b) The concept of perceptual realignment also suggests that participants are somewhat aware of the treadmill's changing conditions; therefore, as a thought experiment, if the authors have asked participants throughout/during the experiment whether they are trying different strategies, would they predict that some behaviour is under deliberate control?

We have expanded the discussion to explicitly acknowledge that our testing methodology for assessing explicit strategies may have limitations, recognizing the factors mentioned by the reviewer. Moreover, as mentioned in response to comment (1), we leveraged the framework from Tsay et al., 2024 and its definition of explicit strategies to ensure a robust and consistent approach in interpreting the survey responses.

We revised the Experiment 2 Results section, lines 515-518, to specify that we are evaluating the presence of explicit strategies according to the criteria of intentionality and reportability:

“A recent framework for motor learning by Tsay et al. defines explicit strategies as motor plans that are both intentional and reportable (Tsay et al., 2024). Within this framework, Tsay et al. clarify that "intentional" means participants deliberately perform the motor plan, while "reportable" means they are able to clearly articulate it.”

We then reorganized the Discussion to include a separate section “Mapping operates independently of explicit control”, lines 646-661, where we discuss limitations of the survey methodology and interpretation of the results according to Tsay et al., 2024:

“Here, we show that explicit strategies are not systematically used to adapt step length asymmetry and Δ motor output: the participants in our study either did not know what they did, reported changes that did not actually occur or would not lead symmetry. Only one person reported “leaning” on the left (slow) leg for as much time as possible, which is a relevant but incomplete description for how to walk with symmetry. Four reports mentioned pressure or weight, which may indirectly influence symmetry (Hirata et al., 2019; Lauzière et al., 2014), but they were vague and conflicting (e.g., “making heavy steps on the right foot” or “put more weight on my left foot”). All other responses were null, explicitly wrong or irrelevant, or overly generic, like wanting to “stay upright” and “not fall down”. We acknowledge that our testing methodology has limitations. First, it may introduce biases related to memory recall or framing of the questionnaire. Second, while it focuses on participants' intentional use of explicit strategies to control walking, it does not rule out the possibility of passive awareness of motor adjustments or treadmill configurations. Despite these limitations, the motor adjustments reported by participants consistently fail to meet the criteria for explicit strategies as outlined by Tsay et al.: reportability and intentionality (Tsay et al., 2024). Together with existing literature, this supports the interpretation that stimulus response mapping operates automatically.”

We also made the following addition to the “Limitations” section of the Discussion (lines 917-919):

“While mapping differs from explicit strategies as they are currently defined, we still lack a comprehensive framework to capture the varying levels and nuanced characteristics of intentionality and awareness of different mechanisms (Tsay et al., 2024).”

We finally note that “Unlike explicit strategies, which are rapidly acquired and diminish over time, this mapping mechanism exhibits prolonged learning beyond 15 minutes, with a rate comparable to recalibration” (Discussion, lines 632-634).

(3) The distinction between structural and memory-based differences in the two subgroups was based on the notion that memory-based strategies increase asymmetry. However, an alternative explanation could be that unfamiliar perturbations, due to the ramping up, trigger a surprise signal that leads to greater asymmetry due to reactive corrections to prevent one's fall - not because participants are generalizing from previously learned representations (e.g., (Iturralde & Torres-Oviedo, 2019)).

We agree that reactive corrections could contribute to the walking pattern in response to split-belt perturbations, as detailed by Iturralde & Torres-Oviedo, 2019. We also acknowledge that reactive corrections are rapid, flexible, feedback-driven, and automatic – characteristics that make them appear similar to stimulus-response mapping. However, a detailed evaluation of our results suggests that the behaviors observed in the ramp tasks cannot be fully explained by reactive corrections. Reactive corrections occur almost immediately, quickly adjusting the walking pattern to reduce error and improve stability. This excludes the possibility that what we identified as stimulusresponse mapping could instead be reactive corrections, because the stimulus-response mapping observed in our study is acquired slowly at a rate comparable to recalibration. It also excludes the possibility that the increased asymmetry in the Ramp Up & Down could be due to reactive corrections, because these would operate alongside mapping to help reduce asymmetry rather than exacerbate it.

We made substantial revisions to the Discussion and included the section “Stimulus-response mapping is flexible but requires learning” to explain this interpretation (lines 595-622):

“The mapping mechanism observed in our study aligns with the corrective responses described by Iturralde and Torres-Oviedo, which operate relative to a recalibrated "new normal" rather than relying solely on environmental cues (Iturralde and Torres-Oviedo, 2019). Accordingly, our findings suggest a tandem architecture: forward model recalibration adjusts the nervous system's "normal state," while stimulus-response mapping computes motor adjustments relative to this "new normal." This architecture explains the sharp transition from flexible to rigid motor adjustments observed in our Ramp Down task. The transition occurs at the configuration perceived as "equal speeds" (~0.5 m/s speed difference) because this corresponds to the recalibrated “new normal”.

In the first half of the Ramp Down, participants adequately modulated their walking pattern to accommodate the gradually diminishing perturbation, achieving symmetric step lengths. Due to the recalibrated “new normal”, perturbations within this range are perceived as congruent with the direction of adaptation but reduced in magnitude. This allows the mapping mechanism to flexibly modulate the walking pattern by using motor adjustments previously learned during adaptation. Importantly, the rapid duration of the Ramp Down task rules out the possibility that the observed modulation may instead reflect washout, as confirmed by the fact the aftereffects measured post-Ramp-Down were comparable to previous work (Kambic et al., 2023; Reisman et al., 2005).

In the second half of the Ramp Down, aftereffects emerged as participants failed to accommodate perturbations smaller than the recalibrated “new normal”. These perturbations were perceived as opposite to the adaptation perturbation and, therefore, novel. Accordingly, the mapping mechanism responded as it would to a newly introduced perturbation, rather than leveraging previously learned adjustments (Iturralde and Torres-Oviedo, 2019). Due to the rapid nature of the Ramp Down, the mapping mechanism lacked sufficient time to learn the novel motor adjustments required for these perturbations – a process that typically takes several minutes, as shown by our baseline ramp tasks and control experiments. As mapping-related learning was negligible, the rigid recalibration adjustments dominated during this phase. Consequently, the walking pattern did not change to accommodate the gradually diminishing perturbation, leading to the emergence of aftereffects.”

(4) Further contextualization: Recognizing the differences in dependent variables (reaching position vs. leg speed/symmetry in walking), could the Proprioceptive/Perceptual Re-alignment model also apply to gait adaptation (Tsay et al., 2022; Zhang et al., 2024)? Recent reaching studies show a similar link between perception and action during motor adaptation (Tsay et al., 2021) and have proposed a model aligning with the authors' correlations between perception and action. The core signal driving implicit adaptation is the discrepancy between perceived and desired limb position, integrating forward model predictions with proprioceptive/visual feedback.

We appreciate the reviewer’s suggestion and agree that the Proprioceptive Re-alignment model (PReMo) and Perceptual Error Adaptation model (PEA), offer valuable insights into the relationship between perception and motor adaptation. To explore whether these frameworks apply to gait adaptation, we conducted an extensive modeling analysis. This is shown in Figure 5 and Supplementary Figures S7-S8, and is detailed in the text of Experiment 1 Results section “Modelling analysis for perceptual realignment” (lines 327–375), Methods section “Proprioceptive re-alignment model (PReMo)” (lines 1181-1221), Methods section “Perceptual Error Adaptation model (PEA)” (lines 1222-1247), Methods section “Perceptuomotor recalibration + mapping (PM-ReMap)” (lines 1248-1286), and SI Appendix section “Evaluation and development of perceptual models.” (lines 99-237).

First, we evaluated how PReMo and PEA models fitted our Ramp Down data. We translated the original variables to walking adaptation variables using a conceptual equivalence explained by one of the features explored by Tsay et al. (2022). Specifically, the manuscript provides guidance on extending the PReMo model from visuomotor adaptation in response to visual-proprioceptive discrepancies, to force-field adaptation in response to mechanical perturbations – which share conceptual similarities with split-belt treadmill perturbations. The manuscript also discusses that, if vision is removed, the proprioceptive shift decays back to zero according to a decay parameter. This description entails that proprioceptive shift cannot increase or develop in the absence of vision. We applied the models to split-belt adaptation in accordance with this information, as described in the SI Appendix: “PReMo variables equivalents for walking adaptation”. As reported in Experiment 1 Results “Modelling analysis for perceptual realignment” (lines 327–375) and Figure 5, neither PReMo nor PEA adequately captured the key features of our Ramp Down data: “The models could not capture the matching-then-divergent behavior of Δ motor output, performing significantly worse than the recalibration + mapping model (PReMo minus recalibration+mapping BIC difference = 24.591 [16.483, 32.037], PEA minus recalibration+mapping BIC difference = 6.834 [1.779, 12.130], mean [CI]). Furthermore, they could not capture the perceptual realignment and instead predicted that the right leg would feel faster than the left throughout the entire Ramp Down”.

