Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The study by Yasmin & colleagues tackles an important question, what is the molecular nature of specificity that arises from otherwise highly similar proteins. In this case, they focus on two proteins with epigenetic activity, DNMT3A and DNMT3B, using a functional readout of their ability to methylate DNA in a model that specifically requires DNMT3B function at a subset of the genome, i.e. DNMT3B-dependent regions. This includes characterizing the role of DNMT3B in these regions in stem cell-to-embryoid body differentiation experiments, using genomic assays to probe DNA methylation dynamics. By removing DNMT3B and ectopically expressing a variety of sophisticated mutants, the authors attempt to show the protein domains required for specificity. However, several questions remain about the strength of the data to support the claims, particularly with respect to the ectopically expressed mutant DNMT3B proteins.

Major comments:

1. The strength of this study is in the very nice addback strategy for probing DNMT3B specificity, where the designed mutants seem highly useful to ask critical questions. However, the stability of the mutant proteins (i.e. cellular expression levels) and question of protien levels in the nucleus are insufficient evidence for the conclusions stated in the paper. With the exception of the Dnmt3b1-KI clones (top panel fig 3B), it seems like most mutants are not expressed at wildtype levels. How much of the results are driven by differences in expression, relative to the wildtype protein? While this a technically challenging problem, there are various methods to establish roughly matched expression such as integration into a stronger locus for expression or tuning the promoter sequence for expression of a construct. Given the mutants are key for the main conclusions of the study, this seems critical to address, though would substantially increase effort required for the paper.
2. Characterisation of the datasets supporting effects seems lacking in several instances. For example, the text states that DMNT3B null cells behave similarly to wildtype cells but supporting data (FigS2A-C) or that Dnmt3b1-KI and Dnmt3b3-KI behave normally with respect to differentiation (FigS4C), seem insufficient evidence for this, with largely summary plots supporting the statements. Similarly, several of the MBDseq datasets seem discordant, such as FigS2G or FigS4D(right panel) where the x-y axis for scatterplots are clearly not equivalent suggesting global effects on the data. The authors should also clearly demonstrate the levels of DNMT3A throughout their EB timecourse for mutant lines, where this seems especially important given their readout is DNA methylation dynamics.
3. An optional analysis that could support the claims of the paper would be to contrast the effect sizes in their cellular model with existing datasets that profile DNA methylation dynamics in vivo, where these have been captured at early developmental levels. This would nicely show that their functional readout in relation to normal processes.

Minor comments:

1. Several figures require addressing, listed here:
   • Fig1B the points are not so legible when overplotted, consider reducing the size of the datapoint circles or turning into "*" representations.
   • Fig4I seems not to have a figure legend.
   • FigS2G should be represented as a square and not as a rectangle, as this visually condenses on axis relative to the other.
   • FigS3A is unclear, could more be added to the legend to describe what exactly is the schematic representing?
   • FigS3D the axis seems not aligned with the barplot positioning?
2. The Dnmt3b-PAS-KI clone 1 does not seem to well-cluster with the 2nd and 3rd clone, could this be a clonal effect at the global level?
3. The text states (page 7, third paragraph) that in the two differentiation models the identity of the CGIs that exhibit different dynamics largely match, though no direct comparison (i.e. delta-delta effect) is show, rather a summary plot of either is presented side-by-side. This seems insufficient evidence of the statement, and a direct comparison of the fold changes would help.
4. The clonal effect sizes would benefit from more quantitative comparisons throughout the manuscript, broken down to raw data. For example, the statement in page 8 paragraph two that the effects on independent clones were fully consistent is show from largely a PCA plot, which seems incomplete evidence that replicates behave consistently. More transparent analysis of clonality from the raw data would be helpful for the reader.
5. The statement in the discussion that the authors experimental system affords 'homogeneous and highly synchronised onset and progression of XCI", but it seems unclear from the data provided in the manuscript that cells exhibit differentiation in a synchronized manner. Softening this statement seems apt here.

Significance

The question of specificity is highly important, not just to the field of epigenetics and DNA methylation where this study is particularly relevant, but also to a broader audience. Many of our cells proteins are highly homologous but have nevertheless highly divergent activities. Molecular explanations of specificity are therefore critical to understand phenotype and how traits can be acquired through gene paralogue evolution. Here, by focusing on a particularly apt example, the similar DNMT3A/B proteins, this study offers a nice breakdown with the potential to tie back the results to locus specific activity in the genome. The strongest aspect is the comparison of sophisticated mutants in a matched experimental setting, however, the experiments do not seem sufficient to support the broad conclusions of the study. From a genomics standpoint, the experimental setup is impressive, but requires additional work to show that matched expression levels of wildtype/mutant proteins still maintains the phenotypes reported.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

DNA methylation controls gene expression and genome stability during development and in healthy adult cells. It is frequently abnormal in diseases including cancer. Therefore, there is a clear impetus for the community to better understand the mechanisms that underlie DNA methylation patterns in mammalian genomes, including how the mark is deposited on specific regions. Of particular interest are CpG islands, which correspond to the majority of promoters, and are typically devoid of methylation. However, in specific cases, including during differentiation and X chromosome inactivation, CpG islands acquire extensive DNA methylation in a de novo methylation process. There are 2 de novo DNA methyltransferases expressed in the embryo: DNMT3A and DNMT3B. They are globally similar, but only DNMT3B contributes to de novo DNA methylation during X inactivation (Gendrel 2012, from the authors' lab). The question of this paper is: why is that?

The model system used by the authors is female mouse ES cells, which during differentiation in vitro inactivate one of their X chromosomes. They use a hybrid line to distinguish parental alleles, and a genetic trick to ensure that the same chromosome is inactivated in all cells. Figure 1 validates the system, showing that CGIs on the inactivated X acquire DNA methylation during the differentiation process into EBs, along with some autosomal CGIs (this is done by MBD-seq). Some are fast to gain methylation, some slower. A similar analysis is carried out during differentiation into NPCs.

The authors then move on to functional experiments by knocking out DNMT3B in their system. The KO clones are extensively characterized (Fig 2 and S2). They lack DNMT3B but retain DNMT3A levels similar to WT. However, they fail to methylate the majority of CGIs during X inactivation, confirming that DNMT3B (and not DNMT3A) is the principal actor in this process.

The next question is: which domain(s) of DNMT3B is/are involved in this function. For this, the authors rescue their KO clones with cDNAs encoding different isoforms of DNMT3B, namely 3B1 (active) and 3B3 (inactive). They found that 3B1 fully rescued proper DNA methylation and gene expression during differentiation, whereas 3B3 had no effect (Fig 3 and S4).

Having found that 3B1 expression fully rescues the DNMT3B KO, they move on to a more precise delineation of the important domains (Fig 4). Domain-swapping experiments show that the catalytic domain itself does not contribute to the specificity of DNMT3B (see my note on this experiment below). A similar strategy is then employed to test the contribution of the PWWP and ADD domains to 3B function. I did not find this part very clear. My understanding is that the swapped rescue construct has some activity on gene bodies even before differentiation. I gather from the lower part of Panel 4F that the PAS construct mostly fails to rescue DNA methylation during differentiation, but I am a little confused by the phrasing. It would be very easy to solve this with a summary graphic.

Lastly, they move on to an examination of the Nter part of DNMT3B, using their domain swapping/rescue approach (Figures 5 and S6). A first experiment suggests that the Nter of 3A cannot substitute for the Nter of 3B (see note below). A second set of experiments shows that the presence of residues 1-218 is necessary for DNMT3B to function, and that smaller deletions within this region also inactivate the protein (see notes below).

The paper is very well written. The figures are clear. The experiments are well controlled and correctly interpreted, except for the points below.

Fig 4A: it is looking like the DNMT3B1-3A-Cat rescue construct is vastly overexpressed relative to the endogenous protein. This is surprising as the DNMT3B1 rescue construct is not (Figure 3B). What is the reason for this? Is a different promoter or rescue method being used? I feel that the overexpression level weakens the conclusion that the catalytic domain of 3A can perfectly replace that of 3B.

Fig 4G: similarly, the two different DNMT3B-PAS-KI clones show widely different levels of DNMT3B expression. Are they both used to generate the data of Fig 4H? A third rescue clone is shown in Fig S5B. What experiment(s) was it used for?

The clarity of the section concerning DNMT3B-PAS-KI clones can be improved easily.

Fig 5B: the DNMT3a2(N)3b-KI clones show 3B expression levels that are ~10% of endogenous. I find it hard to conclude from this that the Nter of 3A cannot replace the Nter of 3B. Unless the authors can show that a 10% level of 3B expression is enough to fully rescue the KO of 3B.

Fig S6D: same comment for the DeltaA and DeltaB construct. I am not seeing the data for DeltaC. In the absence of expression data, the methylation data for this mutant are not interpretable.

Fig 5B. Is it clear that the Delta 1-218 protein is nuclear? What about the Delta A-E mutants?

I suggest that the authors add the following paper to their reference list: Wapenaar EMBO Rep 2024 PMID: 39528729

Significance

I feel that the target audience is somewhat specialized. In addition, the novelty of the paper is diminished by the existence of published papers, in particular: DNMT3B PWWP mutations cause hypermethylation of heterochromatin. Taglini F, ..., Sproul D. EMBO Rep. 2024 Mar;25(3):1130-1155. doi: 10.1038/s44319-024-00061-5. Epub 2024 Jan 30. PMID: 38291337

This paper uses a different system (human cancer cells), but arrives at the same conclusion, ie the Nter of DNMT3B is necessary for de novo DNA methylation. In addition, it shows that the Nter interacts with HP1.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

In this study the authors used a previously established mESCs iXist-ChrX cell line to investigate the mechanisms underpinning developmentally regulated CGI methylation through differentiation into embryoid bodies and MBD-seq profiling. They show that their system recapitulates developmentally regulated DNA methylation at CGIs on the X chromosome and autosomes before using a knockout and rescue system to determine that this is dependent on the DNMT3B1 isoform. Through domain swap experiments, they then go on to suggest that this requires the PWWP-ADD domain and that the N-terminal region of DNMT3B1.

Major comments:

• Are the key conclusions convincing?
• Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?
• Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.
• Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.
• Are the data and the methods presented in such a way that they can be reproduced?
• Are the experiments adequately replicated and statistical analysis adequate?

Overall, this is an interesting study that provides insights into the role of different domains of DNMT3B in establishing DNA methylation patterns during development. However, it suffers from poor annotation and description throughout. More specific comments are:

1. The manuscript provides some details as to the replication of experiments but fails to show the replicate data in the vast majority of cases. Instead, representative data is presented alongside global correlations for some experiments. However, global correlations can mask differences between replicates. The data from all replicates should be shown in the manuscript and clear details provided regarding the replication strategy in the methods and figure legends. For example, were the different knock-in clones generated from independent DNMT3B knockout clones? For individual experiments this would be a minor point however, collectively this is a major point. It is particularly important given the variation highlighted in point 2 below.
2. Different DNMT3B knock-in clones show high variability in expression levels. Have the authors investigated whether the discrepancy in protein levels contributes to the differences in methylation patterns seen? A non-comprehensive list of examples is given in the minor comments section.
3. There is variation in the level of expression of different DNMT3B constructs detected by Western blot. Could this be caused by differences in protein stability? It would be helpful to see an assessment of protein stability to determine whether this contributes to the variable expression. For example, the DNMT3B3-KI has lower levels than the DNMT3B1-KI (Figure 3B) and this could potentially contribute to the differences in DNA methylation observed.
4. The study lacks statistical tests to support the conclusions drawn from the analysis of the sequencing data. For example, are the differences in CGI methylation between DNMT3B1-KI and DNMT3B3-KI statistically significant? For individual analyses this would be a minor comment but given the lack throughout the study, this classes as a major comment.
5. Chromatin marks play major roles in DNMT3A and DNMT3B recruitment (Tibben & Rothbart 2024) and the N-terminal region, PWWP and ADD domains have direct or indirect chromatin reading activity. However, the manuscript does not detail the chromatin environment of the CGIs studied. This could potentially be addressed through experiments, analysis of existing data or discussion.
6. As the authors state in their discussion, MBD-seq may only detect very dense methylation. This could potentially obscure lower levels of DNA in some conditions. Analysis of a few loci by alternative methods, such as targeted bsPCR or EM-PCR would help support key results and rule out the possibility that some of the rescue constructs are able to partially rescue DNA methylation patterns.
7. While some expression data is shown, there is currently no investigation as to whether the different DNMT3B domain swap constructs have impact on transcriptional silencing on Xi/autosomal sites.
8. The text relating to the section on the PWWP-ADD domain is very brief and currently unclear. Expanding this section and specifying which data are derived from ES cells vs differentiated cells would help to clarify. We also suggest that it would be clearer to move this data to the main figure and to move the results of the catalytic domain experiments, which are negative, to the supplementary.
9. The authors suggest that PWW-ADD domain region of DNMT3B is required for developmental methylation of Xi and autosomal CGIs. However, there is no further dissection as to whether this requirement is due to the PWWP, ADD or the intervening region.
10. Throughout the text autosomal and Xi CGIs are both analysed. The introduction highlights SMCHD1 as important in methylating CGIs on Xi but that PCGF6 complex for the autosomal targets. This suggests separate mechanisms target DNMT3B to these loci. However, based on the results presented here, these two different types of targets have similar DNMT3B1 domain requirements. It would be interesting to see discussion with regard to this point in the manuscript.

Minor comments:

• Specific experimental issues that are easily addressable.
• Are prior studies referenced appropriately?
• Are the text and figures clear and accurate?
• Do you have suggestions that would help the authors improve the presentation of their data and conclusions?
• Introduction: Methylation of tumour suppressor gene CGI promoters is mentioned alongside examples of developmental CGI methylation. It would be useful to place rare event in context. Methylation of CGIs is common in cancer but very few of these correspond to tumour suppressor genes. It would also be useful to discuss how DNMT3B might be involved in these events.
• Introduction: The authors mention the DNMT3A1's N-terminal region recruits it to H2AK119ub1. It would also be useful to discuss recent work on the N-terminal of DNMT3B which can bind HP1-alpha to mediate H3K9me3 recruitment (Taglini et al 2024). This paper is currently only cited in the discussion.
• Throughout the manuscript DNMT3B1 is referred to as the catalytic isoform, however it is not the only catalytic isoform of DNMT3B.
• Page 5: 'the presence of non-catalytic isoforms, notably DNMT3L and DNMT3B3'. This statement incorrectly suggests DNMT3L is a non-catalytic isoform of DNMT3B.
• The authors refer to the protein complex as heterodimers when referring to a DNMT3B -DNMT3L or DNMT3B3-DNMT3A complex. However, the consensus of structural studies is that they form tetramers.
• Marker sizes are not included on blots with the exception of Figure 3B.
• Western blots are cropped closely. It would be useful if full blots were shown in the supplementary particularly given the presence of extra bands in some blots and different DNMT3B isoforms.
• Details about the Western blot methods (eg visualisation and antibodies) are missing.
• It would be useful if the size of different groups was annotated in plots. This is given in Figure 1D for example but not in Figure 2E.
• The authors show data confirming DNMT3B knockout by western blot. However, they do not provide details of the strategy for generating the knockout (ie vector used, sgRNAs, screening process). Could the authors also provide additional details as to whether there is any sequencing to confirm the results on the knockout?
• Page 9: The authors state "Since Dnmt3b-/- cells have normal levels of DNMT3A", but show no data to support this statement. This is particularly relevant as they have generated new DNMT3B knockout clones for this study so they cannot be assumed to behave similarly to previous studies.
• Figure 1B: There is poor PCA clustering between replicates at some timepoints, particularly Day 8.
• Figure 1G: Different colours are used for the different timepoints in this figure. We are unclear if this is deliberate.
• Page 7: The authors state "... the two categories largely matched between the two differentiation systems (Figure S1C-G)". It is difficult to draw this interpretation from the data presented as no explicit overlap is shown.
• Figure 2E: MBD-seq peaks for DNMT3B-independent loci in WT sample have a dip in the middle of the peak (also seen in Figure 5F). Could the authors explain why this might be and why it only appears in some experiments?
• Clarification as to whether the DNMT3B -dependent and -independent loci are located on chromosome X or autosomes (e.g.: Figure 2D, E).
• Figure S2C: the chromatin RNA-seq is not explained in the text or figure.
• Figure S2E: Suggests WT is one of 5 replicates. The authors should show all replicates.
• Figure S2H: What genes are included in the metaplot?
• DNMT3B rescue knock-in clones. As shown in figure 2A, there are two different Dnmt3b-/- cell line clones. Could the authors clarify whether all the CRISPR KI clones are produced from the same parental Dnmt3b-/- cell line clone?
• Figure 3B: The two clones shown for Dnmt3b3-KI have variable expression. Do the individual replicates for the Dnmt3b3-KI clones show similar methylation patterns?
• Figure 3E: in the MBD-seq metaplots, there is a peak present at -/+ 4kb. What are these peaks and why do they appear at 4kb distance? Similar peaks are seen in other metaplots.
• In figure 3F, the signal for both Dnmt3b1-KI and Dnmt3b3-KI at DNMT3B-independent CGIs is higher than in the KO. This suggests that these may not be DNMT3B independent but this point is unaddressed in the text.
• Figure S3A: it is currently unclear what has been modelled in this figure, adding labels of what has been plotted along the x- and y- axis may aid in understanding.
• Figure 4A: Dnmt3b1-3a-Cat-KI appears very highly expressed. Is the WT shown the endogenous protein? Could this higher expression be because the chimeric protein is more stable than DNMT3B1? There are also multiple bands on this blot.
• Plots panels are inconsistently ordered, e.g.: Figure 3F is dependent then independent. 4F is independent then dependent.
• Figure 4G: the expression level of the Dnmt3b-PAS-KI varies significantly between the clones shown. There are also two bands on the blot, both for the wt and KI. Clarify if WT is endogenous.
• Figure 4H: The figure lacks a legend to indicate the scale of the colour density used.
• Figure 4F,H: Could the authors clarifying what data (clones and number of replicates) are presented in the representative plots. Does the different protein levels between the clones result in any differences in DNA methylation?
• Page 12: The authors cite Boyko et al when discussing potential differences between the ADD domains of DNMT3A and B. However, they do not cite the study of Lu et al., 2023 (https://doi.org/10.1093/nar/gkad972) which reaches a different conclusion.
• Figure S4A: The position of the 750 residue is inconsistent across the isoforms in this schematic.
• Figure 5A: Schematic suggests the chimeric protein is DNMT3A2(N)-3B. However, DNMT3A2 lacks the N terminal region so presumably this should be DNMT3A1(N)-3B. This applies other figure panels using this construct.
• Figure 5B: Many lanes on this blot are unlabelled and it would be useful to clarify what these extra lanes show.
• Figure 5C: For Dnmt3a2(N)3b-KI the levels of methylation appear to be lower than Dnmt3b-/- and it would be useful to understand why this might be the case.
• Figure 5D: would be helpful to indicate which CGIs are DNMT3B dependent and independent.
• Figure 5F: Dnmt3a2(N)3b-KI data not included for the autosomal peaks
• Figure 5G, H: It is difficult to see if there are any differences between the deletions in this heatmap. For example, it appears that levels of methylation on autosomal DNMT3B-dependent loci are very similar between the KO and rescue constructs. ∆D also appears to have a lesser effect than the other deletions on the Xi CGIs. A more quantitative representation of the data would help with interpretation.
• S5E has different colour scale to other heatmaps. Red is low and in other heatmaps red is high.
• Figure S6C: sequence conservation is shown for primates. However as mouse Dnmt3b is used throughout the paper, including the mouse NT would be a useful comparison. This is particularly relevant given that the NT is the region that varies the most between mouse and human proteins (Molaro et al 2020 https://doi.org/10.1093/molbev/msaa044 ).
• Figure S6D: There is variable expression levels between the clones of the different deletions. The deletion ∆C is also not shown in this figure meaning that no data is shown to support the statement that it is unstable.
• It would be useful to clarify in the text that "deletion of residues 98-146" corresponds to ∆C. It is also unclear why MBD-seq data for this deletion was included if it is unstable.
• Discussion, page 15: The authors propose that DNMT3B could directly bind to H3K9me3. However, a study they cite, Taglini et al., 2024 (Figure S8C, D), suggests this is not the case.
• Discussion: When discussing regulatory element methylation on Xi. An uncited statement is included: 'This observation may help to explain a prior observation that loss of DNMT3B1 alone does not result in significant de-repression of Xi during embryogenesis'. However this model appears contradictory to the observation that DNMT1 is not required for Xi silencing, given that DNMT1 KO embryos would be expected to have very low DNA methylation (eg Sado et al 2000, https://doi.org/10.1006/dbio.2000.9823 and Sado et al 2004, https://doi.org/10.1242/dev.00995).

Referee cross-commenting

Having reviewed the comments of the other reviewers, we agree that they are very similar and we have no issues with them. We note that Reviewer 3 notes that considering nuclear protein levels is important in the context of this study and we agree that this is an important additional consideration that we did not consider in our review.

Significance

The manuscript is an interesting study on the role of DNMT3B in X inactivation and development. It will be of interest to scientists who work on these fundamental processes. In addition, given the roles of DNA methylation in gene regulation, cancer, aging and disease more generally the findings are likely to be of interest to many others.

Our expertise is in epigenetics and its regulation in disease, with a specific focus on DNA methylation and DNMTs.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Revision Plan (Response to Reviewers)

1. General Statements [optional]

Response: We are pleased the reviewers appreciate the power of this novel proteomics methodology that allowed us to uncover new depths on the complexity of the ribosome ubiquitination code in response to stress. We also appreciate that the reviewers think that this is a "very timely" study and "interesting to a broad audience" that can change the models of translation control currently adopted in the field. Characterizing complex cellular processes is critical to advance scientific knowledge and our work is the first of its kind using targeted proteomics methods to unveil the integrated complexity of ribosome ubiquitin signals in eukaryotic systems. We also appreciate the fairness of the comments received and below we offer a comprehensive revision plan substantially addressing the main points raised by the reviewers. According to the reviewers' suggestions, we will also expand our studies to two additional E3 ligases (Mag2 and Not4) known to ubiquitinate ribosomes, which will create an even more complete perspective of ubiquitin roles in translation regulation.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

        __Response: __We appreciate the evaluation of our work and that the power of our proteomics method established a good foundation for the study. We also understand the reviewer's concerns and we will detail below a plan to enhance quantification and increase systematic comparisons. The experiments presented here were conducted with biological replicates, but in several instances, we focused on presence and absence of bands, or their pattern (mono vs poly-ub) because of the semi-quantitative nature of immunoblots. We will revise the figures and present their quantification and statistical analyses. In additional, we did not intend to use these stressors interchangeably, but instead, to use select conditions to highlight the complexity the stress response. In particular, we followed up with H2O2 *versus* 4-NQO because both chemicals are considered sources of oxidative stress. Even though it is unfeasible to compare every single stress condition in every strain background, in the revised version, we will include additional controls to increase the cohesion of the narrative, and expand the comparison between MMS, H2O2, and 4-NQO, as suggested. Details below.