Second, we used simulations to confirm that PReMo and PEA cannot account for the perceptual realignment observed in our study, and to understand why. At adaptation plateau, PReMo predicts that perceived and actual step length asymmetry converge, as shown in Fig. S7A, top, and as detailed in the SI Appendix “Original PReMo simulations”. We found that this is because PReMo assumes that perceptual realignment arises specifically from mismatches between different sensory modalities. This assumption works for paradigms that introduce an actual mismatch between sensory modalities, such as visuomotor adaptation paradigms with a mismatch between vision and proprioception. This assumption also works for paradigms that indirectly introduce a mismatch between integrated sensory information from different sensory modalities. In force-field adaptation, both proprioceptive and visual inputs are present and realistic, but when these inputs are integrated with sensory predictions, the resulting integrated visual estimate is mismatched compared to the integrated proprioceptive estimate. In contrast, the assumption that perceptual realignment arises from sensory modalities mismatches does not work for paradigms that involve a single sensory modality. Split-belt adaptation only involves proprioception as no visual feedback is given, and perceptual realignment arises from discrepancies between predicted and actual motor outcomes, rather than between integrated sensory modalities.

To overcome this limitation, we reinterpreted the variables of the PReMo model, while keeping the original equations, to account for realignment driven by mismatches of the same nature as the perturbation driving adaptation. As reported in the SI Appendix “Iterative simulations for the development of PM-ReMap”, the simulation (Fig. S7A, middle row) “showed perceptual realignment at adaptation plateau, addressing a limitation of the original model. However, it failed to account for the Ramp Down perceptual results, inaccurately predicting that belt speeds feel equal when they are actually equal (Fig. S7A, middle row, perceived perturbation decays alongside actual perturbation and converge to zero at the end of the Ramp Down). […] This occurs because, under the retained PReMo equations, β_p and β_v change immediately and are proportional to the difference between and on each trial, so that they ramp down to zero in parallel with the perturbation”.

We also noted that the simulations of the original and reinterpreted PReMo models could also not support the operation of the mapping mechanism observed in the Ramp Down (Fig. S7B). We describe that “This occurs because the overall motor output x_p, which includes both recalibration and mapping mechanisms, changes gradually according to the learning rate 𝐾. Consequently, changes in 𝐺 take many trials to be fully reflected in x_p. Hence, we found complementary limitations where PReMo assumes perceptual realignment changes immediately while mapping adjustments develop gradually – but the opposite is true in our data”.

We therefore modified the PReMo equations and developed a new model, called perceptuomotor recalibration + mapping (PM-ReMap) that addresses these limitations and is able to capture our Ramp Down motor and perceptual results. As described in the SI Appendix “Iterative simulations for the development of PM-ReMap”, “we introduced an update equation for β_p so that it changes gradually trial-by-trial according to the learning rate 𝐾. We then removed the learning rate from the update equation for x_p so that it integrates two distinct types of changes: 1) the gradual changes in driven by β_p and representing the recalibration mechanism, and 2) the immediate changes in 𝐺 – representing the mapping mechanism”. The final equations of the PM-ReMap model are as follows:

As reported in Experiment 1 Results, “Modelling analysis for perceptual realignment”, and as shown in Fig. 5C, “the PM-ReMap model captured the Δ motor output in the Ramp Down with performance comparable to that of the recalibration + mapping model (BIC difference = 2.381 [-0.739, 5.147], mean [CI]). It also captured perceptual realignment, predicting that some intermediate belt speed difference in the Ramp Down is perceived as “equal speeds” (, Fig. 5C)”. We also found that the estimated aligned with the empirical measurement of the PSE in the Ramp Down both at group and individual level: “At group level, was comparable to the upper bound of compensation_perceptual (difference = -7 [-15, 1]%, mean [CI]), but significantly larger than the lower bound (difference = 19 [8, 31]%, mean [CI]). Furthermore, we found a significant correlation between individual participants’ and their upper bound of compensation_perceptual (r=0.63, p=0.003), but not their lower bound (r=0.30, p=0.203). Both sets of results are consistent with those observed for the recalibration + mapping model”.

Based on these findings, we summarize that PM-ReMap “extends the recalibration + mapping model by incorporating the ability to account for forgetting – typical of state space models – while still effectively capturing both recalibration and mapping mechanisms. However, performance of the PM-ReMap model does not exceed that of the simpler recalibration + mapping model, suggesting that forgetting and unlearning do not have a substantial impact on the Ramp Down”.

Reviewer #2 (Public review):

Recent findings in the field of motor learning have pointed to the combined action of multiple mechanisms that potentially contribute to changes in motor output during adaptation. A nearly ubiquitous motor learning process occurs via the trial-by-trial compensation of motor errors, often attributed to cerebellar-dependent updating. This error-based learning process is slow and largely unconscious. Additional learning processes that are rapid (e.g., explicit strategy-based compensation) have been described in discrete movements like goal-directed reaching adaptation. However, the role of rapid motor updating during continuous movements such as walking has been either under-explored or inconsistent with those found during the adaptation of discrete movements. Indeed, previous results have largely discounted the role of explicit strategy-based mechanisms for locomotor learning. In the current manuscript, Rossi et al. provide convincing evidence for a previously unknown rapid updating mechanism for locomotor adaptation. Unlike the now well-studied explicit strategies employed during reaching movements, the authors demonstrate that this stimulus-response mapping process is largely unconscious. The authors show that in approximately half of subjects, the mapping process appears to be memory-based while the remainder of subjects appear to perform structural learning of the task design. The participants that learned using a structural approach had the capability to rapidly generalize to previously unexplored regions of the perturbation space.

One result that will likely be particularly important to the field of motor learning is the authors' quite convincing correlation between the magnitude of proprioceptive recalibration and the magnitude error-based updating. This result beautifully parallels results in other motor learning tasks and appears to provide a robust marker for the magnitude of the mapping process (by means of subtracting off the contribution of error-based motor learning). This is a fascinating result with implications for the motor learning field well beyond the current study.

A major strength of this manuscript is the large sample size across experiments and the extent of replication performed by the authors in multiple control experiments.

Finally, I commend the authors on extending their original observations via Experiment 2. While it seems that participants use a range of mapping mechanisms (or indeed a combination of multiple mapping mechanisms), future experiments may be able to tease apart why some subjects use memory versus structural mapping. A future ability to push subjects to learn structurally-based mapping rules has the potential to inform rehabilitation strategies.

Overall, the manuscript is well written, the results are clear, and the data and analyses are convincing. The manuscript's weaknesses are minor, mostly related to the presentation of the results and modeling.

Weaknesses:

The overall weaknesses in the manuscript are minor and can likely be addressed with textual changes.

(1) A key aspect of the experimental design is the speed of the "ramp down" following the adaptation period. If the ramp-down is too slow, then no after-effects would be expected even in the alternative recalibration-only/errorbased only hypothesis. How did the authors determine the appropriate rate of ramp-down? Do alternative choices of ramp-down rates result in step length asymmetry measures that are consistent with the mapping hypothesis?

We thank the reviewer for their insightful comment regarding the rate of the Ramp Down following the adaptation period and its potential impact on aftereffects under different hypotheses. We added a detailed explanation for how we determined the Ramp Down design, including analyses of previous work, to the SI Appendix, “Ramp Down design”, lines 22-98. We also describe the primary points in the main Methods section, “Ramp Tasks”, lines 978-991:

As described in SI Appendix, “Ramp Down design”, the Ramp Down task was specifically designed to measure the pattern of aftereffects in a way that ensured reliable and robust measurements with sufficient resolution across speeds, and that minimized washout to prevent confounding the results. To balance time constraints with a measurement resolution adequate for capturing perceptual realignment, we used 0.05 m/s speed decrements, matching the perceptual sensitivity estimated from our re-analysis of the baseline data from Leech et al. (Leech et al., 2018a). To obtain robust motor aftereffect measurements, we collected three strides at each speed condition, as averaging over three strides represents the minimum standard for consistent and reliable aftereffect estimates in split-belt adaptation (typically used in catch trials) (Leech et al., 2018a; Rossi et al., 2019; Vazquez et al., 2015). To minimize unwanted washout by forgetting and/or unlearning, we did not pause the treadmill between adaptation and the post-adaptation ramp tasks, and ensured the Ramp Down was relatively quick, lasting approximately 80 seconds on average. Of note, the Ramp Down design ensures that even in cases of partial forgetting, the emergence pattern of aftereffects remains consistent with the underlying hypotheses.

In the SI Appendix, we explain that, while we did not test longer ramp-down durations directly, previous data suggest that durations of up to at least 4.5 minutes would yield step length asymmetry measures consistent with our results and the mapping hypothesis. Additionally, our control experiments replicated the behavior observed in the Ramp Down using speed match tasks lasting only 30 seconds, further supporting the robustness of our findings across varying durations.

(2) Overall, the modeling as presented in Figure 3 (Equation 1-3) is a bit convoluted. To my mind, it would be far more useful if the authors reworked Equations 1-3 and Figure 3 (with potential changes to Figure 2) so that the motor output (u) is related to the stride rather than the magnitude of the perturbation. There should be an equation relating the forward model recalibration (i.e., Equation 1) to the fraction of the motor error on a given stride, something akin to u(k+1) = r * (u(k) - p(k)). This formulation is easier to understand and commonplace in other motor learning tasks (and likely what the authors actually fit given the Smith & Shadmehr citation and the derivations in the Supplemental Materials). Such a change would require that Figure 3's independent axes be changed to "stride," but this has the benefit of complementing the presentation that is already in Figure 5.

We reworked these equations (now numbered 4-6, lines 207-209) so that the motor output u is related to stride k as suggested by the reviewer:

We changed Figure 2 and Figure 3 accordingly, adding a “stride” x-axis to the Ramp Down data figure.

Reviewer #2 (Recommendations for the authors):

I think that some changes to the text/ordering could improve the manuscript's readability. In particular:

(1) My feeling is that much of the equations presented in the Methods section should be moved to the Results section. Particularly Equations 9-11. The introduction of these motor measures should likely precede Figure 1, as their definitions form the crux of Figure 1 and the subsequent analyses.

(2) It is unclear to me why many of the analyses and discussion points have been relegated to Supplemental Material. I would significantly revise the manuscript to move much of the content from Supplemental Material to the Methods and Discussion (where appropriate). Even the Todorov and Herzfeld models can likely simply be referenced in the text without a need for their full description in the Supplemental material - as their implementations appear to this reviewer as consistent with those presented in the respective papers. Beyond the Supplementary Tables, my feeling is that nearly all of the content in Supplemental can either be simply cited (e.g. alternative model implementations) or directly incorporated into the main manuscript without compromising the readability of the manuscript.

We reorganized the manuscript and SI Appendix substantially, moving content to the Results or other main text section. The changes included those recommended by the reviewer:

• We moved the equations describing step length asymmetry, perturbation, and Δ motor output (originally numbered Eq. 9-11) to the Results section (Experiment 1, “Motor paradigm and hypothesis”, lines 131-133, now numbered Eq. 1-3).

• We moved Supplementary Methods to the main Methods section

• We moved the most relevant content of the Supplementary Discussion to the main Discussion, and removed the less relevant content altogether.

• We moved the methods describing walking-adaptation specific implementation of the Todorov and Herzfeld models to the main Methods section and removed the portions that were identical to the original implementation.

• We moved the control experiments to the main text (main Results and Methods sections).

• We removed the SI Appendix section “Experiment 1 mechanisms characteristics”

Reviewer #3 (Public review):

Summary:

In this work, Rossi et al. use a novel split-belt treadmill learning task to reveal distinct sub-components of gait adaptation. The task involved following a standard adaptation phase with a "ramp-down" phase that helped them dissociate implicit recalibration and more deliberate SR map learning. Combined with modeling and re-analysis of previous studies, the authors show multiple lines of evidence that both processes run simultaneously, with implicit learning saturating based on intrinsic learning constraints and SR learning showing sensitivity to a "perceptual" error. These results offer a parallel with work in reaching adaptation showing both explicit and implicit processes contributing to behavior; however, in the case of gait adaptation the deliberate learning component does not appear to be strategic but is instead a more implicit SR learning processes.

Strengths:

(1) The task design is very clever and the "ramp down" phase offers a novel way to attempt to dissociate competing models of multiple processes in gait adaptation.

(2) The analyses are thorough, as is the re-analysis of multiple previous data sets.

(3) The querying of perception of the different relative belt speeds is a very nice addition, allowing the authors to connect different learning components with error perception.

(4) The conceptual framework is compelling, highlighting parallels with work in reaching but also emphasizing differences, especially w/r/t SR learning versus strategic behaviors. Thus the discovery of an SR learning process in gait adaptation would be both novel and also help conjoin different siloed subfields of motor learning research.

Weaknesses:

(1) The behavior in the ramp-down phase does indeed appear to support multiple learning processes. However, I may have missed something, but I have a fundamental worry about the specific modeling and framing of the "SR" learning process. If I correctly understand, the SR process learns by adjusting to perceived L/R belt speed differences (Figure 7). What is bugging me is why that process would not cause the SR system to still learn something in the later parts of the ramp-down phase when the perceived speed differences flip (Figure 4). I do believe this "blunted learning" is what the SR component is actually modeled with, given this quote in the caption to Figure 7: "When the perturbation is perceived to be opposite than adaptation, even if it is not, mapping is zero and the Δ motor output is constant, reflecting recalibration adjustments only." It seems a priori odd and perhaps a little arbitrary to me that a SR learning system would just stop working (go to zero) just because the perception flipped sign. Or for that matter "generalize" to a ramp-up (i.e., just learn a new SR mapping just like the system did at the beginning of the first perturbation). What am I missing that justifies this key assumption? Or is the model doing something else? (if so that should be more clearly described).

We concur that this point was confusing, and we performed additional analyses and revised the text to improve clarity. Specifically, we clarify that the stimulus-response mapping does indeed still learn in the second portion of the Ramp Down, when the perceived speed differences flip. However, learning by the mapping mechanism proceeds slowly – at a rate comparable to that of forward model recalibration, taking several minutes. The duration of the task is relatively short, so that learning by the mapping mechanism is limited. We schematize the learning to be zero as an approximation. We have now included an additional modelling analysis (as part of our expanded perceptual modelling analyses), which shows there is no significant improvement in modelling performance when accounting for forgetting of recalibration or learning in the opposite direction by mapping in the second half of the ramp down, supporting this approximation. We explain this and other revisions in detail below.

We include a Discussion section “Stimulus-response mapping is flexible but requires learning” where we improve our explanation of the operation of the mapping mechanism in the Ramp Down by leveraging the framework proposed by Iturralde and Torres-Oviedo, 2019. The section first explains that mapping operates relative to a new equilibrium corresponding to the current forward model calibration (lines 595-603):

“The mapping mechanism observed in our study aligns with the corrective responses described by Iturralde and Torres-Oviedo, which operate relative to a recalibrated "new normal" rather than relying solely on environmental cues (Iturralde and Torres-Oviedo, 2019). Accordingly, our findings suggest a tandem architecture: forward model recalibration adjusts the nervous system's "normal state," while stimulus-response mapping computes motor adjustments relative to this "new normal." This architecture explains the sharp transition from flexible to rigid motor adjustments observed in our Ramp Down task. The transition occurs at the configuration perceived as "equal speeds" (~0.5 m/s speed difference) because this corresponds to the recalibrated “new normal”.”

The following paragraph (lines 604-611) explain how this concept reflects in the first half of the Ramp Down:

“In the first half of the Ramp Down, participants adequately modulated their walking pattern to accommodate the gradually diminishing perturbation, achieving symmetric step lengths. Due to the recalibrated “new normal”, perturbations within this range are perceived as congruent with the direction of adaptation but reduced in magnitude. This allows the mapping mechanism to flexibly modulate the walking pattern by using motor adjustments previously learned during adaptation. Importantly, the rapid duration of the Ramp Down task rules out the possibility that the observed modulation may instead reflect washout, as confirmed by the fact the aftereffects measured post-Ramp-Down were comparable to previous work (Kambic et al., 2023; Reisman et al., 2005).”

The last paragraph (lines 612–622) explain the second half of the Ramp Down in light of the equilibrium concept and of the slow learning rate of mapping:

“In the second half of the Ramp Down, aftereffects emerged as participants failed to accommodate perturbations smaller than the recalibrated “new normal”. These perturbations were perceived as opposite to the adaptation perturbation and, therefore, novel. Accordingly, the mapping mechanism responded as it would to a newly introduced perturbation, rather than leveraging previously learned adjustments (Iturralde and TorresOviedo, 2019). Due to the rapid nature of the Ramp Down, the mapping mechanism lacked sufficient time to learn the novel motor adjustments required for these perturbations – a process that typically takes several minutes, as shown by our baseline ramp tasks and control experiments. As mapping-related learning was negligible, the rigid recalibration adjustments dominated during this phase. Consequently, the walking pattern did not change to accommodate the gradually diminishing perturbation, leading to the emergence of aftereffects.”

We also revised the Discussion section “Mapping operates as memory-based in some people, structure-based in others”, to clarify the processes of interpolation and extrapolation (lines 689-700). This revision helps explain why mapping may generalize to a ramp-up faster than learning a perturbation perceived in the opposite direction (when considered together with the explanation that mapping operates relative to the new recalibrated equilibrium) In the former case (generalize to a ramp-up), a structure-based mapping can use the extrapolation computation: it leverages previous knowledge of which gait parameters should be modified and how – e.g., modulating the positioning our right foot to be more forward on the treadmill – but must extrapolate the specific parameter values – e.g., how more far forward. In the latter case (learning a perturbation perceived in the opposite direction), even a structure-based mapping would need to figure out what gait parameters to change completely anew – e.g., modulating the positioning of the foot in the opposite way, to be less forward, requires a different set of control policies.