To strengthen the work, the following revisions are essential:

R1.1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.

        __Response: __As requested, we will display quantification and statistical analysis of the suggested and new immunoblots that will be conducted during the revision period.

---

R1.3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.

        __Response: __We will follow the reviewers' suggestion and redesign the analysis to increase consistency and prioritize data under identical conditions. To increase confidence in the mRNA data analysis, we intend to perform follow up experiments and analyze protein abundance of *ARG proteins* and *CTT1 *under different conditions. The remaining data using non-parallel comparisons will be moved to supplemental material and de-emphasized in the final version of the manuscript.

R1.4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

        __Response: __To ensure a better comparison across strains and conditions, we will re-run several experiments and focus on our main stress conditions. Specifically:

• 3D: We plan to re-run this experiment and include MMS

• 3E: We plan to perform the same panel of experiments in rad6D ,and display WT data as main figure.

• 4A-B: We plan to perform translation output (HPG incorporation) experiments with MMS as suggested

• 4C: We plan to re-run blots for p-eIF2a under MMS for improved comparison.

Reviewer #1 (Significance (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript the authors use a new target proteomics approach to quantify site-specific ubiquitin modification across the ribosome before and after oxidative stress. Then they validate their findings following in particular ubiquitination of Rps20 and Rps3 and extend their analysis to different forms of oxidative stress. Finally they question the relevance of two known actors of ribosome ubiquitination, Hel2 and Rad6. It is not easy to summarize the observations because in fact the major finding is that the patterns of ribosome ubiquitination occur in a stresser and enyzme specific manner (even when considering only oxidative stress). However, the complexity revealed by this study is very relevant for the field, because it underlies that the ubiquitination code of ribosomes is not easy to interpret with regard to translation dynamics and responses to stress or players involved. It suggests that some of the models that have generally been adopted probably need to be amended or completed. I am not a proteomics expert, so I cannot comment on the validity of the new proteomics approach, of whether the methods are appropriately described to reproduce the experiments. However, for the follow up experiments, the results following Rps20 and Rps3 ubiquitination are well performed, nicely controlled and are appropriately interpreted.

Maybe what one can regret is that the authors have limited their analysis to the study of Hel2 and Rad6, and not included other enyzmes that have already been associated with regulation of ribosome ubiquitination, to get a more complete picture. It may not take that much time to test more mutants, but of course there is the risk that rather than enable to make a working model it might make things even more complex.

        __Response: __We value the positive evaluation of our work. We also appreciate the notion that it meaningfully expands the knowledge on the complexity of the ribosome ubiquitination code, challenges the current models of translation control, and conducted well-performed, and nicely controlled experiments. To address the main concern of the reviewer, we will expand our work by studying two additional enzymes involved in ribosome ubiquitination (Mag2 and Not4) and provide a more comprehensive picture of this integrated system. Specifically, we will generate yeast strains deleted for *MAG2* and *NOT4*, and evaluate their impact in ribosome ubiquitination under our main conditions of stress. We will investigate the role of these additional E3s in translation output (HPG incorporation), and in inducing the integrated stress response via phosphorylated eIF2α and Gcn4 expression. Additional follow up experiments will be performed according to our initial results.

Reviewer #2 (Significance (Required)):

In recent years, regulation of translation elongation dynamics has emerged as a much more relevant site of control of gene expression that previously envisonned. The ribosome has emerged as a hub for control of stress responses. Therefore this study is certainly very timely and interesting for a broad audience. However, it does fall short of giving any simple picture, and maybe the only point one can question is whether it is interesting to publish a manuscript that concludes that regulation is complicated, without really being able to provide any kind of suggestive model.

My feeling is nevertheless that it will impact how scientists in the field design their experiments and what they will conclude. It will certainly also drive new experiments and approaches, and lead to investigations on how all the different players in regulation of ribosome modification talk to each other and signal to signaling pathways.

        __Response: __We appreciate the comments and the balanced view that studies like ours will still be impactful and contribute to a number of fields in multiple and meaningful ways. With the new experiments proposed here, and used of additional mutants and strains, we intend to propose and provide a more unified model that explain this complex and dynamic relationship.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

R3.1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.

        __Response: __We understand the importance of the suggested experiment. We have already requested and kindly received strains expressing these mutations, which will reduce the time required to successfully address this point. We will perform our translation and ISR assays such as the one referred by the reviewer in Figs. 4A-C and 5E, and results will determine the role of individual ribosome ubiquitination sites in translation control.

R3.2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

        __Response: __As was also requested by reviewer 1 and discussed above (point R1.1), we will conduct quantification and display statistical analyses for our immunoblots. In addition, we will re-run the aforementioned experiments to improve quantification following the reviewers' request (same gel & diluted control samples).

Reviewer #3 (Significance (Required)):

• General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

• Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

• Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

• The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

__ Response:__ We appreciate that our work will be valuable to a number of fields in protein dynamics and that our method advances the field by measuring ribosome ubiquitination relatively and stoichiometrically in response to stress.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Response: All requested changes require experiments and data analyses, and a complete revision plan is delineated above in section #2.

• *

4. Description of analyses that authors prefer not to carry out

• *

R1.2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.

        __Response: __Although we understand the interest on the proposed result for consistency, this is the only requested experiment that we do not intend to conduct. Because of the lack of overall ubiquitination of ribosomal proteins in *rad6**D* in response to H2O2 (e.g., Silva et al., 2015, Simoes et al., 2022), we believe that this PRM experiment in unlikely to produce meaningful insight on the ubiquitination code. In this context, we expected that sites regulated by Hel2 will be the ones largely modified in rad6*D *and we followed up on them via immunoblot. Moreover, this experiment would not be time or cost-effective, and resources and efforts could be used to strengthen other important areas of the manuscript, such as including the E3's Mag2 and Not4 into our work.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.
2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

Significance

General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript the authors use a new target proteomics approach to quantify site-specific ubiquitin modification across the ribosome before and after oxidative stress. Then they validate their findings following in particular ubiquitination of Rps20 and Rps3 and extend their analysis to different forms of oxidative stress. Finally they question the relevance of two known actors of ribosome ubiquitination, Hel2 and Rad6.

It is not easy to summarize the observations because in fact the major finding is that the patterns of ribosome ubiquitination occur in a stresser and enyzme specific manner (even when considering only oxidative stress). However, the complexity revealed by this study is very relevant for the field, because it underlies that the ubiquitination code of ribosomes is not easy to interpret with regard to translation dynamics and responses to stress or players involved. It suggests that some of the models that have generally been adopted probably need to be amended or completed. I am not a proteomics expert, so I cannot comment on the validity of the new proteomics approach, of whether the methods are appropriately described to reproduce the experiments. However, for the follow up experiments, the results following Rps20 and Rps3 ubiquitination are well performed, nicely controlled and are appropriately interpreted. Maybe what one can regret is that the authors have limited their analysis to the study of Hel2 and Rad6, and not included other enyzmes that have already been associated with regulation of ribosome ubiquitination, to get a more complete picture. It may not take that much time to test more mutants, but of course there is the risk that rather than enable to make a working model it might make things even more complex.

Significance

In recent years, regulation of translation elongation dynamics has emerged as a much more relevant site of control of gene expression that previously envisonned. The ribosome has emerged as a hub for control of stress responses. Therefore this study is certainly very timely and interesting for a broad audience.

However, it does fall short of giving any simple picture, and maybe the only point one can question is whether it is interesting to publish a manuscript that concludes that regulation is complicated, without really being able to provide any kind of suggestive model.

My feeling is nevertheless that it will impact how scientists in the field design their experiments and what they will conclude. It will certainly also drive new experiments and approaches, and lead to investigations on how all the different players in regulation of ribosome modification talk to each other and signal to signaling pathways.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

To strengthen the work, the following revisions are essential:

1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.
2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.
3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.
4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

Significance

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The response appears in a PDF document, which will be easier to read than plain text

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This article investigates the phenomenon of intracellular protein agglomeration. The authors distinguish between agglomeration and aggregation, both physically characterising them and developing a simple but elegant assay to differentiate the two. Using microscopy and structural analysis, this research demonstrates that unlike aggregates, agglomerates retain their folded structures (and are not misfolded), and do not colocalise with chaperones or interact with the proteostasis machinery which targets and breaks down misfolded proteins. The inert nature of agglomerates was further confirmed in fitness assays, though they were observed to disrupt the yeast proteome. Overall, agglomerated proteins were described and characterised, and shown to be largely neutral in vivo.

The claims and conclusions were well supported by the data. Microscopy and CD spectra (previously published) were used to confirm the nature of agglomerates and to rule out colocalistion with proteostasis machinery. This was confirmed by testing ubiquitination.

The fitness of yeast cells carrying enzymatically-inactive agglomerates was assayed by generating growth curves over 24 hours. The growth rate and doubling time were taken from these growth curves as a proxy for relative fitness. The authors mention not wanting to mask differences in lag, log or stationary phases between mutants. This could be achieved by using the area under each growth curve, rather than growth rate or doubling time alone. No further experimentation would be needed, and area under the curve may provide a more holistic metric to measure fitness by.

The results indicate that agglomerates confer a slight fitness advantage. The authors do not speculate on a reason for this. I would be interested to know why they thought this might be.

Referees cross-commenting

I have read the reports from the other reviewers and agree with their comments.

Significance

Protein filamentation is observed across the tree of life, and contributes greatly to cell structure and organisation (Wagstaff, J., Löwe, J. Prokaryotic cytoskeletons: protein filaments organizing small cells. Nat Rev Microbiol 16, 187-201 (2018).). Recent work in this field has shown that self-assembly is also important for enzyme function (S. Lim, G. A. Jung, D. J. Glover, D. S. Clark, Enhanced Enzyme Activity through Scaffolding on Customizable Self-Assembling Protein Filaments. Small 2019, 15, 1805558.). Previous work from several of these authors demonstrated that the ability of a protein to filament is subject to selection (Garcia-Seisdedos H, Empereur-Mot C, Elad N, Levy ED. Proteins evolve on the edge of supramolecular self-assembly. Nature. 2017 Aug 10;548(7666):244-247. doi: 10.1038/nature23320. Epub 2017 Aug 2. PMID: 28783726.). It has become increasingly clear that protein assemblies are ubiquitous, evolvable and perhaps overlooked in research.

This research explores a specific type of filamentation, named agglomeration, unique in that the protein which assemble are not misfolded (Romero-Romero ML, Garcia-Seisdedos H. Agglomeration: when folded proteins clump together. Biophys Rev. 2023;15: 1987-2003.). This is particularly of biomedical interest due to its role in disease, such as sickle cell anaemia (J. Hofrichter, P.D. Ross, & W.A. Eaton, Kinetics and Mechanism of Deoxyhemoglobin S Gelation: A New Approach to Understanding Sickle Cell Disease*, Proc. Natl. Acad. Sci. U.S.A. 71 (12) 4864-4868, https://doi.org/10.1073/pnas.71.12.4864 (1974).) The current research adds to the field by specifically exploring agglomerates in the most detailed methodology to date.

The novelty of this research lies especially in two areas; (1) establishing a method for distinguishing between aggregation and agglomeration, and (2) the finding that agglomerates are largely innocuous in vivo. The method established for defining agglomerates is simple, elegant and well-described in this paper's methods. The authors then probe cellular responses to agglomeration via both proteostasis machinery and cellular fitness. They noted no disruption to fitness and observed little targeting of agglomerates by chaperones. The experiments were thorough, conclusive, and resulted in interesting findings.

The inertia of this type of protein filament is unexpected; agglomerates are large and have been associated with disease. The results of this study, however, indicate that agglomerates are non-toxic and well-tolerated in vivo. The authors speculate that agglomerates may have evolved in a non-adaptive process, which is evolutionary very interesting. They also posit that these results could lead to synthetic biology applications such as a tracking expression or as a molecular sensor. This work is of great interest and impact both in cell biology, biomedicine and in-vivo biology.

Personal note: I come from a background of enzyme evolution and have viewed the work in this light.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This is a very interesting paper investigating the fitness and cellular effects of mutations that drive dihedral protein complex into forming filaments. The Levy group have previously shown that this can happen relatively easily in such complexes and this paper now investigates the cellular consequences of this phenomenon. The study is very rigorous biophysically and very surprisingly comes up empty in terms of an effect: apparently this kind of self-assembly can easily be tolerated in yeast, which was certainly not my expectation. This is a very interesting result, because it implies that such assemblies may evolve neutrally because they fulfill the two key requirements for such a trajectory: They are genetically easily accessible (in as little as a single mutation), and they have perhaps no detrimental effect on fitness. This immediately poses two very interesting questions: Are some natural proteins that are known to form filaments in the cell perhaps examples of such neutral trajectories? And if this trait is truly neutral (as long as it doesn't affect the base biochemical function of the protein in question), why don't we observe more proteins form these kinds of ordered assemblies.

I have no major comments about the experiments as I find that in general very carefully carried out. I have two more general comments:

1. The fitness effect of these assemblies, if one exists, seems very small. I think it's worth remembering that even very small fitness effects beyond even what competition experiments can reveal could in principle be enough to keep assembly-inducing alleles at very low frequencies in natural populations. Perhaps this could be acknowledged in the paper somewhere.
2. The proteins used in this study I think were chosen such that they do not have an important function in yeast that could be disrupted by assembly This allows the effect of the large scale assemblies to be measured in isolation. If I deduced this correctly, this should probably be pointed out agin in this paper (I apologise if I missed this).
3. The model system in which these effects were tested for is yeast. This organism has a rigid cell wall and I was wondering if this makes it more tolerant to large scale assemblages than wall-less eukaryotes. Could the authors comment on this?

Minor points:

In Figure 2D, what are the fits? And is there any analysis that rules out expression effects on the mutant caused by higher levels of the wild-type? The error bars in Figure 2E are not defined.

Significance

This is a remarkably rigours paper that investigates whether self-assembly into large structures has any fitness effect on a single celled organism. This is very relevant, because a landmark paper from the Levy group showed that many proteins are very close in genetic terms to forming such assemblies. The general expectation I think would have been that this phenomenon is pretty harmful. This would have explained why such filaments are relatively rare as far as we know. This paper now does a large number of highly rigours experiments to first prove beyond doubt that a range of model proteins really can be coaxed into forming such filaments in yeast cells through a very small number of mutations. Its perhaps most surprising result is that this does not negatively affect yeast cells.

From an evolutionary perspective, this is a very interesting and highly surprising result. It forces us to rethink why such filaments are not more common in Nature. Two possible answers come to mind: First, it's possible that filamentation is not directly harmful to the cell, but that assembling proteins into filaments can interfere with their basic biochemical function (which was not tested for here).

Second, perhaps assembly does cause a fitness defect, but one so small that it is hard to measure experimentally. Natural selection is very powerful, and even fitness coefficients we struggle to measure in the laboratory can have significant effects in the wild. If this is true, we might expect such filaments to be more common in organisms with small effective population sizes, in which selection is less effective.

A third possibility is of course that the prevalence of such self-assembly is under-appreciated. Perhaps more proteins than we currently know assemble into these structures under some conditions without any benefit or detriment to the organism.

These are all fascinating implications of this work that straddle the fields of evolutionary genetics and biochemistry and are therefore relevant to a very wide audience. My own expertise is in these two fields. I also think that this work will be exciting for synthetic biologists, because it proves that these kinds of assemblies are well tolerated inside cells.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this work, the authors used yeast cell as a model system to study the abovementioned question. They established a model protein system using fluorescently labeled proteins that can form both agglomerates and aggregates. Using imaging experiments, they arguably showed that agglomerates do not colocalize with the proteostasis machinery, echoing what was observed by proteomics results. The proteomics results after pull down assay to study the interactome revealed that agglomerate-size-dependent changes were dependent on the cell-wall and plasma-membrane proteins. On the other hand, as expected, the misfolded proteins (aggregates) showed heavy involvement of proteostasis network components.

Although the experiments still lack some controls and failed to support some of the conclusions, I found this work is a nice complement of the field to emphasize the point that "aggregates" and "agglomerates" are two different states, which is often mistaken by lots of researchers in recent years, in particular with the membraneless organelles (LLPS). I support its publication after the authors may consider the following suggestions and make necessary improvement.

Major concerns:

My major concern was raised by the lack of evidence to support the model system's folding state in the cell. 1. In Figure 1 and 2, I found the evidence to distinguish the folded state of proteins in the cells was limited. The concept of using hybrid imaging technique to prove the folding state is not a common experiment. The description of Figure 2 was very limited. I am sure the general audience can be convinced that the model proteins were actually folded and form agglomeration. 2. In addition, for mutants formed aggregates, the authors may consider to perform fractionation or crosslinking or native page experiment to show the evidence of protein misfolding and aggregation. 3. Have the authors considered to use FRAP assay to distinguish "aggregates" and "agglomerates" states in the cell? Does each of the state display different dynamics in the cell?

Minor concerns:

1. In Figure 3, it is very interesting to see such patten. I wonder why some of the chaperones were not responsive to misfolded proteins but some were very addicted to proteostasis. Could you elaborate more on this point? Are they chaperone sensitive, namely selective to 60/10, 70/40 or 90 system?
2. In Figure 6, I suggest to add GO analysis and KEGG analysis to distinguish pathways and functional mismatch between "aggregates" and "agglomerates" interactomics.
3. The quantity control for the proteomics studies is needed, namely the reproducibility of 3 repeats?
4. This may beyond the scope of this work. I am interested whether the authors could point out whether similar works can be done in mammalian cells. What is the model system for mammalian cell that can form "agglomerates".

Referees cross-commenting

I read through the other two reviewers' comments, which I found reasonable. It seems like all reviewers agreed that this work is of enough significance for the field only with several technical concerns.

Significance

The submitted manuscript emphasized on a very important but often misleading concept: "aggregates" and "agglomerates" are two different states of protein structures in the cell with distinct physiological roles. However, these two states are of very similar phenotype: punctate structure in the cell. While the proteostasis network has been well-established for its central role of protein quality control and coping with misfolded and aggregated proteome, the authors attempted to profile the mechanism and physiological impact of mutation-induced folded-state protein filamentation, namely a model of "agglomerates". Such overarching goal of this work clearly pointed out the novelty of this work. Clearly, this is a new angle and aspect remained to be clarified for the field.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02465

Corresponding author(s): Saravanan, Palani

1. General Statements

We would like to thank the Review Commons Team for handling our manuscript and the Reviewers for their constructive feedback and suggestions. In our revised manuscript, we have addressed and incorporated all the major suggestions of the reviewers, and we have also added new significant data on the role of Tropomyosin in regulation of endocytosis through its control over actin monomer pool maintenance and actin network homeostasis. We believe that with all these additions, our study has significantly gained in quality, strength of conclusions made, and scope for future work.

2. Point-by-point description of the revisions

Reviewer #1

Evidence, reproducibility and clarity

There are 2 Major issues -

Having an -ala-ser- linker between the GFP and tropomyosin mimics acetylation. This is not the case, and more likely the this linker acts as a spacer that allows tropomyosin polymers to form on the actin, and without it there is steric hindrance. A similar result would be seen with a simple flexible uncharged linker. It has been shown in a number of labs that the GFP itself masks the effect of the charge on the amino terminal methionine. This is consistent with NMR, crystallographic and cryo structural studies. Biochemical studies should be presented to demonstrate that the impact of a linker for the conclusions stated to be made, which provide the basis of a major part of this study.

Response: We would like to clarify that all mNG-Tpm constructs used in our study contain a 40 amino-acid (aa) flexible linker between the N-terminal mNG fluorescent protein and the Tpm protein as per our earlier published study (Hatano et al., 2022). During initial optimization, we have also experimented with linker length and the 40aa-linker length works optimally for clear visualization of Tpm onto actin cable structures in budding yeast, fission yeast (both S. pombe and S. japonicus), and mammalian cells (Hatano et al., 2022). These constructs have also been used since in other studies (Wirshing et al., 2023; Wirshing and Goode, 2024) and currently represents the best possible strategy to visualize Tpm isoforms in live cells. In our study, we characterized these proteins for functionality and found that both mNG-Tpm1 and mNG-Tpm2 were functional and can rescue the synthetic lethality observed in Dtpm1Dtpm2 cells. During our study, we observed that mNG-Tpm1 expression from a single-copy integration vector did not restore full length actin cables in Dtpm1 cells (Fig. 1B, 1C). We hypothesized that this could be a result of reduced binding affinity of the tagged tropomyosin due to lack of normal N-terminal acetylation which stabilizes the N-terminus. The 40aa linker is unstructured and may not be able to neutralize the charge on the N-terminal Methionine, thus, we tried to insert -Ala-Ser- dipeptide which has been routinely used in vitro biochemical studies to stabilize the N-terminal helix and impart a similar effect as the N-terminal acetylation (Alioto et al., 2016; Palani et al., 2019; Christensen et al., 2017) by restoring normal binding affinity of Tpm to F-actin (Monteiro et al., 1994; Greenfield et al., 1994). We observed that addition of the -Ala-Ser- dipeptide to mNG-Tpm fusion, indeed, restored full length actin cables when expressed in Dtpm1 cells, performing significantly better in our in vivo experiments (Fig. 1B, 1C). We agree with the reviewer that the -AS- dipeptide addition may not mimic N-terminal acetylation structurally but as per previous studies, it may stabilize the N-terminus of Tpm and allow normal head-to-tail dimer formation (Greenfield et al., 1994; Monteiro et al., 1994; Frye et al., 2010). We have discussed this in our new Discussion section (Lines 350-372). Since, the addition of -AS- dipeptide was referred to as "acetyl-mimic (am)" in a previous study (Alioto et al., 2016), we continued to use the same nomenclature in our study. Now as per your suggestions and to be more accurate, we have renamed "mNG-amTpm" constructs as "mNG-ASTpm" throughout the study to not confuse or claim that -AS- addition mimics acetylation. In any case, we have not seen any other ill effect of -AS- dipeptide introduction in addition to our 40 amino acid linker suggesting that it can also be considered part of the linker. Although, we agree with the reviewer that biochemical characterization of the effect of linker would be important to determine, we strongly believe that it is currently outside the scope of this study and should be taken up for future work with these proteins. Our study has majorly aimed to understand the functionality and utility of these mNG-Tpm fusion proteins for cell biological experiments in vivo, which was not done earlier in any other model system.

My major issue however is making the conclusions stated here, using an amino-terminal fluorescent protein tag that s likely to impact any type of isoform selection at the end of the actin polymer. Carboxyl terminal tagging may have a reduced effect, but modifying the ends of the tropomyosin, which are integral in stabilising end to end interactions with itself on the actin filament, never mind any section systems that may/maynot be present in the cell, is not appropriate.