We mentioned above that this illustration of the mapping mechanism relies on the assumption that the additional learning of the mapping mechanism in the second half of the Ramp Down is negligible. As part of our revisions for the “Modelling analysis for perceptual realignment”, we developed a new model – the perceptuomotor recalibration + mapping model (PM-ReMap) that extends the recalibration + mapping model by accounting for the possibility that Δ motor output is not constant in the second half of the Ramp Down (main points are at lines 355-275, and Figure 5; see response to Reviewer #1 (Public review), Comment 4, for a detailed explanation). We find that performance of the PM-ReMap model does not exceed that of the simpler recalibration + mapping model, suggesting that the Δ motor output does not change substantially in the second half of the Ramp Down. Note that, if the Δ motor output decayed in this phase, it could be due to forgetting or unlearning of the recalibration mechanism, or also it could be due to the mapping mechanism learning in the opposite direction than it did in adaptation. In the Results section, we focused on describing recalibration forgetting/unlearning for simplicity. However, in the Discussion section “Mapping may underly savings upon re-exposure to the same or different perturbation”, we explain in detail how the motor aftereffects also depend on the mapping mechanism learning in the opposite direction, as corroborated by our Control experiments and previous work. Therefore, the finding that the PM-ReMap model performance does not exceed that of the simpler recalibration + mapping model suggest that both effects – recalibration forgetting/unlearning and opposite-direction-learning of mapping – are not significant, nor is their combined effect on the Δ motor output.

(2) A more minor point, but given the sample size it is hard to be convinced about the individual difference analysis for structure learning (Figure 5). How clear is it that these two groups of subjects are fully separable and not on a continuum? The lack of clusters in another data set seems like a somewhat less than convincing control here.

We performed an additional analysis – a silhouette analysis – to confirm the presence of these clusters in our data (Methods, lines 1070-1072). The results, reported in Experiment 2 Results, lines 487-490, confirmed that there is strong evidence for the presence of these clusters:

“A silhouette analysis confirmed strong evidence for these clusters: the average silhouette score was 0.90, with 19 of 20 participants scoring above 0.7 – considered strong evidence – and one scoring between 0.5 and 0.7 – considered reasonable evidence (Dalmaijer et al., 2022; Kaufman and Rousseeuw, 1990; Rousseeuw, 1987).”

Reviewer #3 (Recommendations for the authors):

(1) I think there is far too much content pushed into the supplement. The other models and full model comparison should be in the main text, as should the re-analysis of previous data sets. Also, key discussion points should not be in the supplement either.

We reorganized the manuscript and SI Appendix substantially, including the changes recommended by the reviewer. Please refer to our response to “Reviewer #2 - Recommendations for the authors” for a detailed explanation.

(2) Line 649: in reaching the calibration system does respond to different error sizes; why not here?

We apologize for the confusion. Similar to reaching adaptation, the recalibration in walking adaptation also scales based on the error size experienced in adaptation. What we meant to convey is that, once a calibration has been acquired in adaptation, the recalibration process is rigid in that it can only change gradually. So if we jump the perturbation to a different value, the original calibration is transiently used until the system has the time to recalibrate again. For example, if we jump abruptly from the adaptation perturbation to a perturbation of zero in postadaptation, the adaptation calibration persists resulting in aftereffects.

We revised the manuscript to clarity these points. First, we explicitly report that forward model recalibration scales based on the error size experienced in adaptation:

“We next compared Medium Descend and Small Abrupt (1m/s or 0.4m/s perturbation), and found that recalibration contributed significantly more for the smaller perturbation (larger compensation_perceptual / compensation_motor-total in Small Abrupt than Medium Descend, Fig. 8A middle and Table S6).” (Control experiments Results, lines 422-425)

“the mapping described here shares some characteristics with explicit mechanisms, such as flexibility and modulation by error size” (Discussion, lines 630-631)

Additionally, we leverage the framework proposed by Tsay et al., 2024, to improve our explanation of the characteristics of the different learning mechanisms. Please refer to our response to “Reviewer #1 (Public review)”, Comment (1).

(3) It would be nice to see bar graphs showing model comparison results for each individual subject in the main text, and to see how many subjects are best fit by the SR+calibration model.

We included the recommended bar graphs to Figure 3 and Figure 5.

(4) Why exactly does the "perturbation" in Figure 3 have error bars?

In walking adaptation, the perturbation that participants experienced is closely dictated by the treadmill belt speeds, but not exactly, because participants are free to move their feet as they like, so that their ankle movement may not always match the treadmill belts exactly. Therefore, we record the perturbation that is actually experienced by each participant’s feet using markers. We then display the mean and standard error of this perturbation.

We moved the equation describing the perturbation measure from the Methods to the Experiment 1 Results (lines 131-133, Eq. 1-3). We believe this change will help the reader understand the measures depicted.

Author Response

The following is the authors’ response to the original reviews.

eLife assessment

This valuable work provides a near-complete description of the mechanosensory bristles on the Drosophila melanogaster head and the anatomy and projection patterns of the bristle mechanosensory neurons that innervate them. The data presented are solid. The study has generated numerous invaluable resources for the community that will be of interest to neuroscientists in the field of circuits and behaviour, particularly those interested in mechanosensation and behavioural sequence generation.

We express our gratitude to the Reviewers for their valuable suggestions, which significantly enhanced the manuscript. The revisions were undertaken, not with the expectation of acceptance, but rather driven by our sincere belief that these revisions would enhance the manuscript's impact for future readers.

Public Reviews:

Reviewer #1 (Public Review):

Sensory neurons of the mechanosensory bristles on the head of the fly project to the sub esophageal ganglion (SEZ). In this manuscript, the authors have built on a large body of previous work to comprehensively classify and quantify the head bristles. They broadly identify the nerves that various bristles use to project to the SEZ and describe their region-specific innervation in the SEZ. They use dye-fills, clonal labelling, and electron microscopic reconstructions to describe in detail the phenomenon of somatotopy - conserved peripheral representations within the central brain - within the innervation of these neurons. In the process they develop novel tools to access subsets of these neurons. They use these to demostrate that groups of bristles in different parts of the head control different aspects of the grooming sequence.

Reviewer #2 (Public Review):

The authors combine genetic tools, dye fills and connectome analysis techniques to generate a "first-of-its-kind", near complete, synaptic resolution map of the head bristle neurons of Drosophila. While some of the BMN anatomy was already known based on previous work by the authors and other researchers, this is the first time a near complete map has been created for the head BMNs at electron microscopy resolution.

Strengths:

(1) The authors cleverly use techniques that allow moving back and forth between periphery (head bristle location) and brain, as well as moving between light microscopy and electron microscopy data. This allows them to first characterize the pathways taken by different head BMNs to project to the brain and also characterize anatomical differences among individual neurons at the level of morphology and connectivity.

(2) The work is very comprehensive and results in a near complete map of all I’m head BMNs.

(3) Authors also complement this anatomical characterization with a first-level functional analysis using optogenetic activation of BMNs that results in expected directed grooming behavior.

Weaknesses:

(1) The clustering analysis is compelling but cluster numbers seem to be arbitrarily chosen instead of by using some informed metrics.

We made revisions to the manuscript that address this concern. Please see our response to “recommendations for authors” for a description of these revisions.

(2) It could help provide context if authors revealed some of the important downstream pathways that could explain optogenetics behavioral phenotypes and previously shown hierarchical organization of grooming sequences.

We made revisions to the manuscript that address this recommendation. Please see our response to “recommendations for authors” for a description of these revisions.

(3) In contrast to the rigorous quantitative analysis of the anatomical data, the behavioral data is analyzed using much more subjective methods. While I do not think it is necessary to perform a rigorous analysis of behaviors in this anatomy focused manuscript, the conclusions based on behavioral analysis should be treated as speculative in the current form e.g. calling "nodding + backward walking" as an avoidance response is not justified as it currently stands. Strong optogenetic activation could lead to sudden postural changes that due to purely biomechanical constraints could lead to a couple of backward steps as seen in the example videos. Moreover since the quantification is manual, it is not clear what the analyst interprets as backward walking or nodding. Interpretation is also concerning because controls show backward walking (although in fewer instances based on subjective quantification).

While unbiased machine vision-based methods would nicely complement the present work, this type of analysis is not yet working to distinguish between different head grooming movements. Therefore, we are currently limited to manual annotation for our behavioral analysis. That said, we do not believe that our manual annotation is subjective. The grooming movements that we examine in this work are distinguishable from each other through frame-by-frame manual annotation of video at 30 fps. Our annotation of the grooming and backward motions performed by flies are based on previous publications that established a controlled vocabulary defining each movement (Hampel et al., 2020a, 2017, 2015; Seeds et al., 2014). In this work, we added head nodding to this controlled vocabulary that is described in the Materials and methods. We have added additional text to the third paragraph of the Material and methods section entitled “Behavioral analysis procedures” that we hope better describes our behavioral analysis. This description now reads:

Head nodding was annotated when the fly tilted its head downward by any amount until it returned its head back in its original position. This movement often occurred in repeated cycles. Therefore, the “start” was scored at the onset of the first forward movement and the “stop” when the head returned to its original position on the last nod.