Response: We agree with the reviewer that N-terminal tagging of tropomyosin may have effects on its function, but these constructs represent the only fluorescently tagged functional tropomyosin constructs available currently while C-terminal fusions are either non-functional (we were unable to construct strains with endogenous Tpm1 gene fused C-terminally to GFP) or do not localize clearly to actin structures (See Figure R1 showing endogenous C-terminally tagged Tpm2-yeGFP that shows almost no localization to actin cables). To our knowledge, our study represents a first effort to understand the question of spatial sorting of Tpm isoforms, Tpm1 and Tpm2, in S. cerevisiae and any future developments with better visualization strategies for Tpm isoforms without compromising native N-terminal modifications and function will help improve our understanding of these proteins in vivo. We have also discussed these possibilities in our new Discussion section (Lines 391-396).

Significance

This paper explores the role of formin in determining the localisation of different tropomyosins to different actin polymers and cellular locations within budding yeast. Previous studies have indicated a role for the actin nucleating proteins in recruiting different forms of tropomyosin within fission yeast. In mammalian cells there is variation in the role of formins in affiecting tropomyosin localisation - variation between cell type. There is also evidence that other actin binding proteins, and tropomyosin abundance play roles in regulating the tropomyosin-actin association according to cell type. Biochemical studies have previously been undertaken using budding yeast and fission yeast that the core actin polymerisation domain of formins do not interact with tropomyosin directly. The significance of this study, given the above, and the concerns raised is not clear to this reviewer.

Response: __Our study explores multiple facets of Tropomyosin (Tpm) biology. The lack of functional tagged Tpm has been a major bottleneck in understanding Tpm isoform diversity and function across eukaryotes. In our study, we characterize the first functional tagged Tpm proteins (Fig. 1, Fig. S1) and use them to answer long-standing questions about localization and spatial sorting of Tpm isoforms in the model organism S. cerevisiae (Fig. 2, Fig. 3, Fig. S2, Fig. S3). We also discover that the dual Tpm isoforms, Tpm1 and Tpm2, are functionally redundant for actin cable organization and function, while having gained divergent functions in Retrograde Actin Cable Flow (RACF) (Fig. 4, Fig. 5A-D, Fig. S4, Fig. S5, Fig. S6). We have now added new data on role of global Tpm levels controlling endocytosis via maintenance of normal linear-to-branched actin network homeostasis in S. cerevisiae (Fig. 5E-G)__. We respectfully differ with the reviewer on their assessment of our study and request the reviewer to read our revised manuscript which discusses the significance, limitations, and future perspectives of our study in detail.

Reviewer #2

Evidence, reproducibility and clarity

This manuscript by Dhar, Bagyashree, Palani and colleagues examines the function of the two tropomyosins, Tpm1 and Tpm2, in the budding yeast S. cerevisiae. Previous work had shown that deletion of tpm1 and tpm2 causes synthetic lethality, indicating overlapping function, but also proposed that the two tropomyosins have distinct functions, based on the observation that strong overexpression of Tpm2 causes defects in bud placement and fails to rescue tpm1∆ phenotypes (Drees et al, JCB 1995). The manuscript first describes very functional mNeonGreen tagged version of Tpm1 and Tpm2, where an alanine-serine dipeptide is inserted before the first methionine to mimic acetylation. It then proposes that the Tpm1 and Tpm2 exhibit indistinguishable localization and that low level overexpression (?) of Tpm2 can replace Tpm1 for stabilization of actin cables and cell polarization, suggesting almost completely redundant functions. They also propose on specific function of Tpm2 in regulating retrograde actin cable flow.

Overall, the data are very clean, well presented and quantified, but in several places are not fully convincing of the claims. Because the claims that Tpm1 and Tpm2 have largely overlapping function and localization are in contradiction to previous publication in S. cerevisiae and also different from data published in other organisms, it is important to consolidate them. There are fairly simple experiments that should be done to consolidate the claims of indistinguishable localization, and levels of expression, for which the authors have excellent reagents at their disposal.

1. Functionality of the acetyl-mimic tagged tropomyosin constructs: The overall very good functionality of the tagged Tpm constructs is convincing, but the authors should be more accurate in their description, as their data show that they are not perfectly functional. For instance, the use of "completely functional" in the discussion is excessive. In the results, the statement that mNG-Tpm1 expression restores normal growth (page 3, line 69) is inaccurate. Fig S1C shows that tpm1∆ cells expressing mNG-Tpm1 grow more slowly than WT cells. (The next part of the same sentence, stating it only partially restores length of actin cables should cite only Fig S1E, not S1F.) Similarly, the growth curve in Fig S1C suggests that mNG-amTpm1, while better than mNG-Tpm1 does not fully restore the growth defect observed in tpm1∆ (in contrast to what is stated on p. 4 line 81). A more stringent test of functionality would be to probe whether mNG-amTpm1 can rescue the synthetic lethality of the tpm1∆ tpm2∆ double mutant, which would also allow to test the functionality of mNG-amTpm2.

__Response: __We would like to thank the reviewer for his feedback and suggestions. Based on the suggestions, we have now more accurately described the growth rescue observed by expression of mNG-ASTpm1 in Dtpm1 cells in the revised text. We have also removed the use of "completely functional" to describe mNG-Tpm functionality and corrected any errors in Figure citations in the revised manuscript.

As per reviewers' suggestion, we have now tested rescue of synthetic lethality of Dtpm1Dtpm2 cells by expression of all mNG-Tpm variants and we find that all of them are capable of restoring the viability of Dtpm1Dtpm2 cells when expressed under their native promoters via a high-copy plasmid (pRS425) (Fig. S1E) but only mNG-Tpm1 and mNG-ASTpm1 restored viability of Dtpm1Dtpm2 cells when expressed under their native promoters via an integration plasmid (pRS305) (Fig. S1F). These results clearly suggest that while both mNG-Tpm1 and mNG-Tpm2 constructs are functional, Tpm1 tolerates the presence of the N-terminal fluorescent tag better than Tpm2. These observations now enhance our understanding of the functionality of these mNG-Tpm fusion proteins and will be a useful resource for their usage and experimental design in future studies in vivo.

It would also be nice to comment on whether the mNG-amTpm constructs really mimicking acetylation. Given the Ala-Ser peptide ahead of the starting Met is linked N-terminally to mNG, it is not immediately clear it will have the same effect as a free acetyl group decorating the N-terminal Met.

Response: __We agree with the reviewer's observation and for the sake of clarity and accuracy, we have now renamed "mNG-amTpm" with "mNG-ASTpm". The use of -AS- dipeptide is very routine in studies with Tpm (Alioto et al., 2016; Palani et al., 2019; Christensen et al., 2017) and its addition restores normal binding affinities to Tpm proteins purified from E. coli (Monteiro et al., 1994). We agree with the reviewer that the -AS- dipeptide addition may not mimic N-terminal acetylation structurally but as per previous studies, it may help neutralize the impact of a freely protonated Met on the alpha-helical structure and stabilize the N-terminus helix of Tpm and allow normal head-to-tail dimer formation (Monteiro et al., 1994; Frye et al., 2010; Greenfield et al., 1994). Consistent with this, we also observe a highly significant improvement in actin cable length when expressing mNG-ASTpm as compared to mNG-Tpm in Dtpm1 cells, suggesting an improvement in function probably due to increased binding affinity (Fig. 1B, 1C). We have also discussed this in our answer to Question 1 of Reviewer 1 and the revised manuscript (Lines 350-372)__.

__ Localization of Tpm1 and Tpm2:__Given the claimed full functionality of mNG-amTpm constructs and the conclusion from this section of the paper that relative local concentrations may be the major factor in determining tropomyosin localization to actin filament networks, I am concerned that the analysis of localization was done in strains expressing the mNG-amTpm construct in addition to the endogenous untagged genes. (This is not expressly stated in the manuscript, but it is my understanding from reading the strain list.) This means that there is a roughly two-fold overexpression of either tropomyosin, which may affect localization. A comparison of localization in strains where the tagged copy is the sole Tpm1 (respectively Tpm2) source would be much more conclusive. This is important as the results are making a claim in opposition to previous work and observation in other organisms.

Response: __We thank the reviewer for this observation and their suggestions. We agree that relative concentrations of functional Tpm1 and Tpm2 in cells may influence the extent of their localizations. As per the reviewer's suggestion, we have now conducted our quantitative analysis in cells lacking endogenous Tpm1 and only expressing mNG-ASTpm1 from an integrated plasmid copy at the leu2 locus and the data is presented in new __Figure S3. We compared Tpm-bound cable length (Fig. S3A, S3B) __and Tpm-bound cable number (Fig. S3A, S3C) along with actin cable length (Fig. S3D, S3E) and actin cable number (Fig. S3D, S3F) in wildtype, Dbnr1, and Dbni1 cells. Our analysis revealed that mNG-ASTpm1 localized to actin cable structures in wildtype, Dbnr1, and Dbni1 cells and the decrease observed in Tpm-bound cable length and number upon loss of either Bnr1 or Bni1, was accompanied by a corresponding decrease in actin cable length and number upon loss of either Bnr1 or Bni1. Thus, this analysis reached the same conclusion as our earlier analysis (Fig. 2) that mNG-ASTpm1 does not show preference between Bnr1 and Bni1-made actin cables. mNG-ASTpm2 did not restore functionality, when expressed as single integrated copy, in Dtpm1Dtpm2 cells (new results in __Fig. S1E, S1F, S5A) thus, we could not conduct a similar analysis for mNG-ASTpm2. This suggests that use of mNG-ASTpm2 would be more meaningful in the presence of endogenous Tpm2 as previously done in Fig. 2D-F.

We have now also performed additional yeast mating experiments with cells lacking bnr1 gene and expressing either mNG-ASTpm1 or mNG-ASTpm2 and the data is shown in new Figure 3. From these observations, we observe that both mNG-ASTpm1 and mNG-ASTpm2 localize to the mating fusion focus in a Bnr1-independent manner (Fig. 3B, 3D) and suggests that they bind to Bni1-made actin cables that are involved in polarized growth of the mating projection. These results also add strength to our conclusion that Tpm1 and Tpm2 localize to actin cables irrespective of which formin nucleates them. Overall, these new results highlight and reiterate our model of formin-isoform independent binding of Tpm1 and Tpm2 in S. cerevisiae.

In fact, although the authors conclude that the tropomyosins do not exhibit preference for certain actin structures, in the images shown in Fig 2A and 2D, there seems to be a clear bias for Tpm1 to decorate cables preferentially in the bud, while Tpm2 appears to decorate them more in the mother cell. Is that a bias of these chosen images, or does this reflect a more general trend? A quantification of relative fluorescence levels in bud/mother may be indicative.

Response: __We thank the reviewer for pointing this out. Our data and analysis do not suggest that Tpm1 and Tpm2 show any preference for decoration of cables in either mother or bud compartment. As per the reviewer's suggestion, we have now quantified the ratio of mean mNG fluorescence in the bud to the mother (Bud/Mother) and the data is shown in __Figure. S2G. The bud-to-mother ratio was similar for mNG-ASTpm1 and mNG-ASTpm2 in wildtype cells, and the ratio increased in Dbnr1 cells and decreased in Dbni1 cells for both mNG-ASTpm1 and mNG-ASTpm2 (Fig. S2G). __This is consistent with the decreased actin cable signal in the mother compartment in Dbnr1 cells and decreased actin cable signal in the bud compartment in Dbni1 cells (Fig. S2A-D). Thus, our new analysis shows that both mNG-ASTpm1 and mNG-ASTpm2 have similar changes in their concentration (mean fluorescence) upon loss of either formins Bnr1 and Bni1 and show similar ratios in wildtype cells as well, suggesting no preference for binding to actin cables in either bud or mother compartment. The preference inferred by the reviewer seems to be a bias of the current representative images and thus, we have replaced the images in __Fig. 2A, 2D to more accurately represent the population.

The difficulty in preserving mNG-amTpm after fixation means that authors could not quantify relative Tpm/actin cable directly in single fixed cells. Did they try to label actin cables with Lifeact instead of using phalloidin, and thus perform the analysis in live cells?

__Response: __We did not use LifeAct for our analysis as LifeAct is known to cause expression-dependent artefacts in cells (Courtemanche et al., 2016; Flores et al., 2019; Xu and Du, 2021) and it also competes with proteins that regulate normal cable organization like cofilin. Use of LifeAct would necessitate standardization of expression to avoid such artefacts in vivo. Also, phalloidin staining provides the best staining of actin cables and allows for better quantitative results in our experiments. The use of LifeAct along with mNG-Tpm would also require optimization with a red fluorescent protein which usually tend to have lower brightness and photostability. However, during the revision of our study, a new study from Prof. Goode's lab has developed and optimized expression of new LifeAct-3xmNeonGreen constructs for use in S. cerevisiae (Wirshing and Goode, 2024). Thus, a similar strategy of using tandem copies of bright and photostable red fluorescent proteins can be explored for use in combination with mNG-Tpm in the future studies.

__ Complementation of tpm1∆ by Tpm2:__

I am confused about the quantification of Tpm2 expression by RT-PCR shown in Fig S3F. This figure shows that tpm2 mRNA expression levels are identical in cells with an empty plasmid or with a tpm2-encoding plasmid. In both strains (which lack tpm1), as well as in the WT control, one tpm2 copy is in the genome, but only one strain has a second tpm2 copy expressed from a centromeric plasmid, yet the results of the RT-PCR are not significantly different. (If anything, the levels are lower in the tpm2 plasmid-containing strain.) The methods state that the primers were chosen in the gene, so likely do not distinguish the genomic from the plasmid allele. However, the text claims a 1-fold increase in expression, and functional experiments show a near-complete rescue of the tpm1∆ phenotype. This is surprising and confusing and should be resolved to understand whether higher levels of Tpm2 are really the cause of the observed phenotypic rescue.

The authors could for instance probe for protein levels. I believe they have specific nanobodies against tropomyosin. If not, they could use expression of functional mNG-amTpm2 to rescue tpm1∆. Here, the expression of the protein can be directly visualized.

Response: __We thank the reviewer for pointing this out. We would like to clarify that in our RT-qPCR experiments, the primers were chosen within the Tpm1 and Tpm2 gene and do not distinguish between transcripts from endogenous or plasmid copy. We have now mentioned this in the Materials and Methods section of the revised manuscript. So, they represent a relative estimate of the total mRNA of these genes present in cells. We were consistently able to detect ~19 fold increase in Tpm2 total mRNA levels as compared to wildtype and ∆tpm1 cells (Fig. S4D) when tpm2 was expressed from a high-copy plasmid (pRS425). This increase in Tpm2 mRNA levels was accompanied by a rescue in growth (Fig. S4A) and actin cable organization (Fig. S4B) of ∆tpm1 cells containing pRS425-ptpm2TPM2. When tpm2 was expressed from a low-copy number centromeric plasmid (pRS316), we detected a ~2 fold increase in Tpm2 transcript levels when using the tpm1 promoter and no significant change was detected when using tpm2 promoter (Fig. S4E)__. We have made sure that these results are accurately described in the revised manuscript.

As per the reviewer's suggestion, we have now conducted a more extensive analysis to ascertain the expression levels of Tpm2 in our experiments and the data is now presented in new Figure S5. We used mNG-ASTpm1 and mNG-ASTpm2 to rescue growth of ∆tpm1 (Fig. S5A) and correlated growth rescue with protein levels using quantified fluorescence intensity (Fig. S5B, S5C) and western blotting (anti-mNG) (Fig. S5D, S5E). We find that ∆tpm1 cells containing pRS425-ptpm1mNG-ASTpm1 had the highest protein level followed by pRS425-ptpm2 mNG-ASTpm2, pRS305-ptpm1mNG-ASTpm1, and the least protein levels were found in pRS305-ptpm2 mNG-ASTpm2 containing ∆tpm1 cells in both fluorescence intensity and western blotting quantifications (Fig. S5C, S5E). Surprisingly, we were not able to detect any protein levels in ∆tpm1 cells containing pRS305-ptpm2 mNG-ASTpm2 with western blotting (Fig. S5D) which was also accompanied by a lack of growth rescue (Fig. S5A). This most likely due to weak expression from the native Tpm2 promoter which is consistent with previous literature (Drees et al., 1995). Taken together, this data clearly shows that the rescue observed in ∆tpm1 cells is caused due to increased expression of mNG-ASTpm2 in cells and supports our conclusion that increase in Tpm2 expression leads to restoration of normal growth and actin cables in ∆tpm1 cells.

__ Specific function of Tpm2:__

The data about the retrograde actin flow is interpreted as a specific function of Tpm2, but there is no evidence that Tpm1 does not also share this function. To reach this conclusion one would have to investigate retrograde actin flow in tpm1∆ (difficult as cables are weak) or for instance test whether Tpm1 expression restores normal retrograde flow to tpm2∆ cells.

Response: __We agree with the reviewer and as per the reviewer's suggestion, we have performed another experiment which include wildtype, ∆tpm2 cells containing empty pRS316 vector or pRS316-ptpm2TPM1 or pRS316-ptpm1TPM1. We find that RACF rate increased in ∆tpm2 cells as compared to wildtype and was restored to wildtype levels by exogenous expression of Tpm2 but not Tpm1 (Fig. S6E, S6F). Since, actin cables were not detectable in ∆tpm1 cells, we measured RACF rates in ∆tpm1 cells expressing Tpm1 or Tpm2 from a plasmid copy, which restored actin cables as shown previously in __Fig. 5A-C. We observed that RACF rates were similar to wildtype in ∆tpm1 cells expressing either Tpm1 or Tpm2 (Fig. S6E, S6F), suggesting that Tpm1 is not involved in RACF regulation. Taken together, these results suggest a specific role for Tpm2, but not Tpm1, in RACF regulation in S. cerevisiae, consistent with previous literature (Huckaba et al., 2006).

Minor comments: __1.__The growth of tpm1∆ with empty plasmid in Fig S3A is strangely strong (different from other figures).

Response: We thank the reviewer for pointing this out. We have now repeated the drop test multiple times (Fig. R2), but we see similar growth rates as the drop test already presented in Fig. S4A. __At this point, it would be difficult to ascertain the basis of this difference observed at 23{degree sign}C and 30{degree sign}C, but a recent study that links leucine levels to actin cable stability (Sing et al., 2022) might explain the faster growth of these ∆tpm1 cells containing a leu2 gene carrying high-copy plasmid. However, there is no effect on growth rate at 37{degree sign}C which is consistent with other spot assays shown in __Fig. S1D, S4F, S5A.

Significance

I am a cell biologist with expertise in both yeast and actin cytoskeleton.

The question of how tropomyosin localizes to specific actin networks is still open and a current avenue of study. Studies in other organisms have shown that different tropomyosin isoforms, or their acetylated vs non-acetylated versions, localize to distinct actin structures. Proposed mechanisms include competition with other ABPs and preference imposed by the formin nucleator. The current study re-examines the function and localization of the two tropomyosin proteins from the budding yeast and reaches the conclusion that they co-decorate all formin-assembled structures and also share most functions, leading to the simple conclusion that the more important contribution of Tpm1 is simply linked to its higher expression. Once consolidated, the study will appeal to researchers working on the actin cytoskeleton.

We thank the reviewer for their positive assessment of our work and the constructive feedback that has greatly improved the quality of our study. After addressing the points raised by the reviewer, we believe that our study has significantly gained in consolidating the major conclusions of our work.

**Referees cross-commenting**

Having read the other reviewers' comments, I do agree with reviewer 1 that it is not clear whether the Ala-Ser linker really mimics acetylation. I am less convinced than reviewer 3 that the key conclusions of the study are well supported, notably the issue of Tpm2 expression levels is not convincing to me.

Response: __We acknowledge the reviewer's point about the effect of Ala-Ser dipeptide and would request the reviewer to refer to our response to Reviewer 1 (Question 1) for a more detailed discussion on this. We have also extensively addressed the question of Tpm2 expression levels as suggested by the reviewer (new data in __Figure S5) which has further strengthened the conclusions of our study.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary:__ The study presents the first fully functional fluorescently tagged Tpm proteins, enabling detailed probing of Tpm isoform localization and functions in live cells. The authors created a modified fusion protein, mNG-amTpm, which mimicked native N-terminal acetylation and restored both normal growth and full-length actin cables in yeast cells lacking native Tpm proteins, demonstrating the constructs' full functionality. They also show that Tpm1 and Tpm2 do not have a preference for actin cables nucleated by different formins (Bnr1 and Bni1). Contrary to previous reports, the study found that overexpressing Tpm2 in Δtpm1 cells could restore growth rates and actin cable formation. Furthermore, it is shown that despite its evolutionary divergence, Tpm2 retains actin-protective functions and can compensate for the loss of Tpm1, contributing to cellular robustness.

Major and Minor Comments: 1. The key conclusions of this paper are convincing. However, I suggest that more detail be provided regarding the image analysis used in this study. Specifically, since threshold settings can impact the quality of the generated data and, therefore, its interpretation, it would be useful to see a representative example of the quantification methods used for actin cable length/number (as in refs. 80 and 81) and mitochondria morphology. These could be presented as Supplemental Figures. Additionally, it would help to interpret the results if the authors could be more specific about the statistical tests that were used.

Response: __We agree with the reviewer's suggestions and have now updated our Materials and Methods section to describe the image analysis pipelines used in more detail. We have also added examples of quantification procedure for actin cable length/number and mitochondrial morphology as an additional Supplementary __Figure S7. Briefly, the following pipelines were used:

• Actin cable length and number analysis: This was done exactly as mentioned in McInally et al., 2021, McInally et al., 2022. Actin cables were manually traced in Fiji as shown in __ S7A__, and then the traces files for each cell were run through a Python script (adapted from McInally et al., 2022) that outputs mean actin cable length and number per cell.
• Mitochondria morphology: Mitochondria Analyzer plug-in in Fiji was used to segment out the mitochondrial fragments. The parameters used for 2D segmentation of mitochondria were first optimized using "2D Threshold Optimize" to find the most accurate segmentation and then the same parameters were run on all images. After segmentation of the mitochondrial network, measurements of fragment number were done using "Analyze Particles" function in Fiji. An example of the overall process is shown in __ S7B.__ As per the reviewer's suggestion, we have now included the description of the statistical test used in the Figure Legends of each Figure in the revised manuscript. We have used One-Way Anova with Tukey's Multiple Comparison test, Kruskal-Wallis test with Dunn's Multiple Comparisons, and Unpaired Two-tailed t-test using the in-built functions in GraphPad Prism (v.6.04).

**Referees cross-commenting**

I agree with both reviewers 1 and 2 regarding the issues with the Ala-Ser acetylation mimic and Tpm2 expression levels, respectively. I think the authors should be more careful in how they frame the results, but I consider that these issues do not invalidate the main conclusions of this study.

Response: __We acknowledge the reviewer's concern about the Ala-Ser dipeptide and would request them to refer our earlier discussion on this in response to Reviewer 1 (Question 1) and Reviewer 2 (Question 2). We would also request the reviewer to refer to our answer to Reviewer 2 (Question 6) where we have extensively addressed the question of Tpm2 expression levels and their effect on rescue of Dtpm1 cells. This data is now presented as new __Figure S5 in our revised manuscript.