We do not make any firm conclusions about the head movements (nodding) and backwards motions. We refer to nodding as a descriptive term that would allow the reader to better understand what the behavior looks like. We make no firm conclusions about any behavioral functional role that either the nodding or the backward motions might have, with the exception of nodding in the context of grooming. We only suggest that the behaviors appear to be avoidance responses. Furthermore, backward walking was not mentioned. Instead we refer to backward motions. We are only reporting our annotations of these movements that do occur, and are significantly different from controls. We speculate that these could be avoidance responses based on support from the literature. Future studies will be required to understand whether these movements serve real behavioral roles.

Summary:

The authors end up generating a near-complete map of head BMNs that will serve as a long-standing resource to the Drosophila research community. This will directly shape future experiments aimed at modeling or functionally analyzing the head grooming circuit to understand how somatotopy guides behaviors.

Reviewer #3 (Public Review):

Eichler et al. set out to map the locations of the mechanosensory bristles on the fly head, examine the axonal morphology of the bristle mechanosensory neurons (BMNs) that innervate them, and match these to electron microscopy reconstructions of the same BMNs in a previously published EM volume of the female adult fly brain. They used BMN synaptic connectivity information to create clusters of BMNs that they show occupy different regions of the subesophageal zone brain region and use optogenetic activation of subsets of BMNs to support the claim that the morphological projections and connectivity of defined groups of BMNs are consistent with the parallel model for behavioral sequence generation.

The authors have beautifully cataloged the mechanosensory bristles and the projection paths and patterns of the corresponding BMN axons in the brain using detailed and painstaking methods. The result is a neuroanatomy resource that will be an important community resource. To match BMNs reconstructed in an electron microscopy volume of the adult fly brain, the authors matched clustered reconstructed BMNs with light-level BMN classes using a variety of methods, but evidence for matching is only summarized and not demonstrated in a way that allows the reader to evaluate the strength of the evidence. The authors then switch from morphology-based categorization to non-BMN connectivity as a clustering method, which they claim demonstrates that BMNs form a somatotopic map in the brain. This map is not easily appreciated, and although contralateral projections in some populations are clear, the distinct projection zones that are mentioned by the authors are not readily apparent. Because of the extensive morphological overlap between connectivity-based clusters, it is not clear that small projection differences at the projection level are what determines the post-synaptic connectivity of a given BMN cluster or their functional role during behavior. The claim the somatotopic organization of BMN projections is preserved among their postsynaptic partners to form parallel sensory pathways is not supported by the result that different connectivity clusters still have high cosine similarity in a number of cases (i.e. Clusters 1 and 3, or Clusters 1 and 2). Finally, the authors use tools that were generated during the light-level characterization of BMN projections to show that specifically activating BMNs that innervate different areas of the head triggers different grooming behaviors. In one case, activation of a single population of sensory bristles (lnOm) triggers two different behaviors, both eye and dorsal head grooming. This result does not seem consistent with the parallel model, which suggests that these behaviors should be mutually exclusive and rely on parallel downstream circuitry.

We made revisions to the manuscript that address this recommendation. Please see our response to “recommendations for authors” for a description of these revisions.

This work will have a positive impact on the field by contributing a complete accounting of the mechanosensory bristles of the fruit fly head, describing the brain projection patterns of the BMNs that innervate them, and linking them to BMN sensory projections in an electron microscopy volume of the adult fly brain. It will also have a positive impact on the field by providing genetic tools to help functionally subdivide the contributions of different BMN populations to circuit computations and behavior. This contribution will pave the way for further mechanistic study of central circuits that subserve grooming circuits.

Recommendations for the authors:

All three reviewers appreciated the work presented in this manuscript. There were also a few overlapping concerns that were raised that are summarised below, should the authors wish to address them:

Somatotopy: We recommend that the authors describe the extent of prior knowledge in more detail to highlight their contribution better.

We made revisions that better highlight the extent of prior knowledge about somatotopy. We describe how previous studies showed bristle mechanosensory neurons in insects are somatotopically organized, but these studies were not comprehensive descriptions of complete somatotopic maps for the head or body. To our knowledge, our study provides the first comprehensive and synaptic resolution somatotopic map of a head for any animal. This sets the stage for the complete definition of the interface between somatotopically-organized mechanosensory neurons and postsynaptic circuits, which has broad implications for future studies on aimed grooming, and mechanosensation in general. Below we itemize revisions to the Introduction, Discussion, and Figures to provide a clearer statement of the significance of our study as it relates to somatotopy.

(1) Newly added Figure 1 – figure supplement 1 more explicitly grounds the study in somatotopy, providing a working model of the organization of the circuit pathways that produce the grooming sequence. This model features somatotopy as shown in Figure 1 – figure supplement 1C.

(2) Figure 1 – figure supplement 1 is incorporated into the Introduction in the second, third, and fourth paragraphs, the first paragraph of the Results section titled “Somatotopically-organized parallel BMN pathways”, and the second and third paragraphs of the last Discussion section titled “Parallel circuit architecture underlying the grooming sequence”.

(3) We added text to the end of the fourth paragraph of the Introduction that now reads: “In this model, parallel-projecting mechanosensory neurons that respond to stimuli at specific locations on the head or body could connect with somatotopically-organized parallel circuits that elicit grooming of those locations (Figure 1 – figure supplement 1A-C). The previous discovery of a mechanosensory-connected circuit that elicits aimed grooming of the antennae provides evidence of this organization (Hampel 2015). However, the extent to which distinct circuits elicit grooming of other locations is unknown, in part, because the somatotopic projections of the mechanosensory neurons have not been comprehensively defined for the head or body.”

(4) There is a Discussion section that further explains the extent of prior knowledge and our contributions on somatotopy that is titled “A synaptic resolution somatotopic map of the head BMNs”. Additionally, the previous version of this section had a paragraph on the broader implications of our work as it relates to somatotopy across species. In light of the reviewer comments, we decided to make this paragraph into its own Discussion section to better highlight the broader significance of our work. This section is titled “First synaptic resolution somatotopic map of the head”.

The somatotopy isn't overtly obvious - perhaps they could try mapping presynaptic sites and provide landmarks to improve visualisation.

We made the following revisions to better highlight the head BMN somatotopy. One point of confusion from the previous manuscript version stemmed from us not explicitly defining the somatotopic organization that we observed. There seemed to be confusion that we were defining the head somatotopy based only on the small projection differences among BMNs from neighboring head locations. While we believe that these small differences indeed correspond to somatotopy, we failed to highlight that there are overt differences in the brain projections of BMNs from distant locations on the head. For example, Figure 5B (right panel) shows the distinct projections between the LabNv (brown) and AntNv (blue) BMNs that innervate bristles on the ventral and dorsal head, respectively. Thus, BMN types innervating neighboring bristles show overlapping projections with small projection differences, whereas those innervating distant bristles show non overlapping projections into distinct zones.

Our analysis of postsynaptic connectivity similarity also shows somatotopic organization among the BMN postsynaptic partners, as BMN types innervating the same or neighboring bristle populations show high connectivity similarity (Figure 8, old Figure 7). Below we highlight major revisions to the text and Figures that hopefully better reveal the head somatotopy.

(1) In the last paragraph of the Introduction we added text that explicitly frames the experiments in terms of somatotopic organization: “This reveals somatotopic organization, where BMNs innervating neighboring bristles project to the same zones in the CNS while those innervating distant bristles project to distinct zones. Analysis of the BMN postsynaptic connectome reveals that neighboring BMNs show higher connectivity similarity than distant BMNs, providing evidence of somatotopically organized postsynaptic circuit pathways.”

(2) We mention an example of overt somatotopy from Figure 5 in the Results section titled “EM-based reconstruction of the head BMN projections in a full adult brain”. The text reads “For example, BMNs from the Eye- and LabNv have distinct ventral and anterior projections, respectively. This shows how the BMNs are somatotopically organized, as their distinct projections correspond to different bristle locations on the head (Figure 5B,C).”

(3) In new Figure 8 (part of old Figure 7), we modified panels that correspond to the cosine similarity analysis of postsynaptic connectivity. The major revision was to plot the cosine similarity clusters onto the head bristles so that the bristles are now colored based on their clusters (C). This shows how neighboring BMNs cluster together, and therefore show similar postsynaptic connectivity. We believe that this provides a nice visualization of somatotopic organization in BMN postsynaptic connectivity. We also added the clustering dendrogram as recommended by Reviewer #2 (Figure 8A).