Reviewer#3 (Significance (Required)):

The finding that Tpm2 can compensate for the loss of Tpm1, restoring actin cable organization and normal growth rates, challenges previous assumptions about the non-redundant functions of these isoforms in Saccharomyces cerevisiae (ref. 16). It also supports a concentration-dependent and formin-independent localization of Tpm isoforms to actin cables in this species. The development of fully functional fluorescently tagged Tpm proteins is a significant methodological advancement. This advancement overcomes previous visualization challenges and allows for accurate in vivo studies of Tpm function and regulation in S. cerevisiae.

The findings will be of particular interest to researchers in the field of cellular and molecular biology who study actin cytoskeleton dynamics. Additionally, it will be relevant for those utilizing advanced microscopy and live-cell imaging techniques.

As a researcher, my experience lies in cytoskeleton dynamics and protein interactions, though I do not have specific experience related to tropomyosin. I use different yeast species as models and routinely employ live-cell imaging as a tool.

We thank the reviewer for their positive outlook and assessment of our study. We have incorporated all their suggestions, and we are confident that the revised manuscript has significantly improved in quality due to these additions.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The study presents the first fully functional fluorescently tagged Tpm proteins, enabling detailed probing of Tpm isoform localization and functions in live cells. The authors created a modified fusion protein, mNG-amTpm, which mimicked native N-terminal acetylation and restored both normal growth and full-length actin cables in yeast cells lacking native Tpm proteins, demonstrating the constructs' full functionality. They also show that Tpm1 and Tpm2 do not have a preference for actin cables nucleated by different formins (Bnr1 and Bni1). Contrary to previous reports, the study found that overexpressing Tpm2 in Δtpm1 cells could restore growth rates and actin cable formation. Furthermore, it is shown that despite its evolutionary divergence, Tpm2 retains actin-protective functions and can compensate for the loss of Tpm1, contributing to cellular robustness.

Major and Minor Comments:

The key conclusions of this paper are convincing. However, I suggest that more detail be provided regarding the image analysis used in this study. Specifically, since threshold settings can impact the quality of the generated data and, therefore, its interpretation, it would be useful to see a representative example of the quantification methods used for actin cable length/number (as in refs. 80 and 81) and mitochondria morphology. These could be presented as Supplemental Figures. Additionally, it would help to interpret the results if the authors could be more specific about the statistical tests that were used.

Referees cross-commenting

I agree with both reviewers 1 and 2 regarding the issues with the Ala-Ser acetylation mimic and Tpm2 expression levels, respectively. I think the authors should be more careful in how they frame the results, but I consider that these issues do not invalidate the main conclusions of this study.

Significance

The finding that Tpm2 can compensate for the loss of Tpm1, restoring actin cable organization and normal growth rates, challenges previous assumptions about the non-redundant functions of these isoforms in Saccharomyces cerevisiae (ref. 16). It also supports a concentration-dependent and formin-independent localization of Tpm isoforms to actin cables in this species. The development of fully functional fluorescently tagged Tpm proteins is a significant methodological advancement. This advancement overcomes previous visualization challenges and allows for accurate in vivo studies of Tpm function and regulation in S. cerevisiae.

The findings will be of particular interest to researchers in the field of cellular and molecular biology who study actin cytoskeleton dynamics. Additionally, it will be relevant for those utilizing advanced microscopy and live-cell imaging techniques.

As a researcher, my experience lies in cytoskeleton dynamics and protein interactions, though I do not have specific experience related to tropomyosin. I use different yeast species as models and routinely employ live-cell imaging as a tool.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This manuscript by Dhar, Bagyashree, Palani and colleagues examines the function of the two tropomyosins, Tpm1 and Tpm2, in the budding yeast S. cerevisiae. Previous work had shown that deletion of tpm1 and tpm2 causes synthetic lethality, indicating overlapping function, but also proposed that the two tropomyosins have distinct functions, based on the observation that strong overexpression of Tpm2 causes defects in bud placement and fails to rescue tpm1∆ phenotypes (Drees et al, JCB 1995). The manuscript first describes very functional mNeonGreen tagged version of Tpm1 and Tpm2, where an alanine-serine dipeptide is inserted before the first methionine to mimic acetylation. It then proposes that the Tpm1 and Tpm2 exhibit indistinguishable localization and that low level overexpression (?) of Tpm2 can replace Tpm1 for stabilization of actin cables and cell polarization, suggesting almost completely redundant functions. They also propose on specific function of Tpm2 in regulating retrograde actin cable flow.

Overall, the data are very clean, well presented and quantified, but in several places are not fully convincing of the claims. Because the claims that Tpm1 and Tpm2 have largely overlapping function and localization are in contradiction to previous publication in S. cerevisiae and also different from data published in other organisms, it is important to consolidate them. There are fairly simple experiments that should be done to consolidate the claims of indistinguishable localization, and levels of expression, for which the authors have excellent reagents at their disposal.

Functionality of the acetyl-mimic tagged tropomyosin constructs:

The overall very good functionality of the tagged Tpm constructs is convincing, but the authors should be more accurate in their description, as their data show that they are not perfectly functional. For instance, the use of "completely functional" in the discussion is excessive. In the results, the statement that mNG-Tpm1 expression restores normal growth (page 3, line 69) is inaccurate. Fig S1C shows that tpm1∆ cells expressing mNG-Tpm1 grow more slowly than WT cells. (The next part of the same sentence, stating it only partially restores length of actin cables should cite only Fig S1E, not S1F.) Similarly, the growth curve in Fig S1C suggests that mNG-amTpm1, while better than mNG-Tpm1 does not fully restore the growth defect observed in tpm1∆ (in contrast to what is stated on p. 4 line 81). A more stringent test of functionality would be to probe whether mNG-amTpm1 can rescue the synthetic lethality of the tpm1∆ tpm2∆ double mutant, which would also allow to test the functionality of mNG-amTpm2.

It would also be nice to comment on whether the mNG-amTpm constructs really mimicking acetylation. Given the Ala-Ser peptide ahead of the starting Met is linked N-terminally to mNG, it is not immediately clear it will have the same effect as a free acetyl group decorating the N-terminal Met.

Localization of Tpm1 and Tpm2:

Given the claimed full functionality of mNG-amTpm constructs and the conclusion from this section of the paper that relative local concentrations may be the major factor in determining tropomyosin localization to actin filament networks, I am concerned that the analysis of localization was done in strains expressing the mNG-amTpm construct in addition to the endogenous untagged genes. (This is not expressly stated in the manuscript, but it is my understanding from reading the strain list.) This means that there is a roughly two-fold overexpression of either tropomyosin, which may affect localization. A comparison of localization in strains where the tagged copy is the sole Tpm1 (respectively Tpm2) source would be much more conclusive. This is important as the results are making a claim in opposition to previous work and observation in other organisms.

In fact, although the authors conclude that the tropomyosins do not exhibit preference for certain actin structures, in the images shown in Fig 2A and 2D, there seems to be a clear bias for Tpm1 to decorate cables preferentially in the bud, while Tpm2 appears to decorate them more in the mother cell. Is that a bias of these chosen images, or does this reflect a more general trend? A quantification of relative fluorescence levels in bud/mother may be indicative.

The difficulty in preserving mNG-amTpm after fixation means that authors could not quantify relative Tpm/actin cable directly in single fixed cells. Did they try to label actin cables with Lifeact instead of using phalloidin, and thus perform the analysis in live cells?

Complementation of tpm1∆ by Tpm2:

I am confused about the quantification of Tpm2 expression by RT-PCR shown in Fig S3F. This figure shows that tpm2 mRNA expression levels are identical in cells with an empty plasmid or with a tpm2-encoding plasmid. In both strains (which lack tpm1), as well as in the WT control, one tpm2 copy is in the genome, but only one strain has a second tpm2 copy expressed from a centromeric plasmid, yet the results of the RT-PCR are not significantly different. (If anything, the levels are lower in the tpm2 plasmid-containing strain.) The methods state that the primers were chosen in the gene, so likely do not distinguish the genomic from the plasmid allele. However, the text claims a 1-fold increase in expression, and functional experiments show a near-complete rescue of the tpm1∆ phenotype. This is surprising and confusing and should be resolved to understand whether higher levels of Tpm2 are really the cause of the observed phenotypic rescue. The authors could for instance probe for protein levels. I believe they have specific nanobodies against tropomyosin. If not, they could use expression of functional mNG-amTpm2 to rescue tpm1∆. Here, the expression of the protein can be directly visualized.

Specific function of Tpm2:

The data about the retrograde actin flow is interpreted as a specific function of Tpm2, but there is no evidence that Tpm1 does not also share this function. To reach this conclusion one would have to investigate retrograde actin flow in tpm1∆ (difficult as cables are weak) or for instance test whether Tpm1 expression restores normal retrograde flow to tpm2∆ cells.

Minor comments:

The growth of tpm1∆ with empty plasmid in Fig S3A is strangely strong (different from other figures).

Referees cross-commenting

Having read the other reviewers' comments, I do agree with reviewer 1 that it is not clear whether the Ala-Ser linker really mimics acetylation. I am less convinced than reviewer 3 that the key conclusions of the study are well supported, notably the issue of Tpm2 expression levels is not convincing to me.

Significance

I am a cell biologist with expertise in both yeast and actin cytoskeleton.

The question of how tropomyosin localizes to specific actin networks is still open and a current avenue of study. Studies in other organisms have shown that different tropomyosin isoforms, or their acetylated vs non-acetylated versions, localize to distinct actin structures. Proposed mechanisms include competition with other ABPs and preference imposed by the formin nucleator. The current study re-examines the function and localization of the two tropomyosin proteins from the budding yeast and reaches the conclusion that they co-decorate all formin-assembled structures and also share most functions, leading to the simple conclusion that the more important contribution of Tpm1 is simply linked to its higher expression. Once consolidated, the study will appeal to researchers working on the actin cytoskeleton.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

There are 2 Major issues:

1. Having an -ala-ser- linker between the GFP and tropomyosin mimics acetylation. This is not the case, and more likely the this linker acts as a spacer that allows tropomyosin polymers to form on the actin, and without it there is steric hindrance. A similar result would be seen with a simple flexible uncharged linker. It has been shown in a number of labs that the GFP itself masks the effect of the charge on the amino terminal methionine. This is consistent with NMR, crystallographic and cryo structural studies. Biochemical studies should be presented to demonstrate that the impact of a linker for the conclusions stated to be made, which provide the basis of a major part of this study.
2. My major issue however is making the conclusions stated here, using an amino-terminal fluorescent protein tag that s likely to impact any type of isoform selection at the end of the actin polymer. Carboxyl terminal tagging may have a reduced effect, but modifying the ends of the tropomyosin, which are integral in stabilising end to end interactions with itself on the actin filament, never mind any section systems that may/maynot be present in the cell, is not appropriate.

Significance

This paper explores the role of formin in determining the localisation of different tropomyosins to different actin polymers and cellular locations within budding yeast. Previous studies have indicated a role for the actin nucleating proteins in recruiting different forms of tropomyosin within fission yeast. In mammalian cells there is variation in the role of formins in affiecting tropomyosin localisation - variation between cell type. There is also evidence that other actin binding proteins, and tropomyosin abundance play roles in regulating the tropomyosin-actin association according to cell type. Biochemical studies have previously been undertaken using budding yeast and fission yeast that the core actin polymerisation domain of formins do not interact with tropomyosin directly.

The significance of this study, given the above, and the concerns raised is not clear to this reviewer.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Lupu et al have identified a role for the RNA binding protein SRSF3 in epicardial development. In a well written manuscript containing many experiments the authors show that this protein is required at different time points in epicardial development to control a range of processes, in particular cell proliferation. This advances understanding of the complex roles of this RNA binding protein in the heart - and raises an important message about how incomplete Cre recombination needs to be considered in interpreting conditional mutant phenotypes. The following points should be addressed.

1. SRSF3 is known to play essential developmental roles in the myocardium where it regulates capping of transcripts involved in contraction. This point should be mentioned in addition to roles in proliferation. To facilitate understanding, the authors should say more about the subset of cardiomyocytes labelled by Gata5-Cre. For example, is this the result of stochastic activation of the transgene or is a specific subset of cells labelled? How much of the myocardium is targeted?
2. The authors show failure of ventricular compaction at E13.5 using Wt1-CreERT2 and go on to assess proliferation in epicardial cells. As epicardial-derived signals are known to promote compact myocardial growth, they should also show whether there are indirect defects in proliferation in compact layer myocardium that might explain the non-compaction phenotype. The authors should also indicate if any of the large number of genes bound by SRSF3 encode known or potential pro-proliferative signals from the epicardium or EPDCs to the myocardium and potentially validate their altered expression in mutant hearts.
3. The rescue by expansion of non-recombined cells is a most interesting aspect of this study. Can the authors see any such outcompeting in the explant experiments (for example in Figure 2)? Do the authors consider this to be an exclusively in vivo competition phenomenon? Given the known roles of Myc in cell competition can the authors use their single cell transcriptomic data to score Myc expression levels in cells from Srsf3 iKO hearts or determine if Myc transcripts are bound by SRSF3?
4. The authors suggest that this rescue occur by upregulation of Srsf3 in non-recombined cells. It would be helpful to provide additional lines of evidence supporting the hypothesis that SRSF3 expression is upregulated due to hypoxia. Do the CLIP experiments reveal whether SRSF3 binds to it's own transcript?
5. The authors imply that SRSF3 may regulate Ccnd1 mRNA stability. Can the authors directly evaluate this point? Please clarify if this gene is also affected in the knock-down experiments in MEC1 cells.
6. Please brighten the immunofluorescence panels in Figure 1 to more clearly show nuclear labelling and tissue structure.
7. Given the broad roles of SRSF3 is the adjective key necessary in the title?

Significance

This ms advances understanding of the complex roles of this RNA binding protein in the heart - and raises an important message about how incomplete Cre recombination needs to be considered in interpreting conditional mutant phenotypes. This study would be of interest to reseachers in the fields of heart development and RNA-protein interactions. Although there are a number of major points to be addressed, these could be potentially dealt with rapidly.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this study, Lupu et al. analyzed the role of the RNA-binding protein SRSF3 for epicardial development. The authors found that Srsf3 is highly expressed in the proepicardial organ and during early stages of epicardial layer formation. Conditional inactivation of SRSF3 in the proepicardial organ stage using a Gata5-Cre driver line resulted in defective formation of the epicardium, accompanied by a proliferation arrest of the proepicardium, resulting in embryonic lethality at E12.5. In contrast, epicardial-specific Srsf3 deletion at later stages using the inducible Wt1CreERT2 line caused a less severe phenotype indicated by impaired coronary vasculature formation, reduced cardiac compaction, and myocardial hypoxia. Mosaic recombination yielded a small population of epicardial cells that upregulate Srsf3, hyperproliferate and compensate for the depleted Srsf3 negative lineage. Single-cell RNA sequencing of control and epicardial Srsf3 knock out hearts, combined with infrared CLIP to map SRSF3 binding sites in the transcriptome identified a number of putative SRSF3 targets involved in mitotic cell cycle control. Among others, SRSF3 binds directly to transcripts encoding key regulators of proliferation, such as Cyclin D1, and senescence, including MAP4K4. The authors conclude that SRSF3 exerts different functions in processing of RNAs, including splicing.

Overall, this is a well-written and well-organized manuscript, describing interesting findings in the field of epicardial development. However, the mechanistic part is not overly strong. The authors detected some moderate changes in the distribution of different Map4k4 splicing isoforms after knockdown of Srsf3 in an immortalized epicardial cell line but did not go any deeper. The cause for the reduced presence of transcripts for the SRSF3-target Ccnd1 after knockdown of Srsf3 remains enigmatic.

Significance

The authors raise a number of speculations why remaining Srsf3-expressing cell start to hyperproliferate after inactivation of Srsf3 but it does not become clear which mechanism is critical. How do non-targeted epicardial cells in the mosaic recombination model sense the loss of SRSF3 knock out cells, resulting in hyperproliferation and enhanced Srsf3 expression? Is a loss of lateral inhibition, e.g. by activated YAP/TAZ, causative for enhanced proliferation of the remaining epicardial cells and an elevated expression level of WT1 and SRSF3? Immunofluorescence staining and/or qRT-PCR of YAP/TAZ and TEADs might provide an answer.

Is the elevated expression of Srsf3 in non-targeted epicardial cells due to enhanced transcription and/or by altered post-transcriptional processes? How does this observation fit to previous reports indicating that Srsf3 overexpression promotes inclusion of an autoregulatory cassette exon (exon 4) containing a premature (in-frame) stop codon in Srsf3, thereby confining this SRSF3 isoform to nonsense mediated decay (NMD) (doi: 10.1093/emboj/16.16.5077, doi: 10.1186/gb-2012-13-3-r17, doi: 10.1038/srep14548, doi: 10.1161/CIRCRESAHA.118.31451)? In contrast, Srsf1 as well as PTBP1/2 have been previously reported to regulate Srsf3 expression by promoting exon 4 skipping. The authors should perform RNA seq and/or qRT-PCRs validation to check the inclusion of Srsf4 Exon4 as well as Srsf1 and PTBP1/2 expression levels in control and knock out epicardial cells.

It remains unclear by which mechanisms (alternative splicing, alternative polyadenylation, miRNA processing, or others) SRFS3 mainly exerts its function in the embryonic epicardial lineage. The selection and validation of Map4k4 as a splicing target is not based on an unbiased splicing analysis. In my opinion it is mandatory to provide a full assessment of splicing changes in Srsf3-deficient cells, either by long-range sequencing or by analysis of exon-junction reads.

Likewise, it is completely enigmatic what SRFS3 does to Ccnd1 transcripts. Does SRFS3 increase the half-life of Ccnd1, does it impact trafficking? At least, the authors have to determine changes in the half-life of Ccnd1 after depletion of SRFS3.

An unbiased bioinformatics analysis addressing alternative splicing, alternative polyadenylation, and mRNA processing is necessary. Ideally, primary epicardial cells should be used and not an immortalized epicardial cell lines. It is well known, that splicing in cell lines differs substantially from splicing in primary cells.

I am not convinced that the moderate changes of different Map4k4 splicing isoforms after knockdown of Srsf3 are really responsible for the rather drastic phenotype. Additional experiments are needed to prove a decisive function of a shift in Map4k4 splicing isoforms for hyperproliferation of epicardial cells.

The authors claim that inactivation of Srsf3 inhibits cell proliferation and causes a senescence-like phenotype. The claim for acquisition of senescence is solely based on transcriptional changes. No attempts were made to visualize an increase of senescent cells in Srsf3-mutant embryos. The authors need to perform SA-bGAL assays or use other techniques to analyse the appearance of senescent cells in the mutants.

Fig. 2E indicates that WT1 positive / tdTom negative epicardial cell population is enriched in a specific region of the pre-epicardial organ from Srsf3KOs. However it is not clear whether these cells proliferate. The authors should quantify Ki67positive cells in both the WT1-positive / tdTom-positive and the WT1positive / tdTom-negative epicardial population.

In the headline on page 6, the authors stated that "SRSF3 depletion in the PEO results in impaired ... migration of epicardial progenitor cells", which they deduced from the reduced outgrowth of ventricular epicardial explants. However, the reduced outgrowth from the PEO could be caused by both, reduced proliferation and/or reduced migration. Therefore, the authors should provide additional data clearly indicating reduced migration, e.g. by blocking transcription. Scratch assays of SRFS3 knockout/knockdown vs. control epicardial cells would strengthen the analysis, Is there a change in the GO term "regulation of migration"?

To prove the reduced proliferation ratio in Figure 4B, quantification of Cyclin D1 positive cells in both SRSF3 positive and negative cells is required.

Minor issues

Abstract line 12: a full stop is missing at the end of the sentence.

Figure 1A: E11.5, figure label 'DAPI WT1' is missing.

Page 8: Bracket in front of Fig. 4B is missing

Page 8: G2M phase change uniformly to G2/M phase

Page 9: 'Srsf3-depleted hearts also demonstrated an increased abundance of epicardial cells with upregulated expression of genes associated with quiescence, such as Clu48, and senescence, for example Map4k4, Tmem30a and Pofut249 (Fig. 4F)'. The sentence is misleading, implying Srfs3 inactivation in all cardiac cell types ('Srsf3-depleted hearts').

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

The study by Irina Lupu and colleagues highlights SRSF3 as a key regulator of epicardial development by regulating epicardial cell proliferation. This was demonstrated via two murine knockout models; the first to assimilate the role SRSF3 plays in epicardial formation as a whole, and the second to address its importance post a pivotal maturation point. Through scRNA sequencing and irCLIP, several SRSF3 targets were ascertained and identified as cell cycle regulators. Those epicardial cells that did not lose SRSF3 compensated the loss of some of their mates by increasing SRSF3 expression and over-proliferating. Overall, the paper is interesting and the conclusions are largely supported by the provided data.

Major comments:

1. Authors claim that a "reduction in SRSF3 expression levels coincided with the downregulation of WT1 in the epicardium". This was evidenced by immunofluorescence imaging (figure 1A). I suggest conducting a qRT-PCR to quantify Wt1 expression over time, similar to the experiment they performed in figure 1B.
2. A western blot depicting SRSF3 protein production in controls compared to the knockout model may provide stronger evidence of its depletion (figure 1E).
3. Authors state that they were unable to directly identify the absence of exons 2 and 3 in individual cells. Please provide evidence that exons 2 and 3 have been knocked out, at least by performing a qRT-PCR.
4. To prove the functional implication in the observed phenotype of the identified SRSF3 targets, please interfere with Map4k4 activity or expression and check whether the defective epicardial cell proliferation is reverted. This should be done at least in vitro, ideally in vivo.

Minor comments:

1. Several minor typos and spacing issues were observed. Please correct.
2. It would be good for the reader if the authors would simplify their rationale for the use of the two mouse models. It is slightly convoluted and not easy to follow.
3. In figure 4, it is recommended to add a stacked bar plot to represent the percentage of each cell cluster/population after 4A. This would help the reader
4. Figure 4B. It is confusing for the reader to understand the fact that the majority of tdTomato+ sorted cells in Srsf3 iKO keep expressing Srsf3. Including the quantification of the image could help.

Significance

The paper will be of interest to readers in the field of cardiology, embryology and molecular biology. It will advance the field especially in the study of the development of the epicardium. The models are sophisticated and the experiments carefully performed.

My field is molecular cardiology, with interest in RNA-binding proteins.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02835

Corresponding author: Eglantine, Heude

---

RESPONSE TO REVIEWERS

---

Minor changes include:

- concerns of Reviewer #1 with additional work to complete a new Supp. Fig. 3.

- some text modifications to avoid ambiguity and misunderstanding, and to address the concerns of Reviewers #1 and #2.

- addition of three references to develop some aspects of the discussion following the suggestions from Reviewer #3.

---

---

Reviewer #1

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This manuscript investigates the role of Dlx5/6 in development of the ear and vocal tract components. They generate a new conditional mouse knockout of Dlx5/6 using a Sox10 driver, which deletes functional forms of these factors in the otic placode (the primordium of the entire inner ear) and in neural crest cells, which contribute to the middle and outer ear, the otic capsule, as well as the craniofacial skeleton and cartilage including that of the larynx, and tendons.

The authors describe the ear and craniofacial phenotypes in these mutants using marker analysis on sections, some whole mounts and microCT imaging. They confirm previously described ear and craniofacial phenotypes, and add additional information on tongue, thyroid and hyoid cartilages, and associated mesoderm-derived muscles. Finally, they assess innervation finding subtle changes innervations patterns.

They conclude that Dlx5/6 are required for normal development of auditory and vocal structures and suggest that these factors coordinate both systems and that this is important for their co-evolution.

The experiments presented are straightforward phenotypic analysis, which is well presented with high quality imaging. However, it is not clear how well the conclusions are supported by the data because

i) The number of animals analysed and showing the reported phenotypes is not clearly stated; some figure legends list some n numbers, but it is sometimes ambiguous whether they refer to controls or experimental specimen. In other cases, n=2 is too low to draw firm conclusions.

We thank the reviewer for highlighting this point and for giving us the opportunity to clarify it.

The number of specimens mentioned correspond to the number of mutants analyzed versus the equivalent number of controls. For most figures, at least 3 mutants and 3 corresponding controls have thus been studied independently, unless for the microCT analysis (n=2 each condition) that however complete histological data on sections (n=6 each condition) presented at equivalent stage. Our data showing highly penetrant and reproducible results, the n numbers appear significant in support of our conclusions. To avoid ambiguity, we have systematically completed the n numbers with the mention 'each condition' in the figure legends.

ii) The nature of the controls is not stated, so we cannot know what is compared. Are controls WT, heterozygous for Dlx5 and -6, single KOs, single hets?

The lack of clarity in methods and controls does not allow others to reproduce the findings easily.

To address this concern, we have now added the genotype of controls in the Materials and Methods section. For all experiments, the controls of mutant specimens were from the same littermate. We included as control genotypes the specimens heterozygous for the Cre and/or flox or homozygous for the flox only (genotypes equivalent to breeders), which we had previously validated as showing a normal phenotype compared to non-genetically modified 'wild-type' specimens. This observation is consistent with what had been described previously in constitutive inactivated Dlx5/6 mouse models, for which the development of heterozygous animals was unaffected by the mutation (e.g. Heude et al. 2010 PNAS).

---

Based on the fairly limited data presented here, the authors make some far-reaching suggestions that are not supported by experimental findings. For example, they propose that their study points to "co-adaptation of the effector and receptor organs during acoustic communication diversification in land vertebrates" and that Dlx5/6 play a role in this process. This idea appears to be the main motivation for the current study, however, there is little evidence to support such conclusions.

Likewise, they suggest that a "common Sox10-Dlx5/6-BMP signaling axis coordinates the morphogenesis of both CNCC and otic placode derivatives for the proper formation of the vocal and auditory complex". There is no evidence provided that Sox10 controls any of the other genes and functional experiment related to Sox10 are not carried out, while the Dlx5/6-BMP link has been established previously.

Overall, the authors need to improve numbers, experimental information and controls, and given the descriptive nature of the manuscript they should refrain from wide-ranging conclusions.

We regret that Reviewer #1 took our discussion points as conclusions, which was not our purpose. Rather, Reviewer #3 stated that 'The authors are careful in the way they present the evidence, and only go beyond their data to add a few 'speculative' paragraphs at the end of the Discussion' (see below). We believe that the Discussion section is the appropriate space to formulate hypotheses based on experimental results, mainly while speculating on evolutionary aspects, that cannot be technically tested but are still conceptually of interest.

Concerning the Sox10-Dlx5/6-BMP axis, we agree with Reviewer #1 that we do not provide evidences that this common genetic program regulates the formation of vocal and auditory systems. We have added additional nuance to the text to avoid ambiguity, and we expect that our propositions will inspire future research on the issues raised.

The analysis could also be strengthened by focusing on the novel aspects, providing a developmental time series as to when phenotypes are first observed combined with marker analysis e.g. looking at different compartments of the otic vesicle, cell types etc., and providing higher magnification images to describe the phenotypes (e.g. those in figure 1) will strengthen the paper and might provide some novelty.

Reviewer #1 (Significance (Required)):

The authors present a new conditional knock out line for Dlx5/6. The major limitation of the study is that the data presented largely confirm what has already been published including the regulation of BMP pathway components and non-cell autonomous effects on muscle. As far as I am aware, the only new additions to the literature are the analysis of the cartilages and tongue.

The phenotypic analysis is minimal, there is no developmental time series and no other functional experiments to address some of the suggestions made by the authors.

My expertise is in ear formation, and I am aware of the craniofacial literature.

The novelty of our study is based on the generation of a new conditional mouse model to address the pleiotropic role of Dlx5/6 in coordinating the development of both the vocal tract and auditory components, an aspect that has never been considered before.

To this end, we have combined a variety of high-resolution imaging techniques to achieve an unprecedented analysis of critical markers of placode, neural crest and mesoderm derivatives (including a neural crest lineage analysis). We initially selected key early embryonic (E10.5-E11.5), fetal (E14.5) and perinatal (E16.5-E17.5) stages, but we agree with Reviewer #1 that an additional stage is relevant in our context to complete this developmental time series. We have now added in a new Supp. Fig. 3 including the analysis of a late embryonic stage (E12.5). In particular, our results show that primary chondrogenic condensations are affected at the level of the vocal tract and auditory compartments. Furthermore, our data reveal that the differentiation defects of the masticatory and tongue musculature occur at the transition between early (E11.5) and late (E12.5) embryonic stages. We are grateful to Reviewer #1 for his comments, which allowed us to present these new and interesting results.

Reviewer #2

---

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript, Sanchez-Garrido and colleagues analyze the consequences of Dlx5/6 deletion on cranial neural crest- and otic placode-derived structures development. The authors show that Dlx5/6 inactivation causes broad defects of the outer, middle and inner ear as well as malformations of the jaw, pharynx and larynx musculoskeletal systems. The authors conclude that Dlx5/6 play an important role in the regulation of vocal and auditory systems development in mammals.

Reviewer #2 (Significance (Required)):

The work is well-presented and illustrated, with appropriate description of methods. The role of this gene family in regulating craniofacial structures is not completely novel. The analysis is largely descriptive providing little mechanistic insights. The authors propose a possible link between Dlx5/6 activity and BMP signaling as the culprit for the defects observed in the mutants, however this is not supported experimentally. The work is viewed as too preliminary at this stage.

---

We are appreciative of the feedback from Reviewer #2. Our aim was to gain insight into the genetic basis of the coordinated development of the vocal tract and hearing complex. As mentioned above for Reviewer #1, we have added modifications to the discussion to nuance and develop our propositions. We hope that our study will help to further develop some of the aspects we have explored.

Reviewer #3

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

This is a very good paper presenting results examining the role of DLX5/6 in the formation of ear and vocal tract structures by means of mutant mice. I believe the evidence offered in this paper in favor of such a role is strong and well presented.

Reviewer #3 (Significance (Required)):

I think this is a very good study, contributing to our knowledge of the action of DLX5/6, giving us insights into the genes' role in ear and vocal tract structures, critical for vocalization and the pairing of auditory and oral capacities. The authors are careful in the way they present the evidence, and only go beyond their data to add a few 'speculative' paragraphs at the end of the Discussion.

I would encourage the authors to consider making mention of Gokhman et al's (2021) work, where genes similarly impacting craniofacial and oral capacities are discussed: https://www.nature.com/articles/s41467-020-15020-6

I would also have liked the authors to cite work on GLI3 and larynx formation (given that both GLI3 and HAND2 control DLX5 and 6) and play an important role in Neural Crest-related processes: https://elifesciences.org/articles/77055 / https://elifesciences.org/articles/56450

In this context, the authors mention work on the role of Dlx5/6 in GABAergic neurons. Do the authors believe there to be a connection (mediated by the SHH pathway) between the ganglionic eminence and neural crest development, or are these two findings only convergent at the level of the phenotype?

We are grateful to Reviewer #3 for his positive assessment of our study and for his very relevant suggestions of works to consider in the context of our research. This has allowed us to complete and develop exciting points in our discussion, which we hope will stimulate further lines of research regarding the diversification of acoustic communication abilities during mammalian evolution, including the emergence of speech in humans.

To answer the reviewer's question, we do not believe that there is a link between the ganglionic eminences and neural crest derivatives, which develop independently in two embryonic compartments under the regulation of distinct genetic programs. However, the observation that Dlx5/6 expression in the brain regulates socialization and vocalization behaviors in adult mice suggest that the genes have a broader pleiotropic role in acoustic communication, both during development and in the adult at the cognitive level.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This is a very good paper presenting results examining the role of DLX5/6 in the formation of ear and vocal tract structures by means of mutant mice. I believe the evidence offered in this paper in favor of such a role is strong and well presented.

Significance

I think this is a very good study, contributing to our knowledge of the action of DLX5/6, giving us insights into the genes' role in ear and vocal tract structures, critical for vocalization and the pairing of auditory and oral capacities. The authors are careful in the way they present the evidence, and only go beyond their data to add a few 'speculative' paragraphs at the end of the Discussion.

I would encourage the authors to consider making mention of Gokhman et al's (2021) work, where genes similarly impacting craniofacial and oral capacities are discussed. https://www.nature.com/articles/s41467-020-15020-6

I would also have liked the authors to cite work on GLI3 and larynx formation (given that both GLI3 and HAND2 control DLX5 and 6) and play an important role in Neural Crest-related processes: https://elifesciences.org/articles/77055 https://elifesciences.org/articles/56450 In this context, the authors mention work on the role of Dlx5/6 in GABAergic neurons. Do the authors believe there to be a connection (mediated by the SHH pathway) between the ganglionic eminence and neural crest development, or are these two findings only convergent at the level of the phenotype?

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Sanchez-Garrido and colleagues analyze the consequences of Dlx5/6 deletion on cranial neural crest- and otic placode-derived structures development. The authors show that Dlx5/6 inactivation causes broad defects of the outer, middle and inner ear as well as malformations of the jaw, pharynx and larynx musculoskeletal systems. The authors conclude that Dlx5/6 play an important role in the regulation of vocal and auditory systems development in mammals.

Significance

The work is well-presented and illustrated, with appropriate description of methods. The role of this gene family in regulating craniofacial structures is not completely novel. The analysis is largely descriptive providing little mechanistic insights. The authors propose a possible link between Dlx5/6 activity and BMP signaling as the culprit for the defects observed in the mutants, however this is not supported experimentally. The work is viewed as too preliminary at this stage.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript investigates the role of Dlx5/6 in development of the ear and vocal tract components. They generate a new conditional mouse knockout of Dlx5/6 using a Sox10 driver, which deletes functional forms of these factors in the otic placode (the primordium of the entire inner ear) and in neural crest cells, which contribute to the middle and outer ear, the otic capsule, as well as the craniofacial skeleton and cartilage including that of the larynx, and tendons.

The authors describe the ear and craniofacial phenotypes in these mutants using marker analysis on sections, some whole mounts and CT imaging. They confirm previously described ear and craniofacial phenotypes, and add additional information on tongue, thyroid and hyoid cartilages, and associated mesoderm-derived muscles. Finally, they assess innervation finding subtle changes innervation patterns.

They conclude that Dlx5/6 are required for normal development of auditory and vocal structures and suggest that these factors coordinate both systems and that this is important for their co-evolution. The experiments presented are straightforward phenotypic analysis, which is well presented with high quality imaging. However, it is not clear how well the conclusions are supported by the data because

i) The number of animals analysed and showing the reported phenotypes is not clearly stated; some figure legends list some n numbers, but it is sometimes ambiguous whether they refer to controls or experimental specimen. In other cases, n=2 is too low to draw firm conclusions.

ii) The nature of the controls is not stated, so we cannot know what is compared. Are controls WT, heterozygous for Dlx5 and -6, single KOs, single hets?

The lack of clarity in methods and controls does not allow others to reproduce the findings easily. Based on the fairly limited data presented here, the authors make some far-reaching suggestions that are not supported by experimental findings. For example, they propose that their study points to "co-adaptation of the effector and receptor organs during acoustic communication diversification in land vertebrates" and that Dlx5/6 play a role in this process. This idea appears to be the main motivation for the current study, however, there is little evidence to support such conclusions. Likewise, they suggest that a "common Sox10-Dlx5/6-BMP signaling axis coordinates the morphogenesis of both CNCC and otic placode derivatives for the proper formation of the vocal and auditory complex". There is no evidence provided that Sox10 controls any of the other genes and functional experiment related to Sox10 are not carried out, while the Dlx5/6-BMP link has been established previously. Overall, the authors need to improve numbers, experimental information and controls, and given the descriptive nature of the manuscript they should refrain from wide-ranging conclusions. The analysis could also be strengthened by focusing on the novel aspects, providing a developmental time series as to when phenotypes are first observed combined with marker analysis e.g. looking at different compartments of the otic vesicle, cell types etc., and providing higher magnification images to describe the phenotypes (e.g. those in figure 1) will strengthen the paper and might provide some novelty.

Significance

The authors present a new conditional knock out line for Dlx5/6. The major limitation of the study is that the data presented largely confirm what has already been published including the regulation of BMP pathway components and non-cell autonomous effects on muscle. As far as I am aware, the only new additions to the literature are the analysis of the cartilages and tongue. The phenotypic analysis is minimal, there is no developmental time series and no other functional experiments to address some of the suggestions made by the authors.

My expertise is in ear formation, and I am aware of the craniofacial literature.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors first use a Bio-ID approach to search for interactors of the basket proteins TPR and NUP153, identifying proteins involved in various nuclear process, including many splicing components, and confirm some of these interactions using IP and PLA assays. PLA experiments further suggest that these interactions occur primary at or close to the nuclear periphery. Moreover, inhibiting splicing, but not transcription, reduced these interactions. The authors then investigated the role of NUP153 in loading of the splicing machinery and found a lower association of the NUP98/SF3A1 but not AQR interaction (measured through PLA). Furthermore, DamID experiments identified NUP153 bound genes proximal to LAD domains that are actively transcribed, contain overall longer introns with low GC-content, and fall within a group of genes located at the outermost shell of the nucleus (when compared to previously published LaminI ID /PGseq data). Lastly, they interrogate whether depletion of NUP153 results in a splicing defect for NUP153 bound genes.

The authors identify many proteins in their BioID interaction screen, however, only a single nucleoporin (Nup35, an inner ring protein). Previous BioID studies have identified NUP153 in BioID experiments including proteins of the Y-complex (PMID: 24927568 and others).

The BioID list summarises interactions of proteins present across five datasets and among two cell lines, HEK293T and Jurkat T cells. As the reviewer pointed out, these stringent criteria excluded proteins originally present in the BioID datasets. Indeed, the original datasets across the cell lines include a wide range of nucleoporins Nup37, Nup43, Nup53, Nup85, Nup107, Nup188, and Nup205. Apart from this, several other proteins were consistently found on the BioID of the basket nucleopore that had been previously found in the literature, namely transcription-related and export-related proteins, as the reviewer can depict on Fig 1 C.

To ensure that the BioID experiments indeed probe for interactions of NUP152 and TPR at the NPC, the authors should include control experiments that show that their NUP153 and TPR-BirA fusions primarily localize to the NPC. If a significant fraction is not NPC bound, this has to be taken into account when interpreting/discussing their data.

Before conducting the DamID experiment, we validated the peripheral localization of the constructs used here. We provide here the images showing the distribution of the two nucleoporins

Figure R1: Immunofluorescent images obtained for Nup153 and TPR in Jurkat cells prior to BioID experiment.

The authors describe DDX39b/UAP56 as an early splicing factor; DDX39b/UAP56 main role however seems to be in mRNA export and mRNP compaction. The authors might want to include this in the interpretation of their data.

The reviewer is right; the role of DDX39b/UAP56 goes beyond pre-mRNA splicing. This is indeed involved in posttranscriptional maturation and export from the nucleus to the cytoplasm of cellular RNAs. Originally identified as a helicase involved in pre-mRNA splicing, UAP56 has been shown to facilitate the formation of the A complex during spliceosome assembly. This is the reason why we included it in our study. Additionally, DDX39b/UAP56 has been found to be critical for interactions between components of the exon junction and transcription and export complexes to promote the loading of export receptors, while more recently has also been identified as a DNA:RNA helicase involved in the resolution of R-loops (PMID:32439635). At present, we indeed cannot distinguish between multiple functions of this protein at the NPC, and this is now acknowledged in our manuscript.

• Concerning PLA experiment controls, the authors perform TPR-PML as a negative control, however, no negative control for NUP153 is shown (Figure 2). Such a control should be added to allow evaluating the specificity of NUP153 PLA interactions, and/or discussed why this was not done. We would like to clarify better how we selected the antibodies for the control PLA experiments. In PLA, antibodies from different species have to be used. Nup153 is an anti-mouse antibody, and the control we used, PML, is also an anti-mouse; therefore, the two antibodies cannot be used in the PLA experiment. The controls used (mouse) were conjugated with rabbit antibodies, either TPR (PML) or AQR (B23). Both TPR:PML and AQR:B23 showed insignificant PLA signals. Therefore, we can confidently conclude that the PLA spots seen for the NPC/splicing proteins are of measurable quantities. Moreover, the conclusion that there’s a pool of splicing machinery associated with the NPC is sustained on several pieces of evidence accumulated through other experiments, not only PLA. We have supporting evidence from super-resolution microscopy, coIPs as well as the referred PLA.

Figure R2: Control PLA assay with anti-AQR (rabbit) and Nucleophosmin (B23)(mouse) antibodies.

• Quantification of the distance of PLA-NUP153/TPR interactions show interactions mainly close to the nuclear periphery. The imaging data shown in Figure 2b indeed shows that TPR/NUP153 interactions are exclusively at the nuclear periphery, whereas NUP153/splicing factor interactions are sometimes at the edge of the DAPI signal, but mostly somehow internalized (Figure 2B, S2b). Quantification (Figure 2d) shows these distributions to be very similar, likely due to the way the quantification was performed / the bin size of plotting the relative distance of a spot to the nuclear periphery was chosen. Looking at the scale bar/nuclear size and the position of the PLA spots for the NUP153/splicing factors, it appears that spots are often hundreds of nanometers away from the periphery. As the nuclear basket is thought to reach only about 100nm onto the nuclear interior, the conclusion by the authors that these interactions occur at the NPC would not be consistent with the data. The authors should better incorporate this in their interpretation of the data. Nup153/TPR are peripherally located at the most outer shells (0, 40% of signal and 1, 60% of signal). Consistently, based on figure 2d, Nup153:DDX39b is similarly distributed (0, 40% of signal and 1, 60% of signal) and the vast majority of the Nup153:AQR and Nup153:SF31A1 signal is also present in these two shells (20% and 40%). Although our super-resolution microscopy excludes the presence of internal Nup153 staining we cannot exclude that PLA potentially increases the signal of a possible internal Nup153/splicing interaction due to the rolling circle amplification reaction. However, as referred above this is not where primarily our interaction is occurring.

• The conclusion ' Nup153 aids the loading of splicing machinery' is not sufficiently supported by the data. The authors observed a reduction in PLA signal for the NUP98-AQR interaction, but not the NUP98-SF3A1 (Figure 3g). Their conclusion has to reflect this discrepancy in their data. Moreover, the studies focus is to determine the role of NUP153/TPR in recruiting the splicing machinery to the NPC. As in the experiments the authors interrogate the interaction of only NUP98, who has to a large extend splicing factor interactions within the nuclear interior and not at the periphery, the relevance of the experiments in Figure 3 towards the main focus of the paper is unclear. The reviewer is right to point out that the study focuses on determining the role of Nup153/TPR in recruiting the splicing machinery to the NPC. As TPR is docked to the NPC through Nup153 (PMID: 12802065; 39127037) we investigated Nup98 and performed internal controls to show that 1) Nup98 wasn’t disrupted by shNup153 (Fig below) and technically 2) Nup98 was used in the shNup153 studies because of the availability of a reliable mouse antibody that could be coupled with the various rabbit antibodies used previously for SF3A1, DDX39, XAB2, and AQR in PLA experiments' for which mouse counterparts do not exist and would therefore hinder the continuation of the study as explained above. For TPR only a reliable rabbit antibody exists that works in our hands and therefore we wouldn’t have been able to perform the PLA experiments shown here

Figure R3: Nup98 staining in wild type and shNup153 depleted Jurkat cells. In (a) co-immunofluorescence between AQR and Nup98 showing predominant positioning of Nup98 in the nuclear periphery (at the NPC). (b) Nup98 staining at the periphery persists also in shNup153 depleted cells, indicating that this Nup can be used as an NPC marker in Nup153 depleted conditions.

When investigating the effect of NUP153 depletion on splicing, the authors observe a splicing phenotype for multiple NUP153 genes (Figure 5). The authors however show only a single negative control gene (CBX5). It would significantly strengthen their argument if the authors would investigate splicing defects of periphery located noneNUP153 bound genes as well as for genes located in the nuclear interior to better understand whether this splicing phenotype is indeed specific for NUP153 genes (at the nuclear periphery/NPC).

We agree with the reviewer that expanding our observations to other peripherally located genes would be interesting; however, most of the other known peripheral genes are LAD-associated and mostly not expressed. While it was not our intention to claim that splicing at the periphery is specific only for Nup153-bound genes, we had obviously focused on Nup153-bound genes to understand the dynamics between Nup153/splicing machinery interactions. As stated above, other peripherally located genes within LADs are repressed.

It Is out of the scope of this study to understand other relevant splicing hubs, as the reviewer knows splicing can occur throughout the nucleus at different sites (outside speckles). However, we do understand the reviewer's point of view, and to include more controls as requested by this and other reviewers, we have designed primers for additional non-Nup153 bound genes, and these additional experiments will be included in the manuscript.

Figure R4: Preliminary data showing splicing of other Non-Nup153 Bound genes upon shNup153

The authors state in the text describing the SABER-FISH experiments in Figure 5f that 'were able to visualize the presence of a site of transcription where accumulation of these probes was close to the periphery for all except for GSTK1, which showed a wider nuclear distribution, similar to CBX5 control region not bound by Nup153'. However, their statement is not supported by the images shown in Fig 5f, which show TS in control cells in the nuclear interior. Also, a single cell but no quantification is shown. Moreover, what distance from the periphery is considered as close to the periphery is not defined (see also earlier comment on the question what should be considered a periphery and/or NPC association).