(4) In new Figure 8, we added new panels (D-F) that summarize our anatomical and connectomic analysis showing different somatotopic features of the head BMNs. Different BMN types innervate bristles at neighboring and distant proximities (D). BMNs that innervate neighboring bristles project into overlapping zones (E, example of reconstructed BM-Fr and -Ant neurons with non-overlapping BM-MaPa neurons) and show postsynaptic connectivity similarity (F, example connectivity map of three BM types on cosine similarity data).

(5) To accompany the new Figure 8D-F panels, we added a paragraph to summarize the different somatotopic features of the head BMNs that were identified based on our anatomical and connectomic analysis. This is the last paragraph in the Results section titled “Somatotopically-organized parallel BMN pathways”:

Our results reveal head bristle proximity-based organization among the BMN projections and their postsynaptic partners to form parallel mechanosensory pathways. BMNs innervating neighboring bristles project into overlapping zones in the SEZ, whereas those innervating distant bristles project to distinct zones (example of BM-Fr, -Ant, and -MaPa neurons shown in Figure 8D,E). Cosine similarity analysis of BMN postsynaptic connectivity revealed that BMNs innervating the same bristle populations (same types) have the highest connectivity similarity. Figure 8F shows example parallel connections for BM-Fr, -Ant, and -MaPa neurons (vertical arrows), where the edge width indicates the number of synapses from each BMN type to their major postsynaptic partners. Additionally, BMNs innervating neighboring bristle populations showed postsynaptic connectivity similarity, while BMNs innervating distant bristles show little or none. For example, BM-Fr and -Ant neurons have connections to common postsynaptic partners, whereas BM-MaPa neurons show only weak connections with the main postsynaptic partners of BM-Fr or -Ant neurons (Figure 8F, connections under 5% of total BMN output omitted). These results suggest that BMN somatotopy could have different possible levels of head spatial resolution, from specific bristle populations (e.g. Ant bristles), to general head areas (e.g. dorsal head bristles).

We also refer to Figure 8D-F to illustrate the different somatotopic features in the Discussion. These references can be found in the following Discussion sections titled “A synaptic resolution somatotopic map of the head BMNs (fourth paragraph)”, and “Parallel circuit architecture underlying the grooming sequence (second paragraph)”.

(6) In addition to improving the Figures, we provide additional tools that enable readers to explore the BMN somatotopy in a more interactive way. That is, we provide 5 different FlyWire.ai links in the manuscript Results section that enable 3D visualization of the different reconstructed BMNs (e.g. FlyWire.ai link 1).

Note: In working on old Figure 7 to address this Reviewer suggestion, we also reordered panels A-E. We believe that this was a more logical ordering than in the previous draft. These panels are now the only data shown in Figure 7, as the cosine similarity analysis is now in Figure 8. We hope that splitting these panels into two Figures will improve manuscript readability.

Light EM Mapping: A better description of methods by which this mapping was done would be helpful. Perhaps the authors could provide a few example parallel representations of the EM and light images in the main figure would help the reader better appreciate the strength of their approach.

We have done as the Reviewers suggested and added panels to Figure 6 that show examples of the LM and EM image matching (Figure 6A,B). We added two examples that used different methods for labeling the LM imaged BMNs, including MCFO labeling of an individual BM-InOc neuron and driver line labeling of a major portion of BM-InOm neurons using InOmBMN-LexA. These panels are referred to in the first paragraph of the Results section titled “Matching the reconstructed head BMNs with their bristles”. Note that examples for all LM/EM matched BMN types are shown in Figure 6 – figure supplement 2.

We had provided Figure 6 – figure supplement 2 in the reviewed manuscript that shows all the above requested “parallel representations of the EM and light images”. However, the Reviewer critiques made us realize that the purpose of this figure supplement was not clearly indicated. Therefore, we have revised Figure 6 – figure supplement 2 and its legend to make its purpose clearer. First, we changed the legend title to better highlight its purpose. The legend is now titled: “Matching EM reconstructed BMN projections with light microscopy (LM) imaged BMNs that innervate specific bristles”. Second, we added label designations to the figure panel rows that highlight the LM and EM comparisons. That is, the rows for light microscopy images of BMNs are indicated with LM and the rows for EM reconstructed BMN images are labeled with EM. Reviewer #3 had indicated that it was not clear what labeling methods were used to visualize the LM imaged BM-InOm neurons in Figure 6 – figure supplement 2N. Therefore, we added text to the figure and the legend to better highlight the different methods used. Panels A and B were also cropped to accommodate the above mentioned revisions.

The manuscript also provides an extensive Materials and methods section that describes the different lines of evidence that were used to assign the reconstructed BMNs as specific types. We changed the title to better highlight the purpose of this methods section to “Matching EM reconstructed BMN projections with light microscopy imaged BMNs that innervate specific bristles”. The evidence used to support the assignment of the different BMN types is also summarized in Figure 6 – figure supplement 3.

Parallel circuit model: The authors motivate their study with this. We're recommending that they define expectations of such circuitry, its alternatives (including implications for downstream pathways), and behavior before they present their results. We're also recommending that they interpret their behavioural results in the context of these circuits.

Our primary motivation for doing the experiments described in this manuscript was to help define the neural circuit architecture underlying the parallel model that drives the Drosophila grooming sequence. This manuscript provides a comprehensive assessment of the first layer of this circuit architecture. A byproduct of this work is a contribution that offers immediate utility and significance to the Drosophila connectomics community. Namely, the description of the majority of mechanosensory neurons on the head, with their annotation in the recently released whole brain connectome dataset (FlyWire.ai). In writing this manuscript, we tried to balance both of these things, which was difficult to write. We very much appreciate the Reviewers' comments that have highlighted points of confusion in our original draft. We hope that the revised draft is now clearer and more logically presented. We have made revisions to the text and provided a new figure supplement (Figure 1 - figure supplement 1) and new panels in Figure 8. Below we highlight the major revisions.

(1) The Introduction was revised to more explicitly ground the study in the parallel model, while also removing details that were not pertinent to the experiments presented in the manuscript.

The first paragraph introduces different features of the parallel model. To better focus the reader on the parts of the model that were being assessed in the manuscript, we removed the following sentences: “Performance order is established by an activity gradient among parallel circuits where earlier actions have the highest activity and later actions have the lowest. A winner-take-all network selects the action with the highest activity and suppresses the others. The selected action is performed and then terminated to allow a new round of competition and selection of the next action.” Note that these sentences are included in the third and fourth paragraphs of the last Discussion section titled “Parallel circuit architecture underlying the grooming sequence”.

The first paragraph of the Introduction now introduces a bigger picture view of the model that emphasizes the two main features: 1) a parallel circuit architecture that ensures all mutually exclusive actions to be performed in sequence are simultaneously readied and competing for output, and 2) hierarchical suppression among the parallel circuits, where earlier actions suppress later actions.

(2) Newly added Figure 1 – figure supplement 1 provides a working model of grooming (Reviewer # 1 suggestion). We now more strongly emphasize that the study aimed to define the parallel neural circuit architecture underlying the grooming sequence, focusing on the mechanosensory layer of this architecture. In particular, we refer to the new Figure 1 – figure supplement 1 that has been added to better convey the hypothesized grooming neural circuit architecture. Figure 1 – figure supplement 1 is incorporated into the Introduction (paragraphs two, three, and four), Results section titled “Somatotopically-organized parallel BMN pathways (first paragraph)”, and last Discussion section titled “Parallel circuit architecture underlying the grooming sequence (second and third paragraphs)”.

(3) New panels in Figure 8 update the model of parallel circuit organization as it relates to somatotopy (D-F). These panels show the parallel circuits hypothesized by the model, but also indicate convergence, with different possible levels of head resolution for these circuits. We describe above where these panels are referenced in the text.

(4) We added a new paragraph in the last Discussion section titled “Parallel circuit architecture underlying the grooming sequence” that better incorporates the results from this manuscript into the working model of grooming. This paragraph is shown below.

Here we define the parallel architecture of BMN types that elicit the head grooming sequence that starts with the eyes and proceeds to other locations, such as the antennae and ventral head. The different BMN types are hypothesized to connect with parallel circuits that elicit grooming of specific locations (described above and shown in Figure 1 – figure supplement 1A,C). Indeed, we identify distinct projections and connectivity among BMNs innervating distant bristles on the head, providing evidence supporting this parallel architecture (Figure 8D-F). However, we also find partially overlapping projections and connectivity among BMNs innervating neighboring bristles. Further, optogenetic activation of BMNs at specific head locations elicits grooming of both those locations and neighboring locations (Figure 9). These findings raise questions about the resolution of the parallel architecture underlying grooming. Are BMN types connected with distinct postsynaptic circuits that elicit aimed grooming of their corresponding bristle populations (e.g. Ant bristles)? Or are neighboring BMN types that innervate bristles in particular head areas connected with circuits that elicit grooming of those areas (e.g. dorsal or ventral head)? Future studies of the BMN postsynaptic circuits will be required to define the resolution of the parallel pathways that elicit aimed grooming.