Measurement of the distance of FISH signals to the nuclear periphery for each probe (ie transript) performed in n>100 cells were represented graphically in Fig. 5f. We then compared total number of signals for each probe, obtained in shCtrl and shNup153 cells, and represented them graphically in Fig 5g; representative images shown in Fig 5g are those that were measured in Fig5g and represent the accumulation of the signal in cells upon shNup153 (not necessarily all at the nuclear periphery).

We hope that this clarifies better what is represented.

Limitation of the study does not discuss the limitations of the study but rather reads like the extension of the discussion. This section should be rewritten.

We will take into consideration this comment and will expand the section in the revised manuscript

Minor comments:

Western in Figure 3c does not represent well the quantification in 3e.

Figure S3 is mislabelled (pannel h is panel g).

__Reviewer #1 (Significance (Required)): __

Reviewer #1 (Significance (Required)):

The manuscript interrogates an important question related to the role of the NPC in gene regulation, in particular how interaction of genes/pre-mRNAs with the NPC might stimulate expression of specific genes/mRNAs. Stimulating splicing would be one way that could contribute to efficient gene expression, and this is the question the authors address in this manuscript. This study is therefore important and relevant to a wide audience. However, as outlined in the section above, the conclusions drawn by the authors do not always reflect the experimental data, and it is therefore unclear whether the overall conclusion as stated in the title of the manuscript is valid. Moreover, conceptually, if intron containing genes are transcribed at or near nuclear pores, and splicing often occurs co-transcriptionally, it is to be expected to find splicing components close to nuclear pores. While it is relevant to show that this actually happens, and this is, at least in part, done by the authors. However, the experiments presented do not show that the splicing machinery actually actively docks to the NPC and is not just passively recruited close to NPCs because nascent pre-mRNAs are spliced where they are transcribed (the authors state in their title that the NUP153 docks the splicing machinery at the NPC). Showing this require identifying direct interactions between spliceosome components with NUP153/nuclear basket components to stimulate splicing at the NPC. If this would indeed be the case, these findings would describe a novel mechanistic step to stimulate efficient splicing and subsequently export of a selected set of NPC-associated genes. This would open other questions such as how to achieve specificity for only some pre-mRNAs/introns. While addressing this question is likely beyond the scope of this manuscript, the question whether the process described here is an active or passive process should be incorporated in the interpretation of the data.

We are grateful to the reviewer for highlighting the importance and relevance of our work for a broad audience. We now provide additional experimental evidence that will hopefully aid in substantiating our overall conclusions, as suggested by the reviewer.

__ Reviewer 2:__

Summary:

In this manuscript, using a combination of proximity labelling, immunoprecipitations and imaging, the authors report a physical interaction between splicing factors (SFs) and the nuclear basket of nuclear pore complexes (i.e. NUP153). Using DamID, they further identify a set of NUP153-bound genes characterized by long, GC-poor introns. Finally, based on molecular analyses for a set of candidate loci, they report that inactivation of NUP153 triggers a (modest) reduction of intron splicing, which may specifically affect NUP153-bound genes.

Major comments:

• The BioID experiments (Fig. 1) lack proper controls. Proteins biotinylated by NUP-BirA fusions need to be compared with those modified upon expression of a control BirA protein, as has been done previously, especially when other NUPs were used as baits in BioID experiments (PMID: 24927568, to be cited). This control fusion should ideally be targeted to the same compartment (i.e. the nucleus or the nuclear side of the nuclear envelope). In our experimental setting, we have opted to use an unrelated protein tagged with BirA (Lck-BirA) rather than BirA-only control. The peripheral membrane proteins of the Src family kinase (SFK) Lck and its GPF tagged version (LckN18.GFP) localize predominantly at the plasma membrane (PMID: 29588370), whereas the GFP only (non-biothinylated) shows a broader nuclear distribution. All MS-detected proteins from the Lckn18-BirA and GFP negative control experiments were excluded. Moreover, we have analysed carefully the published data of nuclear transporter receptors binding to the NPC and the respective controls (BirA alone or the shuttling NLS_NES_Dendra with C or N terminal tags) (PMID: 29254951), and we did not find that any of these protein controls interact with the proteins of the splicing machinery.

• Here, the chosen controls are inappropriate as the authors are probing interactions between NPC proteins (NUP153/TPR) and proteins restricted to a different nuclear compartment (e.g., nucleophosmin in the nucleolus). We have used two different controls in our PLA experiments - initially, we used B23 for nucleolus stain and then PML protein, major component of PML NBs, as PML can be found scattered throughout the nucleus and sometimes even resides at the nuclear periphery. Both of these controls showed negligible amounts of PLA spots in all our experiments. We take the opportunity to clarify that in PLA, antibodies from different species have to be used. Nup153 is an anti-mouse antibody, and the control we used, PML is also an anti- mouse; therefore the two antibodies cannot be used in the PLA experiment. The controls used (mouse) were conjugated with rabbit antibodies, either TPR (PML) or AQR (B23). Both TPR:PML and AQR:B23 showed insignificant PLA signals. Therefore we can confidently conclude that the PLA spots seen for the NPC/splicing proteins are of measurable quantities.

• Would a control soluble, diffusible nucleoplasmic protein be detected in the vicinity of the NPC and sometimes colocalized with Nups? Precisely for this purpose we have used PML protein, that can be found both disperse in nucleoplasm as well as in PML NBs.

• In order to assess these possibilities, the authors should perform their immunoprecipitation on extracts treated with benzonase, thus abrogating DNA- and RNA-dependent interactions. We have performed this experiment assessing the binding of Nup153 with the components of the IBC and observed that Nup153 interaction with these splicing factors is DNA or RNA independent, with some factors being more affected than others (probably passively recruited by protein-protein interactions with their splicing counterparts).

__Figure R5: __RNA or DNA do not affect the interaction between Nup153 and splicing proteins. Lysates from HEK293T cells transfected with eGFP-Nup153 or eGFP were treated with RNase or DNase or left untreated prior to Co-IP for GFP-Nup153. The membrane was probed for GFP (confirmed successful transfection), AQR, SF3A1, XAB2, and DDX39B. The graph shows quantified bands of the bound fractions, normalized to the input and untreated control from one experiment.

NUP153 inactivation appears to have a modest effect on splicing (Fig. 5; S6), which is poorly characterized here. It is also unclear whether this effect is direct or caused by side consequences of the depletion of this nucleoporin (e.g., changes in nucleocytoplasmic exchanges or gene expression).

We have indeed asked if the presence of splicing components at the periphery could be a consequence of protein trafficking. To address whether nucleocytoplasmic exchange has a role in these associations, we have pharmacologically inhibited nuclear import by ivermectin (IVM). ​​IVM has been shown to block the importin-α/β-mediated nuclear import by directly interacting with karyopherin importin-α.(PMID: 30826604). HEK293T cells transfected with eGFPNup153 or eGFP alone were treated with IVM for 2hr (24hrs post transfection). As biochemical fractionation demonstrated (data not shown) there were slightly decreased protein levels of AQR, DDX39B and SF3A1 in the nuclear insoluble fraction. However, we barely observed any decrease in interactions between Nup153 and the splicing components we tested, indicating that the interaction with the spliceosomal components is not a consequence of nucleocytoplasmic exchange.

Figure R6: Association of splicing components with Nup153 is not only due to nuclear import. HEK293T cells transfected with eGFP-Nup153 or eGFP and treated with import inhibitor IVM or DMSO were analyzed with Co-IP for GFP-Nup153. The membrane was probed for GFP (confirmed successful transfection), AQR, SF3A1, XAB2 and DDX39 B. Quantified bands of the bound fractions were normalized to the input fraction and DMSO control and the results of two experiments were shown in the graph.

Related to the gene expression levels, we have not observed any significant changes in total expression levels of tested genes, probed with designed exonic primers (as indicated with blue arrows in Figure 5a). Additional control genes will be added as suggested by this and other reviewers.

• To confirm the specificity of the effects of NUP153 depletion on the splicing of NUP153-bound genes, the authors need to provide additional splicing measurements for several genes that are bound by NUP153 "in the nucleoplasm" (e.g. excluded from their analysis by the cutoff of proximity to LAD borders, line 192) and for other "non-NUP153" genes (beyond the unique control shown in Fig. S6a).

We acknowledge the comment of the reviewer that our work will benefit from additional controls. We are currently designing primers and probes to amplify additional regions, Nup153 bound and non-LAD proximal or non-Nup153 bound; Please also see the comment below.

- From the few examples provided, it is difficult to evaluate the type of splicing events affected by NUP153 inactivation. Are they uniquely intron retention events? The authors should analyze available RNA-seq data obtained from NUP153-depleted cells (PMID:32917881) to characterize the types of alternative splicing events that are impacted by NUP153.

In Aksenova et al, only 28 differentially expressed genes were detected during rapid degron Nup153 depletion (2h). With this small number of genes, it is highly unlikely we would be able to perform a statistically significant and detailed analysis. Importantly, the depletion was performed in colorectal adenocarcinoma cell line (DLD-1), whereas here we are reporting on T lymphocytes. Based on the analysis which we performed and explained below, there seem to be significant cell type specific differences in Nup153 associations, as already reported by others (PMID: 27807035; 32451376; 28919367).

Splicing dynamics and speckle localization propensity have been proposed to depend on the overall GC content and the overall average intron size by several studies (PMID: 39413186; 38720076; 35182478; 22832277; 35182477). Prompted by our observation that Nup153 genes have longer than average introns and lower than average GC content (Figure 4f and 4g), we analyzed the data from HeLa cells, where genes were classified into groups A, B and C based on their speckle localization and dynamics (PMID 39413186). We intersected our Nup153 genes with the list of ABC genes from HeLa cells and found that 43 out of our 461 protein-coding genes were represented among non-speckle enriched group C genes, with the lowest GC content and longest average intron length.

Figure R7: Nup153 genes comparison to the A_B_C genes from Wu J et al study. GC content and number of introns, used to classify the identified genes from HeLa cells are plotted on X and Y axis. A genes in Red, B in green or C in turquoise from HeLa cells were compared with Nup153 genes from Jurkat cells in the graph on the left. Nup153 genes are represented as triangles. A subgroup of Nup153 genes, classified as C group genes (long introns with high GC content and spliced away from speckles) are shown as turquoise triangles. Graph on the right shows total pool of Nup153 genes in violet (not identified in HeLa cells) and a subgroup of C Nup153 genes as turquoise. The list of Nup153 C genes is shown below.

query

entrezgene

name

symbol

ENSG00000168615

8754

ADAM metallopeptidase domain 9

ADAM9

ENSG00000139154

121536

AE binding protein 2

AEBP2

ENSG00000112249

10973

activating signal cointegrator 1 complex subunit 3

ASCC3

ENSG00000176788

10409

brain abundant membrane attached signal protein 1

BASP1

ENSG00000153956

781

calcium voltage-gated channel auxiliary subunit alpha2delta 1

CACNA2D1

ENSG00000153113

831

calpastatin

CAST

ENSG00000134371

79577

cell division cycle 73

CDC73

ENSG00000188517

84570

collagen type XXV alpha 1 chain

COL25A1

ENSG00000182158

64764

cAMP responsive element binding protein 3 like 2

CREB3L2

ENSG00000109861

1075

cathepsin C

CTSC

ENSG00000153904

23576

dimethylarginine dimethylaminohydrolase 1

DDAH1

ENSG00000139734

81624

diaphanous related formin 3

DIAPH3

ENSG00000102580

5611

DnaJ heat shock protein family (Hsp40) member C3

DNAJC3

ENSG00000173852

23333

dpy-19 like C-mannosyltransferase 1

DPY19L1

ENSG00000151914

667

dystonin

DST

ENSG00000165891

144455

E2F transcription factor 7

E2F7

ENSG00000138829

2201

fibrillin 2

FBN2

ENSG00000115414

2335

fibronectin 1

FN1

ENSG00000075420

64778

fibronectin type III domain containing 3B

FNDC3B

ENSG00000114861

27086

forkhead box P1

FOXP1

ENSG00000090615

2802

golgin A3

GOLGA3

ENSG00000196591

3066

histone deacetylase 2

HDAC2

ENSG00000071794

6596

helicase like transcription factor

HLTF

ENSG00000145012

4026

LIM domain containing preferred translocation partner in lipoma

LPP

ENSG00000065833

4199

malic enzyme 1

ME1

ENSG00000087053

8898

myotubularin related protein 2

MTMR2

ENSG00000145555

4651

myosin X

MYO10

ENSG00000061676

10787

NCK associated protein 1

NCKAP1

ENSG00000185630

5087

PBX homeobox 1

PBX1

ENSG00000113448

5144

phosphodiesterase 4D

PDE4D

ENSG00000163110

10611

PDZ and LIM domain 5

PDLIM5

ENSG00000070087

5217

profilin 2

PFN2

ENSG00000152952

5352

procollagen-lysine,2-oxoglutarate 5-dioxygenase 2

PLOD2

ENSG00000106772

158471

prune homolog 2 with BCH domain

PRUNE2

ENSG00000173482

5797

protein tyrosine phosphatase receptor type M

PTPRM

ENSG00000164292

22836

Rho related BTB domain containing 3

RHOBTB3

ENSG00000067900

6093

Rho associated coiled-coil containing protein kinase 1

ROCK1

ENSG00000112701

26054

SUMO specific peptidase 6

SENP6

ENSG00000154447

57630

SH3 domain containing ring finger 1

SH3RF1

ENSG00000187164

57698

shootin 1

SHTN1

ENSG00000198887

23137

structural maintenance of chromosomes 5

SMC5

ENSG00000116754

9295

serine and arginine rich splicing factor 11

SRSF11

ENSG00000152818

7402

utrophin

UTRN

Table R1: List of Nup153 genes that are characterized as C group of genes.

Only two Nup153 gene were found among A and B genes (Serine and arginine rich splicing factor 11 SRSF11 among A, and Phosphodiesterase 4D PDE4D among B genes). Despite the cell type specific expression and splicing patterns it is worth noting that we find Nup153 genes enriched among C group genes that are spliced out of speckles. We are currently probing the splicing of some of these genes, and these data will be added to the list of control genes.

Considering all these new observations related to the Nup153 splicing events and the general interest and relevance of our initial observations, a new dedicated study will have to be designed to tackle all these important questions that go beyond these current findings

---

Minor comments:

• Several studies have shown that the nuclear basket contributes to a splicing quality control process preventing the nuclear export of improperly spliced transcripts, both in yeast and mammalian cells (PMID:14718167, 19127978, 24452287, 25845599, 22253824, 22661231). These studies have to be mentioned and discussed here.

• Line 31: "movement of active genes towards the NPC would be favorable for their transcription and export ". Please rephrase: "...transcription and mRNA export".

• Line 163: "NUP153 plays a role in harboring splicing machinery". Please rephrase.

• Line 200-202: Fig. 4d and 4e (instead of S4d and S4e)

• Line 186 and beyond: All conclusions about NUP153-bound genes (e.g., "Majority of NUP153 bound genes are proximal to LADs and expressed") are not accurately phrased since the authors selected NUP153-bound genes with a cutoff of proximity to LAD borders. The conclusions are thus only valid for a subpopulation of NUP153-bound regions located in the vicinity of LADs.

• Line 292: "transport Nups less likely interact with splicing machinery". The term "transport Nup" is not correct. Does this mean "nuclear transport receptors"? Or "FG-Nups" (which interact with NTRs)?

All the comments will be addressed in the revised version of the manuscript

Reviewer #2 (Significance (Required):

It is increasingly recognized that NPCs are involved in a number of cellular processes beyond nucleo-cytoplasmic transport and, in particular, contribute to several genomic functions. In this context, the identification of a physical and functional interaction between NPCs and the splicing machinery could be of conceptual interest in the NPC field, and more generally, in cell and genome biology, although it needs to be (i) carefully controlled and validated in view of the strong limitations mentioned above, and (ii) discussed in line with the known links between the nuclear basket and splicing quality control (see minor comments). This coupling would be particularly relevant for genes that have been shown to be positioned at NPCs during transcriptional activation, in line with the "gene gating" model mentioned by the authors.

We greatly appreciate these insightful comments and suggestions, and the time and effort that this reviewer invested in critical reading of our manuscript. We will certainly take the points into account as we revise the manuscript. Specifically, we will carefully address the concerns related to the NPC-splicing interaction, ensuring that the experimental validation is robust and well-controlled, and we will further discuss the connection between the nuclear basket and splicing quality control in the context of our findings.

Once again, thank you for your thoughtful and constructive feedback.

Reviewer #3

The authors discovered that the splicing machinery and nuclear baskets are sometimes in close proximity using Nup153 as a representative for the nuclear basket. They characterize this interaction using several different methods and propose that NUP153 is required to assemble the splicing machinery on genes that are transcribed in the nuclear periphery, which would supporting the gene gating model.

The manuscript is well written and structured and the experiments are carefully conducted and analyzed.

We thank the reviewer for the appreciation of our work. We have addressed all the major points here and will amend the manuscript text according to the suggestions.

Reviewer #3 (Significance (Required)):

The impression that I get from this manuscript is that we are looking at rather rare events with a small effect size. A definitive proof that the splicing machinery really assembles in the vicinity of NPCs docked via NUP153 is lacking. To assist in the revision process I will raise some questions to discuss but also propose some additional experiments to substantiate the claims.

1. It is not clear what NUP153 really binds to and which domain is important. The experiments shown suggest proximity and indirect interactions (Co-IPs), but it is not clear whether NUP153 binds to DNA, RNA or a specific splicing factor. The bioID experiment might label splicing factors because they are cargos that pass through the pores during import, or it labels splicing factors that remain bound to spliced mRNPs during mRNP export. For example DDX39b, also called UAP56 is an important subunit of the TREX complex and involved the final packaging of mRNPs at the NPC. In my opinion, this protein is not a good choice. Also, the negative control that was used in the experiments is a potassium channel in the plasmamembrane, which can exclude that signal occurs by chance. But it would have been better to use a nuclear protein as control to exclude these possibilities.

In line with a rare event, the Co-IP signals are very weak and barely higher than the GFP control. They should be repeated in the presence of RNase to confirm that this interaction occurs on nascent RNA during splicing and not e.g. to recycle or reroute splicing factors or during import.

We acknowledge all the points, some of them already brought to our attention by other reviewers, that we tried to address throughout our response here, and we will also incorporate our answers in the revised manuscript. In particular, we have already provided some evidence related to the role of DNA, RNA, as shown above in Figure R5 (Response to R2). We have also addressed the effect of nuclear-cytosolic transport by using IVM and described these results in Figure R6, showing that splicing factors interact with Nup153 even when cytosolic transport is blocked with IVM. We have also commented on the use of controls and on the additional control analysis that we performed (also mentioned in more detail in response to R2)

Moreover, we are also further trying to understand the binding of Nup153 to the splicing components. Intron Binding Complex, recently shown to be crucial for the activation of the spliceosome due to the activity of its helicase AQR (PMID: 37165190) is one of the protein complexes that we found bound to the NPC basket. We are interested in different functions of this helicase and have created the previously described mutant that has been shown to be defective in splicing. We have probed the interaction of Nup153 to this mutant, which we also characterized for its splicing inefficiency, and observed that Nup153 interacts with the splicing competent AQR, whereas the interaction with the splicing mutant seems to be less efficientThis set of additional data strengthens the bulk of data present here, details of which remain still to be further elucidated in an additional follow-up study.

__Figure R8 __Lysates from HEK 293T cells, transiently transfected with eGFP-NUP153 and AQR-His-FLAG or AQRK829A-His-FLAG were probed in co-IP experiments. Western blot of the co-IP performed with Dynabeads for His-tagged proteins; input (IN), unbound (U) and bound (B) fractions are shown. Densitometric analysis of the co-IP where eGFP-NUP153 intensity was quantified, bound fractions were normalized to input samples, and the results are expressed relative to the AQR-His-FLAG control (n = 2). pENTER is a control empty plasmid.

I do not understand why a NUP would be required to recruit or tether splicing factors to peripheral genes. Usually, splicing factors hitchhike on transcribing polymerase II or they are delivered by nuclear speckles which could happen also at the periphery. The authors should co-stain with a nuclear speckle marker to exclude this possibility.

__Figure R9 __STED microscopy resolves the position of sc-35, marker of splicing speckles with respect to AQR. Jurkat T cells were stained with anti AQR AB (rabbit) and anti - sc35 antibody (mouse) to probe the positioning of splicing speckles with respect to the splicing helicase AQR.

This is a very interesting remark, which we have addressed through co-staining experiments. Since the Nup153 antibody we used is anti-mouse, like SC-35, we instead co-stained SC-35 with AQR, a representative splicing factor. Our results show that a substantial number of AQR spots are detected away from nuclear speckles and near the nuclear rim, suggesting that a subset of splicing factors localize independently of speckles. While we have not directly stained the nuclear periphery, this pattern is consistent with the idea that splicing factors can be recruited outside of speckle-mediated delivery.

To further confirm that NPC-associated splicing does not rely on nuclear speckles, we are open to performing additional co-staining between Nup153 and SON, another speckle marker in a followup study. Furthermore, emerging evidence supports off-speckle splicing, particularly for genes with long introns and low GC content (PMID: 39413186, 38720076, 35182478, 22832277, 35182477). Our additional analysis (Figure R7) demonstrates that genes associated with Nup153 share characteristics with known off-speckle spliced genes, suggesting that these genes might be processed outside of speckles due to their transcription and splicing kinetics

What would be the advantage to splice in the vicinity of the pore? Given that genes with long introns take a long time to be transcribed, splicing would block the pores for hours and would prevent other activities. It would also be possible that splicing does occur at the periphery but NUP153 picks the mRNPs up at a later stage.

We thank the reviewer for these insightful comments. We would like to add here that there are numerous pores, according to our own estimations around 800 pores in Jurkat cells, which implies that there is also huge heterogeneity of the NPC which is at present largely unexplored. In yeast, some pores are basketless and the assembly of the basket is transcription dependent (PMID: 36220102) - suggesting that there’s more than one pore population. Similarly, looking into the statuses of genes that have been associated with pore, both polycomb-repressed and transcriptionally active genes have been found at pores- again pointing to an heterogeneity. This is a very interesting (and large) question that we pose ourselves and to the NPC field but not something we can address straight away.

One major drawback of the story is that the authors use very long-term depletion of NUP153 via shRNAs that will definitely screw up the import of many nuclear proteins; and a splicing inhibitor that has broad effects on nuclear architecture. Degron lines of Nup153 exist and should be used to substantiate at least some of the conclusions. Alternatively, a NUP153 mutant without zinc finger or IDR could be used to prevent DNA binding or basket association.

The reviewer is right, we have for over 2 years tried tirelessly to use the degron Nup153 system established in DLD-1 cells by Dasso’s Lab. This has been unsuccessful. Jurkats are difficult to transfect and more sensitive than other cell lines and they died with our trials. We have therefore used the shNup153 which has been used in X and Y and shown to not interfere with nucleocytoplasmic trafficking.

However, we understand the reviewers point and since then have tried to use a Nup153 mutant construct, containaing Nup153 N-terminus with and without a zinc finger (McKay et al 2009.), kindly sent by the Ulman Lab. Unfortunately, in our hands, the construct has low levels of GFP:Nup153 expression, not comparable to the ones we used in our coIP experiments and that would make conclusions hard.

We are now planning the cloning of our GFP:Nup153 construct to produce such plasmids,

N-terminus with/without the Zinc finger and use it in coIP experiments to understand the importance of the Zn finger domain in the splicing interactions.

Specific comments: Abstract: suggesting that a fraction of splicing occurs at the NPC. speckle-distant splicing events, it should be nuclear speckles, term not explained Line 20: Super enhancers not introduced

Line 39: Splicing and the different spliceosomal subcomplexes needs more explanation and introduction to understand the selection of proteins that were used in the study. Line 65: Choice of controls. LckN18 The genes should be written once in full and their choice should be explained better.

Line 123: The authors state: 'However, we detected a lower number of interacting spots between Nup98/DDX39b than with its Nup153 counterpart; a similar trend is followed between Nup98 and SF3A1 (Fig. S2g), suggesting that the interaction between splicing proteins and Nup98 might be further apart within the NPC structure.

A more likely explanation is that both proteins are shed from the mRNP at the basket as they are not shuttling with the mRNA and should not enter the pore.

Line 131: Mention right at the beginning of the sentence which splicing proteins were imaged here.

Line 141: Nuclear speckles are still not properly introduced. Why is SF3A1 not expected to be in nuclear speckles? It Should be co-stained for nuclear speckle markers. Lane 149: PlaB drastically changes the nuclear architecture. The reduced interactions can have also different reasons. The authors should image the distribution of the investigated factors in the presence of PlaB. Again the Co-IPs should be performed in the presence of RNase to confirm that the observed interactions depend on RNA. Line 181: The requirement of Nup153 to tether the splicing machinery to the NPC is not convincing from the presented data. The knockdown is way too long and the changes are tiny. Wouldn't it be better to use a Nup153 mutant without Zinc knuckle or IDR to show that now the splicing factors interactions are lost? Alternatively degron lines should be used.

Line 186: I do not understand the logic why NUP153 needs to bind to chromatin to fullfil its function in splicing. It could also bind to RNA with its zinc knuckle or IDRs. The authors should perform iCLIP or RIP to exclude this possibility? I also do not understand the logic to look to look for proximity to repressive LADs as a criterion, while investigating a function of NUP153 in splicing which requests actively transcribing genes. This has to be better motivated. Excluding the nucleoplasmic pool of NUP153 removes important data points that might be functionally relevant. Line 187: The entire paragraph on Dam-ID and all subsequent genome-wide analyses is way too densely written and hard to understand for non-experts. Analysis tools or thresholds are rarely given and it is unclear how the different data sets have been made and by who, what they mean and how they have been integrated. Chromatin patterns and expression profiles are very unspecific terms. Many of the used terms have not been introduced properly. The metaplots show very small differences. In the end I am not sure what we have learned from all the data integration. Is NUP153 bound to DNA, to nucleosomes, to nascent RNA or to splicing factors?

Lane 232: Unclear what NUP153 introns are. Is the entire gene where NUP153 binds to considered or only the intron with a NUP153 peak.

Lane 242: Again, the shRNA knockdown performed in this manuscript is way to long to observe a direct effect on splicing at the pore, which occurs at the level of minutes. Degrons should be used here to confirm this observation.

Line 249: More negative controls are needed for genes not bound by NUP153 for the splicing analysis. RNA-Seq analyzed for intron retention could be helpful.

All the text and specific comments will be addressed as suggested by the reviewer.

The experimental part to be included in the revised manuscript is explained in more detail above.

Reviewer expertise (keywords): nuclear pore complexes, nuclear organization, gene expression, mRNA biology

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The authors discovered that the splicing machinery and nuclear baskets are sometimes in close proximity using Nup153 as a representative for the nuclear basket. They characterize this interaction using several different methods and propose that NUP153 is required to assemble the splicing machinery on genes that are transcribed in the nuclear periphery, which would supporting the gene gating model.

The manuscript is well written and structured and the experiments are carefully conducted and analyzed.

Significance

The impression that I get from this manuscript is that we are looking at rather rare events with a small effect size. A definitive proof that the splicing machinery really assembles in the vicinity of NPCs docked via NUP153 is lacking. To assist in the revision process I will raise some questions to discuss but also propose some additional experiments to substantiate the claims.

1. It is not clear what NUP153 really binds to and which domain is important. The experiments shown suggest proximity and indirect interactions (Co-IPs), but it is not clear whether NUP153 binds to DNA, RNA or a specific splicing factor. The bioID experiment might label splicing factors because they are cargos that pass through the pores during import, or it labels splicing factors that remain bound to spliced mRNPs during mRNP export. For example DDX39b, also called UAP56 is an important subunit of the TREX complex and involved the final packaging of mRNPs at the NPC. In my opinion, this protein is not a good choice. Also, the negative control that was used in the experiments is a potassium channel in the plasmamembrane, which can exclude that signal occurs by chance. But it would have been better to use a nuclear protein as control to exclude these possibilities. In line with a rare event, the Co-IP signals are very weak and barely higher than the GFP control. They should be repeated in the presence of RNase to confirm that this interaction occurs on nascent RNA during splicing and not e.g. to recycle or reroute splicing factors or during import.
2. I do not understand why a NUP would be required to recruit or tether splicing factors to peripheral genes. Usually, splicing factors hitchhike on transcribing polymerase II or they are delivered by nuclear speckles which could happen also at the periphery. The authors should co-stain with a nuclear speckle marker to exlcude this possibility.
3. What would be the advantage to splice in the vicinity of the pore? Given that genes with long introns take a long time to be transcribed, splicing would block the pores for hours and would prevent other activities. It would also be possible that splicing does occur at the periphery but NUP153 picks the mRNPs up at a later stage.
4. One major drawback of the story is that the authors use very long-term depletion of NUP153 via shRNAs that will definitely screw up the import of many nuclear proteins; and a splicing inhibitor that has broad effects on nuclear architecture. Degron lines of Nup153 exist and should be used to substantiate at least some of the conclusions. Alternatively, a NUP153 mutant without zinc finger or IDR could be used to prevent DNA binding or basket association.

Specific comments:

Abstract: suggesting that a fraction of splicing occurs at the NPC. speckle-distant splicing events, it should be nuclear speckles, term not explained

Line 20: Super enhancers not introduced

Line 39: Splicing and the different spliceosomal subcomplexes needs more explanation and introduction to understand the selection of proteins that were used in the study.

Line 65: Choice of controls. LckN18 The genes should be written once in full and their choice should be explained better.

Line 123: The authors state: 'However, we detected a lower number of interacting spots between Nup98/DDX39b than with its Nup153 counterpart; a similar trend is followed between Nup98 and SF3A1 (Fig. S2g), suggesting that the interaction between splicing proteins and Nup98 might be further apart within the NPC structure. A more likely explanation is that both proteins are shed from the mRNP at the basket as they are not shuttling with the mRNA and should not enter the pore.

Line 131: Mention right at the beginning of the sentence which splicing proteins were imaged here.

Line 141: Nuclear speckles are still not properly introduced. Why is SF3A1 not expected to be in nuclear speckles? It Should be co-stained for nuclear speckle markers.

Lane 149: PlaB drastically changes the nuclear architecture. The reduced interactions can have also different reasons. The authors should image the distribution of the investigated factors in the presence of PlaB. Again the Co-IPs should be performed in the presence of RNase to confirm that the observed interactions depend on RNA.

Line 181: The requirement of Nup153 to tether the splicing machinery to the NPC is not convincing from the presented data. The knockdown is way too long and the changes are tiny. Wouldn't it be better to use a Nup153 mutant without Zinc knuckle or IDR to show that now the splicing factors interactions are lost? Alternatively degron lines should be used.

Line 186: I do not understand the logic why NUP153 needs to bind to chromatin to fullfil its function in splicing. It could also bind to RNA with its zinc knuckle or IDRs. The authors should perform iCLIP or RIP to exclude this possibility? I also do not understand the logic to look to look for proximity to repressive LADs as a criterion, while investigating a function of NUP153 in splicing which requests actively transcribing genes. This has to be better motivated. Excluding the nucleoplasmic pool of NUP153 removes important data points that might be functionally relevant.

Line 187: The entire paragraph on Dam-ID and all subsequent genome-wide analyses is way too densely written and hard to understand for non-experts. Analysis tools or thresholds are rarely given and it is unclear how the different data sets have been made and by who, what they mean and how they have been integrated. Chromatin patterns and expression profiles are very unspecific terms. Many of the used terms have not been introduced properly. The metaplots show very small differences. In the end I am not sure what we have learned from all the data integration. Is NUP153 bound to DNA, to nucleosomes, to nascent RNA or to splicing factors?

Lane 232: Unclear what NUP153 introns are. Is the entire gene where NUP153 binds to considered or only the intron with a NUP153 peak.

Lane 242: Again, the shRNA knockdown performed in this manuscript is way to long to observe a direct effect on splicing at the pore, which occurs at the level of minutes. Degrons should be used here to confirm this observation.

Line 249: More negative controls are needed for genes not bound by NUP153 for the splicing analysis. RNA-Seq analyzed for intron retention could be helpful.

There is quite some typos and missing words in the text.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this manuscript, using a combination of proximity labelling, immunoprecipitations and imaging, the authors report a physical interaction between splicing factors (SFs) and the nuclear basket of nuclear pore complexes (i.e. NUP153). Using DamID, they further identify a set of NUP153-bound genes characterized by long, GC-poor introns. Finally, based on molecular analyses for a set of candidate loci, they report that inactivation of NUP153 triggers a (modest) reduction of intron splicing, which may specifically affect NUP153-bound genes.

Major comments:

1. The data presented do not convincingly demonstrate a specific interaction between the NPC basket and the splicing machinery, mainly due to the lack of appropriate controls.
   • The BioID experiments (Fig. 1) lack proper controls. Proteins biotinylated by NUP-BirA fusions need to be compared with those modified upon expression of a control BirA protein, as has been done previously, especially when other NUPs were used as baits in BioID experiments (PMID: 24927568, to be cited). This control fusion should ideally be targeted to the same compartment (i.e. the nucleus or the nuclear side of the nuclear envelope).
   • NUP153 immunoprecipitates only very low levels of splicing factors, which are almost indistinguishable from those detected in control pull-downs (see for example AQR or XAB2 signals in blot images and error bars in the quantifications, Fig. 2a). In addition, in these analyses, the high/saturated signals do not allow comparison of the abundance of the proteins of interest in the input samples under different conditions. These limitations also apply to the interpretation of the changes in NUP153-SF association scored upon splicing inhibition (Fig. 3a), which also seems to affect SF abundance in inputs (e.g. AQR, SF3A1).
   • PLA experiments (e.g. Fig. 2b-d) also lack proper control. PLA has been shown to reveal artefactual signals for abundant proteins present in the same compartment (doi.org/10.1101/411355). Here, the chosen controls are inappropriate as the authors are probing interactions between NPC proteins (NUP153/TPR) and proteins restricted to a different nuclear compartment (e.g., nucleophosmin in the nucleolus). An abundant nucleoplasmic protein/epitope should be used as a control in these experiments. In addition, the authors need to show direct immunofluorescence images for each antibody used in PLA assays, in order to verify that the expression levels or localization of their targets are unchanged between conditions (e.g. upon PladB treatment, Fig. S3g, or NUP153 depletion, Fig. 3g).
   • The proximal localization of NUP153 and splicing factors in super resolution microscopy (Fig. 2e-f) is also not properly controlled. Would a control soluble, diffusible nucleoplasmic protein be detected in the vicinity of the NPC and sometimes colocalized with Nups?
   • Since NUP153 is located in the vicinity of peripheral genes (as also shown here through DamID), some of which contain introns, its association with the spliceosome could be indirect, i.e., mediated by DNA. Of note, the association of Mlp1 (the yeast ortholog of TPR) with the splicing factor SF1 has been shown to be mediated by RNAs (PMID:14718167). In order to assess these possibilities, the authors should perform their immunoprecipitation on extracts treated with benzonase, thus abrogating DNA- and RNA-dependent interactions.
2. NUP153 inactivation appears to have a modest effect on splicing (Fig. 5; S6), which is poorly characterized here. It is also unclear whether this effect is direct or caused by side consequences of the depletion of this nucleoporin (e.g., changes in nucleocytoplasmic exchanges or gene expression).
   • To confirm the specificity of the effects of NUP153 depletion on the splicing of NUP153-bound genes, the authors need to provide additional splicing measurements for several genes that are bound by NUP153 "in the nucleoplasm" (e.g. excluded from their analysis by the cutoff of proximity to LAD borders, line 192) and for other "non-NUP153" genes (beyond the unique control shown in Fig. S6a).
   • From the few examples provided, it is difficult to evaluate the type of splicing events affected by NUP153 inactivation. Are they uniquely intron retention events? The authors should analyze available RNA-seq data obtained from NUP153-depleted cells (PMID:32917881) to characterize the types of alternative splicing events that are impacted by NUP153.

Minor comments:

• Several studies have shown that the nuclear basket contributes to a splicing quality control process preventing the nuclear export of improperly spliced transcripts, both in yeast and mammalian cells (PMID:14718167, 19127978, 24452287, 25845599, 22253824, 22661231). These studies have to be mentioned and discussed here.
• Line 31: "movement of active genes towards the NPC would be favorable for their transcription and export ". Please rephrase: "...transcription and mRNA export".
• Line 163: "NUP153 plays a role in harboring splicing machinery". Please rephrase.
• Line 200-202: Fig. 4d and 4e (instead of S4d and S4e)
• Line 186 and beyond: All conclusions about NUP153-bound genes (e.g., "Majority of NUP153 bound genes are proximal to LADs and expressed") are not accurately phrased since the authors selected NUP153-bound genes with a cutoff of proximity to LAD borders. The conclusions are thus only valid for a subpopulation of NUP153-bound regions located in the vicinity of LADs.
• Line 292: "transport Nups less likely interact with splicing machinery". The term "transport Nup" is not correct. Does this mean "nuclear transport receptors"? Or "FG-Nups" (which interact with NTRs)?

Significance

It is increasingly recognized that NPCs are involved in a number of cellular processes beyond nucleo-cytoplasmic transport and, in particular, contribute to several genomic functions. In this context, the identification of a physical and functional interaction between NPCs and the splicing machinery could be of conceptual interest in the NPC field, and more generally, in cell and genome biology, although it needs to be (i) carefully controlled and validated in view of the strong limitations mentioned above, and (ii) discussed in line with the known links between the nuclear basket and splicing quality control (see minor comments). This coupling would be particularly relevant for genes that have been shown to be positioned at NPCs during transcriptional activation, in line with the "gene gating" model mentioned by the authors.

Reviewer expertise (keywords): nuclear pore complexes, nuclear organization, gene expression, mRNA biology

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The authors first use a Bio-ID approach to search for interactors of the basket proteins TPR and NUP153, identifying proteins involved in various nuclear process, including many splicing components, and confirm some of these interactions using IP and PLA assays. PLA experiments further suggest that these interactions occur primary at or close to the nuclear periphery. Moreover, inhibiting splicing, but not transcription, reduced these interactions. The authors then investigated the role of NUP153 in loading of the splicing machinery and found a lower association of the NUP98/SF3A1 but not AQR interaction (measured through PLA). Furthermore, DamID experiments identified NUP153 bound genes proximal to LAD domains that are actively transcribed, contain overall longer introns with low GC-content, and fall within a group of genes located at the outermost shell of the nucleus (when compared to previously published LaminI ID /PGseq data). Lastly, they interrogate whether depletion of NUP153 results in a splicing defect for NUP153 bound genes.

The authors identify many proteins in their BioID interaction screen, however, only a single nucleoporin (Nup35, an inner ring protein). Previous BioID studies have identified NUP153 in BioID experiments including proteins of the Y-complex (PMID: 24927568 and others). To ensure that the BioID experiments indeed probe for interactions of NUP152 and TPR at the NPC, the authors should include control experiments that show that their NUP153 and TPR-BirA fusions primarily localize to the NPC. If a significant fraction is not NPC bound, this has to be taken into account interpreting/discussing their data.<br /> The authors should be more precise when describing the role of the different splicing factor identified in the BioID screen and their function in specific steps of splicing, as this is important when claiming that they identify factors acting at all steps of splicing. For example, the authors describe DDX39b/UAP56 as an early splicing factor; DDX39b/UAP56 main role however seems to be in mRNA export and mRNP compaction. The authors might want to include this in the interpretation of their data.

Concerning PLA experiment controls, the authors perform TPR-PML as a negative control, however, no negative control for NUP153 is shown (Figure 2). Such a control should be added to allow evaluating the specificity of NUP153 PLA interactions, and/or discussed why this was not done.

Quantification of the distance of PLA-NUP153/TPR interactions show interactions mainly close to the nuclear periphery. The imaging data shown in Figure 2b indeed shows that TPR/NUP153 interactions are exclusively at the nuclear periphery, whereas NUP153/splicing factor interactions are sometimes at the edge of the DAPI signal, but mostly somehow internalized (Figure 2B, S2b). Quantification (Figure 2d) shows these distributions to be very similar, likely due to the way the quantification was performed / the bin size of plotting the relative distance of a spot to the nuclear periphery was chosen. Looking at the scale bar/nuclear size and the position of the PLA spots for the NUP153/splicing factors, it appears that spots are often hundreds of nanometers away from the periphery. As the nuclear basket is thought to reach only about 100nm onto the nuclear interior, the conclusion by the authors that these interactions occur at the NPC would not be consistent with the data. The authors should better incorporate this in their interpretation of the data. The conclusion ' Nup153 aids the loading of splicing machinery' is not sufficiently supported by the data. The authors observed a reduction in PLA signal for the NUP98-AQR interaction, but not the NUP98-SF3A1 (Figure 3g). Their conclusion has to reflect this discrepancy in their data. Moreover, the studies focus is to determine the role of NUP153/TPR in recruiting the splicing machinery to the NPC. As in the experiments the authors interrogate the interaction of only NUP98, who has to a large extend splicing factor interactions within the nuclear interior and not at the periphery, the relevance of the experiments in Figure 3 towards the main focus of the paper is unclear. When investigating the effect of NUP153 depletion on splicing, the authors observe a splicing phenotype for multiple NUP153 genes (Figure 5). The authors however show only a single negative control gene (CBX5). It would significantly strengthen their argument if the authors would investigate splicing defects of periphery located noneNUP153 bound genes as well as for genes located in the nuclear interior to better understand whether this splicing phenotype is indeed specific for NUP153 genes (at the nuclear periphery/NPC). The authors state in the text describing the SABER-FISH experiments in Figure 5f that 'were able to visualize the presence of a site of transcription where accumulation of these probes was close to the periphery for all except for GSTK1, which showed a wider nuclear distribution, similar to CBX5 control region not bound by Nup153'. However, their statement is not supported by the images shown in Fig 5f, which show TS in control cells in the nuclear interior. Also, a single cell but no quantification is shown. Moreover, what distance from the periphery is considered as close to the periphery is not defined (see also earlier comment on the question what should be considered a periphery and/or NPC association).

Limitation of the study does not discuss the limitations of the study but rather reads like the extension of the discussion. This section should be rewritten.

Minor comments:

Western in Figure 3c does not represent well the quantification in 3e.

Figure S3 is mislabelled (pannel h is panel g).

Significance

The manuscript interrogates an important question related to the role of the NPC in gene regulation, in particular how interaction of genes/pre-mRNAs with the NPC might stimulate expression of specific genes/mRNAs. Stimulating splicing would be one way that could contribute to efficient gene expression, and this is the question the authors address in this manuscript. This study is therefore important and relevant to a wide audience. However, as outlined in the section above, the conclusions drawn by the authors do not always reflect the experimental data, and it is therefore unclear whether the overall conclusion as stated in the title of the manuscript is valid. Moreover, conceptually, if intron containing genes are transcribed at or near nuclear pores, and splicing often occurs co-transcriptionally, it is to be expected to find splicing components close to nuclear pores. While it is relevant to show that this actually happens, and this is, at least in part, done by the authors. However, the experiments presented do not show that the splicing machinery actually actively docks to the NPC and is not just passively recruited close to NPCs because nascent pre-mRNAs are spliced where they are transcribed (the authors state in their title that the NUP153 docks the splicing machinery at the NPC). Showing this require identifying direct interactions between spliceosome components with NUP153/nuclear basket components to stimulate splicing at the NPC. If this would indeed be the case, these findings would describe a novel mechanistic step to stimulate efficient splicing and subsequently export of a selected set of NPC-associated genes. This would open other questions such as how to achieve specificity for only some pre-mRNAs/introns. While addressing this question is likely beyond the scope of this manuscript, the question whether the process described here is an active or passive process should be incorporated in the interpretation of the data.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reply to the Reviewers

I would like to thank the reviewers for their comments and interest in the manuscript and the study.

Reviewer #1

1. I would assume that there are RNA-seq and/or ChIP-seq data out there produced after knockdown of one or more of these DBPs that show directional positioning.

The directional positioning of CTCF-binding sites at chromatin interaction sites was analyzed by CRISPR experiment (Guo Y et al. Cell 2015). We found that the machine learning and statistical analysis showed the same directional bias of CTCF-binding motif sequence and RAD21-binding motif sequence at chromatin interaction sites as the experimental analysis of Guo Y et al. (lines 229-253, Figure 3b, c, d and Table 1). Since CTCF is involved in different biological functions (Braccioli L et al. Essays Biochem. 2019 ResearchGate webpage), the directional bias of binding sites may be reduced in all binding sites including those at chromatin interaction sites (lines 68-73). In our study, we investigated the DNA-binding sites of proteins using the ChIP-seq data of DNA-binding proteins and DNase-seq data. We also confirmed that the DNA-binding sites of SMC3 and RAD21, which tend to be found in chromatin loops with CTCF, also showed the same directional bias as CTCF by the computational analysis.

__2. Figure 6 should be expanded to incorporate analysis of DBPs not overlapping CTCF/cohesin in chromatin interaction data that is important and potentially more interesting than the simple DBPs enrichment reported in the present form of the figure. __

Following the reviewer's advice, I performed the same analysis with the DNA-binding sites that do no overlap with the DNA-binding sites of CTCF and cohesin (RAD21 and SMC3) (Fig. 6 and Supplementary Fig. 4). The result showed the same tendency in the distribution of DNA-binding sites. The height of a peak on the graph became lower for some DNA-binding proteins after removing the DNA-binding sites that overlapped with those of CTCF and cohesin. I have added the following sentence on lines 435 and 829: For the insulator-associated DBPs other than CTCF, RAD21, and SMC3, the DNA-binding sites that do not overlap with those of CTCF, RND21, and SMC3 were used to examine their distribution around interaction sites.