Aside from this summary of major concerns, the detailed recommendations are attached below.

Reviewer #1 (Recommendations For The Authors):

I appreciate the quality and exhaustive body of work presented in this manuscript. I have a few comments that the authors may want to consider:

(1) The authors motivate this study by posing that it would allow them to uncover whether the complex grooming behaviour of flies followed a parallel model of circuit function. It would have been nice to have been introduced to what the alternative model might be and what each would mean for organisation of the circuit architecture. Some guiding schematics would go a long way in illustrating this point. Modifying the discussion along these lines would also be helpful.

We made several revisions to the manuscript that address this recommendation. Among these revisions, we added Figure 1 – figure supplement 1 that includes a working model for grooming. Please see above for a description of these revisions.

(2) The authors mention the body of work that has mapped head bristles and described somatotopy. It would be useful to discuss in more detail what these studies have shown and highlight where the gaps are that their study fills.

We made several revisions to the manuscript that address this recommendation. Please see above for a description of these revisions.

(3) The dye-fills and reconstructions that are single colour could use a boundary to demarcate the SEZ. This would help in orienting the reader.

We agree with Reviewer #1 that Figure 4 and its supplements could use some indicator that would orient the reader with respect to the dye filled or stochastically labeled neurons. The images are of the entire SEZ in the ventral brain, and in the case of some panels, the background staining enables visualization of the brain (e.g. Figure 4H,M,N. To help orient the reader in this region, we added a dotted line to indicate the approximate SEZ midline. This also enables the reader to more clearly see which of the BMN types cross the midline.

Midline visual guides were added for Figure 4, Figure 4 – figure supplement 2, Figure 4 – figure supplement 3, Figure 4 – figure supplement 4, Figure 4 – figure supplement 5, Figure 4 – figure supplement 6, Figure 4 – figure supplement 7, Figure 4 – figure supplement 8, Figure 6 – figure supplement 2.

(4) The comparison between the EM and the fills/clones are not obvious. And particularly because they are not directly determined, it would be nice to have the EM reconstruction alongside the dye-fills. This would work very nicely in the supplementary figure with the multiple fills of the same bristles. I think this would really drive home the point.

We made several revisions to the manuscript that address this recommendation. Please see above for a description of these revisions.

(5) Are there unnoticed black error-bars floating around in many of the gray-scale images?

The black bars were masking white scale bars in the images. We have removed the black bars and remade the images without scale bars. This was done for the following Figures: Figure 4, Figure 4 – figure supplement 2, Figure 4 – figure supplement 3, Figure 4 – figure supplement 4, Figure 4 – figure supplement 5, Figure 4 – figure supplement 6, Figure 4 – figure supplement 7, Figure 4 – figure supplement 8, Figure 6 – figure supplement 2.

Reviewer #2 (Recommendations For The Authors):

(1) The only point in the paper I found myself going back and forth between methods/supp and text was when authors discuss about the clustering. I think it would help the reader if a few sentences about cosine clustering used for connectivity based clustering were included in the main text. Also, for NBLAST hierarchical clustering, it would help if some informed metrics could be used for defining cluster numbers (e.g. Braun et al, 2010 PLOS ONE shows how Ward linkage cost could be used for hierarchical clustering).

Depending on where the cut height is placed on the dendrogram for cosine similarity of BMNs, different features of the BMN type postsynaptic connectivity are captured. As the number of clusters is increased (lower cut height), clustering is mainly among BMNs of the same type, showing that these BMNs have the highest connectivity similarity. As the number of clusters is reduced (higher cut height), BMNs innervating neighboring bristles on the head are clustered, revealing three general clusters corresponding to the dorsal, ventral, and posterior head. This reveals somatotopy based clustering among same and neighboring BMN types. The cut height shown in Figure 8 and Figure 8 – figure supplement 2 was chosen because it highlighted both of these features.

The NBLAST clustering shows similar results to the connectivity based clustering with respect to neighboring and distant BMN types. As the number of clusters increases BMNs of the same type are clustered, and these types can be further subdivided into morphologically distinct subtypes. As the number of clusters is reduced, the clustering captures neighboring BMNs. Thus, neighboring BMN types showed high morphology similarity (and proximity) with each other, and low similarity with distant BMN types.

Please see our responses to a Reviewer #3 critique below for further description of the clustering results.

On the same lines it would help if the clustering dendrograms were included in the main figure.

We thank Reviewer #2 for this comment. We have added the dendrogram to Figure 8A, a change that we feel makes this Figure much easier to understand.

(2) It could help provide intuition if the authors revealed some of the downstream targets and their implication in explaining the behavioral phenotypes.

While this will be the subject of at least two forthcoming manuscripts, we have added text to the present manuscript that provides insight into BMN postsynaptic targets. Our previous work (Hampel et al. 2015) described a mechanosensory connected neural circuit that elicits grooming of the antennae. While this previous study demonstrated that the Johnston’s organ mechanosensory neurons are synaptically and functionally connected with this circuit, our preliminary analysis indicates that it is also connected with BM-Ant neurons. We hypothesize that there are additional such circuits that are responsible for eliciting grooming of other head locations.

To better highlight potential downstream targets in the manuscript, we now mention the antennal circuit in the Introduction. This text reads: In this model, parallel-projecting mechanosensory neurons that respond to stimuli at specific locations on the head or body could connect with somatotopically-organized parallel circuits that elicit grooming of those locations (Figure 1 – figure supplement 1A-C). The previous discovery of a mechanosensory-connected circuit that elicits aimed grooming of the antennae provides evidence of this organization (Hampel 2015). However, the extent to which distinct circuits elicit grooming of other locations is unknown, in part, because the somatotopic projections of the mechanosensory neurons have not been comprehensively defined for the head or body.

There is also text in the Discussion that addresses this Reviewer comment. It describes the antennal circuit and mentions the possibility that other similar circuits may exist. This can be found in the third paragraph of the section titled “Circuits that elicit aimed grooming of specific head locations”.

(3) Authors find that opto activation of BMNs leads to grooming of targeted as well as neighboring areas. Is there any sequence observed here? i.e. first clean targeted area and then clean neighboring area? I wonder if the answer to this is something as simple as common post-synaptic targets which is essentially reducing the resolution of the BMN sensory map. Some more speculation on this interesting result could be helpful.

We appreciate and agree with this point from Reviewer #2, and have tried to better emphasize the possible implications for grooming that the overlapping projections and connectivity among BMNs innervating neighboring bristles may have. This is now better addressed in the Results and Discussion sections. Below we highlight where this is addressed:

(1) In the second paragraph of the Results section titled “Activation of subsets of head BMNs elicits aimed grooming of specific locations” we added text that suggests the possibility that grooming of the stimulated and neighboring locations could be due to the overlapping projections and connectivity. This text reads: This suggested that head BMNs elicit aimed grooming of their corresponding bristle locations, but also neighboring locations. This result is consistent with our anatomical and connectomic data indicating that BMNs innervating neighboring bristles show overlapping projections and postsynaptic connectivity similarity (see Discussion).

(2) In the fourth paragraph of the Discussion section titled “A synaptic resolution somatotopic map of the head BMNs”, we added a sentence to the end of the fourth paragraph that alludes to further discussion of this topic. This sentence reads: This overlap may have implications for aimed grooming behavior. For example, neighboring BMNs could connect with common neural circuits to elicit grooming of overlapping locations (discussed more below).

(3) In the fourth paragraph of the Discussion section titled “Circuits that elicit aimed grooming of specific head locations” there is a paragraph that mentions the possibility of mechanosensory convergence onto common postsynaptic circuits to promote grooming of the stimulated area, along with neighboring areas. This paragraph is below.

We find that activation of specific BMN types elicits both aimed grooming of their corresponding bristle locations and neighboring locations. This suggests overlap in the locations that are groomed with the activation of different BMN types. Such overlap provides a means of cleaning the area surrounding the stimulus location. Interestingly, our NBLAST and cosine similarity analysis indicates that neighboring BMNs project into overlapping zones in the SEZ and show common postsynaptic connectivity. Thus, we hypothesize that neighboring BMNs connect with common neural circuits (e.g. antennal grooming circuit) to elicit overlapping aimed grooming of common head locations.

(4) In the new second paragraph of the Discussion section titled “Parallel circuit architecture underlying the grooming sequence” we further discuss the issue of the BMN “sensory map. This paragraph is below.

Here we define the parallel architecture of BMN types that elicit the head grooming sequence that starts with the eyes and proceeds to other locations, such as the antennae and ventral head. The different BMN types are hypothesized to connect with parallel circuits that elicit grooming of specific locations (described above and shown in Figure 1 – figure supplement 1A,C). Indeed, we identify distinct projections and connectivity among BMNs innervating distant bristles on the head, providing evidence supporting this parallel architecture (Figure 8D-F). However, we also find partially overlapping projections and connectivity among BMNs innervating neighboring bristles. Further, optogenetic activation of BMNs at specific head locations elicits grooming of both those locations and neighboring locations (Figure 9). These findings raise questions about the resolution of the parallel architecture underlying grooming. Are BMN types connected with distinct postsynaptic circuits that elicit aimed grooming of their corresponding bristle populations (e.g. Ant bristles)? Or are neighboring BMN types that innervate bristles in particular head areas connected with circuits that elicit grooming of those areas (e.g. dorsal or ventral head)? Future studies of the BMN postsynaptic circuits will be required to define the resolution of the parallel pathways that elicit aimed grooming.

(4) If authors were to include a summary table that shows all known attributes about BMN type as columns that could be very useful as a resource to the community. Table columns could include attributes like "bristle name", "nerve tract", "FlyWire IDs of all segments corresponding to the bristle class". "split-Gal4 line or known enhancer" , etc.

We provided a table that includes much of this information after the manuscript had already gone out for review. We regret that this was not available. This is now provided as Supplementary file 3. This table provides the following information for each reconstructed BMN: BMN name, bristle type, nerve, flywire ID, flywire coordinates, NBLAST cluster (cut height 1), NBLAST cluster (cut height 5), and cosine cluster (cut height 4.5). Note that the driver line enhancers for targeting specific BMN types are shown in Figure 3I.

Specific Points:

Figure 4C-V:

• I find it a bit difficult to distinguish ipsi- from contra-lateral projections. Maybe indicate the midline as a thin, stippled line?

We thank the Reviewer #2 for this suggestion. We have now added lines in the panels in Figure 4C-V to indicate the approximate location of the midline. We also added lines to the Figure 4 – figure supplements as described above.

I think this Fig reference is wrong "the red-light stimulus also elicited backward motions with control flies (Figure 6B,C, control, black trace, Video 5)." should be Fig 8B,C

We have fixed this error.

Reviewer #3 (Recommendations For The Authors):

Introduction:

Motivating this study in terms of understanding the neural mechanisms that execute the parallel model seems to overstate what you will achieve with the current study. If you want to motivate it this way, I suggest focusing on the grooming sequence of the head along (eyes, antennae, proboscis).

We made several revisions to the manuscript that address this recommendation. Please see above for a description of these revisions. Please note that many of the revisions focus on the head grooming sequence. We also made minor revisions to the Introduction that further emphasize the focus on head grooming.

Results:

Figure 1. Please indicate that this is a male fly in either the figure title or in the figure itself.

We added a male symbol to Figure 1A.

Figure 3. Panel J is referenced in the main body text and in the figure caption, but there is no Fig 3J.

Panel J is shown in the upper right corner of Figure 3. We realize that the placement of this panel is not ideal, but this was the only place that we could fit it. Additionally, the panel works nicely at that location to better enable comparison with panel C. We have revised the text in the Figure 3 legend to better highlight the location of this Figure panel: “Shown in the upper right corner of the figure are the aligned expression patterns of InOmBMN-LexA (red), dBMN-spGAL4 (green), and TasteBMN-spGAL4 (brown).”

We also added text to a sentence in the results section entitled “Head BMNs project into discrete zones in the ventral brain” that indicates the panel location. This text reads: To further visualize the spatial relationships between these projections, we computationally aligned the expression patterns of the different driver lines into the same brain space (Figure 3J, upper right corner).

Matching the BMNs to EM reconstructions: why cut the dendrogram at H=5? Would be better to determine cluster number using an unbiased method.

To match the morphologically distinct EM reconstructed BMNs to their specific bristles, we relied on different lines of evidence, including NBLAST results (discussed more below), dye fill/stochastic labeling/driver line labeling matches, published morphology, nerve projection, bristle number, proximity to other BMNs, and postsynaptic connectivity (summarized in Figure 6 – figure supplement 3). The following Materials and methods section provides a detailed description of the evidence used to assign each BMN type in “Matching EM reconstructed BMN projections with light microscopy imaged BMNs that innervate specific bristles”. In many cases, BMN type could be assigned with confidence solely based on morphological comparisons with our light level data (e.g. dye fills), in conjunction with bristle counts to indicate an expected number of BMNs showing similar morphology. Thus, the LM/EM matches and NBLAST clustering were largely complementary.

The EM reconstructed BMNs were matched as particular BMN types, in part based on examination of the NBLAST data at different cut heights. NBLAST clustering of the BMNs revealed general trends at higher and lower cut heights (Figure 6 – figure supplement 1A, Supplementary file 3). The lowest cut heights included mostly BMNs of the same type innervating the same bristle populations, and smaller clusters that subdivided into morphologically distinct subtypes (see Supplementary file 3 for clusters produced at cut height 1). This revealed that BMNs of the same type tended to show the highest morphological similarity with each other, but they also showed intratype morphological diversity. Higher cut heights produced clusters of BMNs innervating neighboring bristles populations (e.g. ventral head BMNs), showing high morphological similarity among neighboring BMN types.

We selected the cut height 5 shown in Figure 6 – figure supplement 1A,B because it captures examples of both same and neighboring type clustering. For example, it captures a cluster of mostly BM-Taste neurons (Cluster 16), and neighboring BMN types, including those from the dorsal head (Cluster 14) or ventral head (Cluster 15).

Based on reviewer comments, we realized that the way we wrote the BMN matching section in the Results indicated more reliance on the NBLAST clustering than what was actually necessary, distorting the way we actually matched the BMNs. Therefore, we softend the first couple of sentences to place less emphasis on the importance of the NBLAST. We also indicated that the readers can find the resulting clusters at different cut heights, referring to Figure 6 – figure supplement 1A and Supplementary file 3. The first two sentences of the first paragraph in the Results section titled “Matching the reconstructed head BMNs with their bristles” now read:

The reconstructed BMN projections were next matched with their specific bristle populations. The projections were clustered based on morphological similarity using the NBLAST algorithm (example clustering at cut height 5 shown in Figure 6 – figure supplement 1A,B, Supplementary file 3, FlyWire.ai link 2) (Costa et al., 2016). Clusters could be assigned as BMN types based on their similarity to light microscopy images of BMNs known to innervate specific bristles.

The number of reconstructed BMNs is remarkably similar to what is expected based on bristle counts for each group except for lnOm. Why do you think there is such a large discrepancy there?

We believe that there is a discrepancy between the number of reconstructed BM-InOm neurons and the number expected based on InOm bristle counts because these bristle counts were based on few flies and these numbers appear to be variable. We did not further investigate the numbers of InOm bristles in this manuscript because we only needed an estimate of their numbers, given that there is over an order of magnitude difference in the eye bristles versus any other head bristle population. Therefore, we could relatively easily conclude that the head BMNs were related to the InOm bristles, based on their sheer numbers and their morphology.

Figure 6 - figure supplement 2N, please describe these panels better. Main text says the upper image is from lnOmBMN-LexA, but the figure legend doesn't agree.

We have added text to the figure legend that now makes the contents of panel 2N clear to the reader. Further, we now indicate in the figure legend for each panel, the method used to obtain the labeled neurons (i.e. fill, MCFO, driver), to avoid similar confusion for the other panels.

Figure 6 - figure supplement 4D. How frequently is there a mismatch between the number of BMNs for a given type across hemispheres?

Although the full reconstruction of the BMNs on both sides of the brain was beyond the scope of this work, the BMNs on both sides have since been reconstructed and annotated (Schlegal et al. 2023). We plan to provide more analysis of BMNs on both sides of the brain in a forthcoming manuscript. However, the BMN numbers tend to show agreement on both sides of the brain. The table below shows a comparison between the two sides:

Author response table 1.

Figures 6 and 7. It would be helpful to include a reference brain in all panels that show cluster morphology. Without landmarks there is nothing to anchor the eye to allow the reader to see the described differences in BMN projection zones and patterns.

While we apologize for not making this specific change, we have made revisions to other parts of the manuscript to better highlight the somatotopic organization among the BMNs (revisions described above). Please note that we now provide FlyWire.ai publicly available links that enable readers to view the BMN projections in 3D. They can also toggle a brain mesh on and off to provide spatial reference.

"BMN somatotopic map": It would be helpful to show or describe in more detail what the unique branch morphology for each zone is. It is quite difficult to appreciate, as the groups also have a lot of overlap. Would the unique regions that the BMN groups innervate be easier to see if you plotted presynaptic sites by group? I am left unsure about whether there is a somatotopic map here.

We made several revisions to the manuscript that address this recommendation. Please see above for a description of these revisions. Please note that we did not examine the fine branch morphological differences between BMN types having overlapping projections. Showing these differences would require more extensive anatomical analysis that is beyond the scope of this work. For showing definitive somatotopy, we focused on the overt differences between BMNs innervating bristles at distant locations on the head.

Overall the strict adherence to the parallel model impacts the interpretation of the data. It would be helpful for the authors to discuss which aspects of the current study are consistent with the parallel model and which results are not consistent.