3. Critically, I would like to see use of Micro-C/Hi-C data and ChIP-seq from these factors, where insulation scores around their directionally-bound sites show some sort of an effect like that presumed by the authors - and many such datasets are publicly-available and can be put to good use here.

As suggested by the reviewer, I have added the insulator scores and boundary sites from the 4D nucleome data portal as tracks in the UCSC genome browser. The insulator scores seem to correspond to some extent to the H3K27me3 histone marks from ChIP-seq (Fig. 4a and Supplementary Fig. 3). We found that the DNA-binding sites of the insulator-associated DBPs were statistically overrepresented in the 5 kb boundary sites more than other DBPs (Fig. 4d). The direction of DNA-binding sites on the genome can be shown with different colors (e.g. red and green), but the directionality of insulator-associated DNA-binding sites is their overall tendency, and it may be difficult to notice the directionality from each binding site because the directionality may be weaker than that of CTCF, RAD21, and SMC3 as shown in Table 1 and Supplementary Table 2. We also observed the directional biases of CTCF, RAD21, and SMC3 by using Micro-C chromatin interaction data as we estimated, but the directionality was more apparent to distinguish the differences between the four directions of FR, RF, FF, and RR using CTCF-mediated ChIA-pet chromatin interaction data (lines 287 and 288).

 I found that the CTCF binding sites examined by a wet experiment in the previous study may not always overlap with the boundary sites of chromatin interactions from Micro-C assay (Guo Y et al. *Cell* 2015). The chromatin interaction data do not include all interactions due to the high sequencing cost of the assay, and include less long-range interactions due to distance bias. The number of the boundary sites may be smaller than that of CTCF binding sites acting as insulators and/or some of the CTCF binding sites may not be locate in the boundary sites. It may be difficult for the boundary location algorithm to identify a short boundary location. Due to the limitations of the chromatin interaction data, I planned to search for insulator-associated DNA-binding proteins without using chromatin interaction data in this study.

 I discussed other causes in lines 614-622: Another reason for the difference may be that boundary sites are more closely associated with topologically associated domains (TADs) of chromosome than are insulator sites. Boundary sites are regions identified based on the separation of numerous chromatin interactions. On the other hand, we found that the multiple DNA-binding sites of insulator-associated DNA-binding proteins were located close to each other at insulator sites and were associated with distinct nested and focal chromatin interactions, as reported by Micro-C assay. These interactions may be transient and relatively weak, such as tissue/cell type, conditional or lineage-specific interactions.

 Furthermore, I have added the statistical summary of the analysis in lines 372-395 as follows: Overall, among 20,837 DNA-binding sites of the 97 insulator-associated proteins found at insulator sites identified by H3K27me3 histone modification marks (type 1 insulator sites), 1,315 (6%) overlapped with 264 of 17,126 5kb long boundary sites, and 6,137 (29%) overlapped with 784 of 17,126 25kb long boundary sites in HFF cells. Among 5,205 DNA-binding sites of the 97 insulator-associated DNA-binding proteins found at insulator sites identified by H3K27me3 histone modification marks and transcribed regions (type 2 insulator sites), 383 (7%) overlapped with 74 of 17,126 5-kb long boundary sites, 1,901 (37%) overlapped with 306 of 17,126 25-kb long boundary sites. Although CTCF-binding sites separate active and repressive domains, the limited number of DNA-binding sites of insulator-associated proteins found at type 1 and 2 insulator sites overlapped boundary sites identified by chromatin interaction data. Furthermore, by analyzing the regulatory regions of genes, the DNA-binding sites of the 97 insulator-associated DNA-binding proteins were found (1) at the type 1 insulator sites (based on H3K27me3 marks) in the regulatory regions of 3,170 genes, (2) at the type 2 insulator sites (based on H3K27me3 marks and gene expression levels) in the regulatory regions of 1,044 genes, and (3) at insulator sites as boundary sites identified by chromatin interaction data in the regulatory regions of 6,275 genes. The boundary sites showed the highest number of overlaps with the DNA-binding sites. Comparing the insulator sites identified by (1) and (3), 1,212 (38%) genes have both types of insulator sites. Comparing the insulator sites between (2) and (3), 389 (37%) genes have both types of insulator sites. From the comparison of insulator and boundary sites, we found that (1) or (2) types of insulator sites overlapped or were close to boundary sites identified by chromatin interaction data.

4. The suggested alternative transcripts function, also highlighted in the manuscripts abstract, is only supported by visual inspection of a few cases for several putative DBPs. I believe this is insufficient to support what looks like one of the major claims of the paper when reading the abstract, and a more quantitative and genome-wide analysis must be adopted, although the authors mention it as just an 'observation'.

According to the reviewer's comment, I performed the genome-wide analysis of alternative transcripts where the DNA-binding sites of insulator-associated proteins are located near splicing sites. The DNA-binding sites of insulator-associated DNA-binding proteins were found within 200 bp centered on splice sites more significantly than the other DNA-binding proteins (Fig. 4e and Table 2). I have added the following sentences on lines 405 - 412: We performed the statistical test to estimate the enrichment of insulator-associated DNA-binding sites compared to the other DNA-binding proteins, and found that the insulator-associated DNA-binding sites were significantly more abundant at splice sites than the DNA-binding sites of the other proteins (Fig 4e and Table 2; Mann‒Whitney U test, p value 5. Figure 1 serves no purpose in my opinion and can be removed, while figures can generally be improved (e.g., the browser screenshots in Figs 4 and 5) for interpretability from readers outside the immediate research field.

I believe that the Figure 1 would help researchers in other fields who are not familiar with biological phenomena and functions to understand the study. More explanation has been included in the Figures and legends of Figs. 4 and 5 to help readers outside the immediate research field understand the figures.

6. Similarly, the text is rather convoluted at places and should be re-approached with more clarity for less specialized readers in mind.

Reviewer #2's comments would be related to this comment. I have introduced a more detailed explanation of the method in the Results section, as shown in the responses to Reviewer #2's comments.

Reviewer #2

1. Introduction, line 95: CTCF appears two times, it seems redundant.

On lines 91-93, I deleted the latter CTCF from the sentence "We examine the directional bias of DNA-binding sites of CTCF and insulator-associated DBPs, including those of known DBPs such as RAD21 and SMC3".

2. Introduction, lines 99-103: Please stress better the novelty of the work. What is the main focus? The new identified DPBs or their binding sites? What are the "novel structural and functional roles of DBPs" mentioned?

Although CTCF is known to be the main insulator protein in vertebrates, we found that 97 DNA-binding proteins including CTCF and cohesin are associated with insulator sites by modifying and developing a machine learning method to search for insulator-associated DNA-binding proteins. Most of the insulator-associated DNA-binding proteins showed the directional bias of DNA-binding motifs, suggesting that the directional bias is associated with the insulator.

 I have added the sentence in lines 96-99 as follows: Furthermore, statistical testing the contribution scores between the directional and non-directional DNA-binding sites of insulator-associated DBPs revealed that the directional sites contributed more significantly to the prediction of gene expression levels than the non-directional sites. I have revised the statement in lines 101-110 as follows: To validate these findings, we demonstrate that the DNA-binding sites of the identified insulator-associated DBPs are located within potential insulator sites, and some of the DNA-binding sites in the insulator site are found without the nearby DNA-binding sites of CTCF and cohesin. Homologous and heterologous insulator-insulator pairing interactions are orientation-dependent, as suggested by the insulator-pairing model based on experimental analysis in flies. Our method and analyses contribute to the identification of insulator- and chromatin-associated DNA-binding sites that influence EPIs and reveal novel functional roles and molecular mechanisms of DBPs associated with transcriptional condensation, phase separation and transcriptional regulation.

3. Results, line 111: How do the SNPs come into the procedure? From the figures it seems the input is ChIP-seq peaks of DNBPs around the TSS.

On lines 121-124, to explain the procedure for the SNP of an eQTL, I have added the sentence in the Methods: "If a DNA-binding site was located within a 100-bp region around a single-nucleotide polymorphism (SNP) of an eQTL, we assumed that the DNA-binding proteins regulated the expression of the transcript corresponding to the eQTL".

4. Again, are those SNPs coming from the different cell lines? Or are they from individuals w.r.t some reference genome? I suggest a general restructuring of this part to let the reader understand more easily. One option could be simplifying the details here or alternatively including all the necessary details.

On line 119, I have included the explanation of the eQTL dataset of GTEx v8 as follows: " The eQTL data were derived from the GTEx v8 dataset, after quality control, consisting of 838 donors and 17,382 samples from 52 tissues and two cell lines". On lines 681 and 865, I have added the filename of the eQTL data "(GTEx_Analysis_v8_eQTL.tar)".

5. Figure 1: panel a and b are misleading. Is the matrix in panel a equivalent to the matrix in panel b? If not please clarify why. Maybe in b it is included the info about the SNPs? And if yes, again, what is then difference with a.

The reviewer would mention Figure 2, not Figure 1. If so, the matrices in panels a and b in Figure 2 are equivalent. I have shown it in the figure: The same figure in panel a is rotated 90 degrees to the right. The green boxes in the matrix show the regions with the ChIP-seq peak of a DNA-binding protein overlapping with a SNP of an eQTL. I used eQTL data to associate a gene with a ChIP-seq peak that was more than 2 kb upstream and 1 kb downstream of a transcriptional start site of a gene. For each gene, the matrix was produced and the gene expression levels in cells were learned and predicted using the deep learning method. I have added the following sentences to explain the method in lines 133 - 139: Through the training, the tool learned to select the binding sites of DNA-binding proteins from ChIP-seq assays that were suitable for predicting gene expression levels in the cell types. The binding sites of a DNA-binding protein tend to be observed in common across multiple cell and tissue types. Therefore, ChIP-seq data and eQTL data in different cell and tissue types were used as input data for learning, and then the tool selected the data suitable for predicting gene expression levels in the cell types, even if the data were not obtained from the same cell types.

6. Line 386-388: could the author investigate in more detail this observation? Does it mean that loops driven by other DBPs independent of the known CTCF/Cohesin? Could the author provide examples of chromatin structural data e.g. MicroC?

As suggested by the reviewer, to help readers understand the observation, I have added Supplementary Fig. S4c to show the distribution of DNA-binding sites of "CTCF, RAD21, and SMC3" and "BACH2, FOS, ATF3, NFE2, and MAFK" around chromatin interaction sites. I have modified the following sentence to indicate the figure on line 501: Although a DNA-binding-site distribution pattern around chromatin interaction sites similar to those of CTCF, RAD21, and SMC3 was observed for DBPs such as BACH2, FOS, ATF3, NFE2, and MAFK, less than 1% of the DNA-binding sites of the latter set of DBPs colocalized with CTCF, RAD21, or SMC3 in a single bin (Fig. S4c).

 In Aljahani A et al. *Nature Communications* 2022, we find that depletion of cohesin causes a subtle reduction in longer-range enhancer-promoter interactions and that CTCF depletion can cause rewiring of regulatory contacts. Together, our data show that loop extrusion is not essential for enhancer-promoter interactions, but contributes to their robustness and specificity and to precise regulation of gene expression. Goel VY et al. *Nature Genetics* 2023 mentioned in the abstract: Microcompartments frequently connect enhancers and promoters and though loss of loop extrusion and inhibition of transcription disrupts some microcompartments, most are largely unaffected. These results suggested that chromatin loops can be driven by other DBPs independent of the known CTCF/Cohesin.

I added the following sentence on lines 569-577: The depletion of cohesin causes a subtle reduction in longer-range enhancer-promoter interactions and that CTCF depletion can cause rewiring of regulatory contacts. Another group reported that enhancer-promoter interactions and transcription are largely maintained upon depletion of CTCF, cohesin, WAPL or YY1. Instead, cohesin depletion decreased transcription factor binding to chromatin. Thus, cohesin may allow transcription factors to find and bind their targets more efficiently. Furthermore, the loop extrusion is not essential for enhancer-promoter interactions, but contributes to their robustness and specificity and to precise regulation of gene expression.

 FOXA1 pioneer factor functions as an initial chromatin-binding and chromatin-remodeling factor and has been reported to form biomolecular condensates (Ji D et al. *Molecular Cell* 2024). CTCF have also found to form transcriptional condensate and phase separation (Lee R et al. *Nucleic acids research* 2022). FOS was found to be an insulator-associated DNA-binding protein in this study and is potentially involved in chromatin remodeling, transcription condensation, and phase separation with the other factors such as BACH2, ATF3, NFE2 and MAFK. I have added the following sentence on line 556: FOXA1 pioneer factor functions as an initial chromatin-binding and chromatin-remodeling factor and has been reported to form biomolecular condensates.

7. In general, how the presented results are related to some models of chromatin architecture, e.g. loop extrusion, in which it is integrated convergent CTCF binding sites?

Goel VY et al. Nature Genetics 2023 identified highly nested and focal interactions through region capture Micro-C, which resemble fine-scale compartmental interactions and are termed microcompartments. In the section titled "Most microcompartments are robust to loss of loop extrusion," the researchers noted that a small proportion of interactions between CTCF and cohesin-bound sites exhibited significant reductions in strength when cohesin was depleted. In contrast, the majority of microcompartmental interactions remained largely unchanged under cohesin depletion. Our findings indicate that most P-P and E-P interactions, aside from a few CTCF and cohesin-bound enhancers and promoters, are likely facilitated by a compartmentalization mechanism that differs from loop extrusion. We suggest that nested, multiway, and focal microcompartments correspond to small, discrete A-compartments that arise through a compartmentalization process, potentially influenced by factors upstream of RNA Pol II initiation, such as transcription factors, co-factors, or active chromatin states. It follows that if active chromatin regions at microcompartment anchors exhibit selective "stickiness" with one another, they will tend to co-segregate, leading to the development of nested, focal interactions. This microphase separation, driven by preferential interactions among active loci within a block copolymer, may account for the striking interaction patterns we observe.

 The authors of the paper proposed several mechanisms potentially involved in microcompartments. These mechanisms may be involved in looping with insulator function. Another group reported that enhancer-promoter interactions and transcription are largely maintained upon depletion of CTCF, cohesin, WAPL or YY1. Instead, cohesin depletion decreased transcription factor binding to chromatin. Thus, cohesin may allow transcription factors to find and bind their targets more efficiently (Hsieh TS et al. *Nature Genetics* 2022). Among the identified insulator-associated DNA-binding proteins, Maz and MyoD1 form loops without CTCF (Xiao T et al. *Proc Natl Acad Sci USA* 2021 ; Ortabozkoyun H et al. *Nature genetics* 2022 ; Wang R et al. *Nature communications* 2022). I have added the following sentences on lines 571-575: Another group reported that enhancer-promoter interactions and transcription are largely maintained upon depletion of CTCF, cohesin, WAPL or YY1. Instead, cohesin depletion decreased transcription factor binding to chromatin. Thus, cohesin may allow transcription factors to find and bind their targets more efficiently. I have included the following explanation on lines 582-584: Maz and MyoD1 among the identified insulator-associated DNA-binding proteins form loops without CTCF.

 As for the directionality of CTCF, if chromatin loop anchors have some structural conformation, as shown in the paper entitled "The structural basis for cohesin-CTCF-anchored loops" (Li Y et al. *Nature* 2020), directional DNA binding would occur similarly to CTCF binding sites. Moreover, cohesin complexes that interact with convergent CTCF sites, that is, the N-terminus of CTCF, might be protected from WAPL, but those that interact with divergent CTCF sites, that is, the C-terminus of CTCF, might not be protected from WAPL, which could release cohesin from chromatin and thus disrupt cohesin-mediated chromatin loops (Davidson IF et al. *Nature Reviews Molecular Cell Biology* 2021). Regarding loop extrusion, the 'loop extrusion' hypothesis is motivated by in vitro observations. The experiment in yeast, in which cohesin variants that are unable to extrude DNA loops but retain the ability to topologically entrap DNA, suggested that in vivo chromatin loops are formed independently of loop extrusion. Instead, transcription promotes loop formation and acts as an extrinsic motor that extends these loops and defines their final positions (Guerin TM et al. *EMBO Journal* 2024). I have added the following sentences on lines 543-547: Cohesin complexes that interact with convergent CTCF sites, that is, the N-terminus of CTCF, might be protected from WAPL, but those that interact with divergent CTCF sites, that is, the C-terminus of CTCF, might not be protected from WAPL, which could release cohesin from chromatin and thus disrupt cohesin-mediated chromatin loops. I have included the following sentences on lines 577-582: The 'loop extrusion' hypothesis is motivated by in vitro observations. The experiment in yeast, in which cohesin variants that are unable to extrude DNA loops but retain the ability to topologically entrap DNA, suggested that in vivo chromatin loops are formed independently of loop extrusion. Instead, transcription promotes loop formation and acts as an extrinsic motor that extends these loops and defines their final positions.

 Another model for the regulation of gene expression by insulators is the boundary-pairing (insulator-pairing) model (Bing X et al. *Elife* 2024) (Ke W et al. *Elife* 2024) (Fujioka M et al. *PLoS Genetics* 2016). Molecules bound to insulators physically pair with their partners, either head-to-head or head-to-tail, with different degrees of specificity at the termini of TADs in flies. Although the experiments do not reveal how partners find each other, the mechanism unlikely requires loop extrusion. Homologous and heterologous insulator-insulator pairing interactions are central to the architectural functions of insulators. The manner of insulator-insulator interactions is orientation-dependent. I have summarized the model on lines 559-567: Other types of chromatin regulation are also expected to be related to the structural interactions of molecules. As the boundary-pairing (insulator-pairing) model, molecules bound to insulators physically pair with their partners, either head-to-head or head-to-tail, with different degrees of specificity at the termini of TADs in flies (Fig. 7). Although the experiments do not reveal how partners find each other, the mechanism unlikely requires loop extrusion. Homologous and heterologous insulator-insulator pairing interactions are central to the architectural functions of insulators. The manner of insulator-insulator interactions is orientation-dependent.

8. Do the authors think that the identified DBPs could work in that way as well?

The boundary-pairing (insulator-pairing) model would be applied to the insulator-associated DNA-binding proteins other than CTCF and cohesin that are involved in the loop extrusion mechanism (Bing X et al. Elife 2024) (Ke W et al. Elife 2024) (Fujioka M et al. PLoS Genetics 2016).

 Liquid-liquid phase separation was shown to occur through CTCF-mediated chromatin loops and to act as an insulator (Lee, R et al. *Nucleic Acids Research* 2022). Among the identified insulator-associated DNA-binding proteins, CEBPA has been found to form hubs that colocalize with transcriptional co-activators in a native cell context, which is associated with transcriptional condensate and phase separation (Christou-Kent M et al. *Cell Reports* 2023). The proposed microcompartment mechanisms are also associated with phase separation. Thus, the same or similar mechanisms are potentially associated with the insulator function of the identified DNA-binding proteins. I have included the following information on line 554: CEBPA in the identified insulator-associated DNA-binding proteins was also reported to be involved in transcriptional condensates and phase separation.

9. Also, can the authors comment about the mechanisms those newly identified DBPs mediate contacts by active processes or equilibrium processes?

Snead WT et al. Molecular Cell 2019 mentioned that protein post-transcriptional modifications (PTMs) facilitate the control of molecular valency and strength of protein-protein interactions. O-GlcNAcylation as a PTM inhibits CTCF binding to chromatin (Tang X et al. Nature Communications 2024). I found that the identified insulator-associated DNA-binding proteins tend to form a cluster at potential insulator sites (Supplementary Fig. 2d). These proteins may interact and actively regulate chromatin interactions, transcriptional condensation, and phase separation by PTMs. I have added the following explanation on lines 584-590: Furthermore, protein post-transcriptional modifications (PTMs) facilitate control over the molecular valency and strength of protein-protein interactions. O-GlcNAcylation as a PTM inhibits CTCF binding to chromatin. We found that the identified insulator-associated DNA-binding proteins tend to form a cluster at potential insulator sites (Fig. 4f and Supplementary Fig. 3c). These proteins may interact and actively regulate chromatin interactions, transcriptional condensation, and phase separation through PTMs.

10. Can the author provide some real examples along with published structural data (e.g. the mentioned micro-C data) to show the link between protein co-presence, directional bias and contact formation?

Structural molecular model of cohesin-CTCF-anchored loops has been published by Li Y et al. Nature 2020. The structural conformation of CTCF and cohesin in the loops would be the cause of the directional bias of CTCF binding sites, which I mentioned in lines 539 - 543 as follows: These results suggest that the directional bias of DNA-binding sites of insulator-associated DBPs may be involved in insulator function and chromatin regulation through structural interactions among DBPs, other proteins, DNAs, and RNAs. For example, the N-terminal amino acids of CTCF have been shown to interact with RAD21 in chromatin loops.

 To investigate the principles underlying the architectural functions of insulator-insulator pairing interactions, two insulators, Homie and Nhomie, flanking the *Drosophila even skipped *locus were analyzed. Pairing interactions between the transgene Homie and the eve locus are directional. The head-to-head pairing between the transgene and endogenous Homie matches the pattern of activation (Fujioka M et al. *PLoS Genetics* 2016).

Reviewer #3

Major Comments:

1. Some of these TFs do not have specific direct binding to DNA (P300, Cohesin). Since the authors are using binding motifs in their analysis workflow, I would remove those from the analysis.

When a protein complex binds to DNA, one protein of the complex binds to the DNA directory, and the other proteins may not bind to DNA. However, the DNA motif sequence bound by the protein may be registered as the DNA-binding motif of all the proteins in the complex. The molecular structure of the complex of CTCF and Cohesin showed that both CTCF and Cohesin bind to DNA (Li Y et al. Nature 2020). I think there is a possibility that if the molecular structure of a protein complex becomes available, the previous recognition of the DNA-binding ability of a protein may be changed. Therefore, I searched the Pfam database for 99 insulator-associated DNA-binding proteins identified in this study. I found that 97 are registered as DNA-binding proteins and/or have a known DNA-binding domain, and EP300 and SIN3A do not directory bind to DNA, which was also checked by Google search. I have added the following explanation in line 257 to indicate direct and indirect DNA-binding proteins: Among 99 insulator-associated DBPs, EP300 and SIN3A do not directory interact with DNA, and thus 97 insulator-associated DBPs directory bind to DNA. I have updated the sentence in line 20 of the Abstract as follows: We discovered 97 directional and minor nondirectional motifs in human fibroblast cells that corresponded to 23 DBPs related to insulator function, CTCF, and/or other types of chromosomal transcriptional regulation reported in previous studies.

2. I am not sure if I understood correctly, by why do the authors consider enhancers spanning 2Mb (200 bins of 10Kb around eSNPs)? This seems wrong. Enhancers are relatively small regions (100bp to 1Kb) and only a very small subset form super enhancers.

As the reviewer mentioned, I recognize enhancers are relatively small regions. In the paper, I intended to examine further upstream and downstream of promoter regions where enhancers are found. Therefore, I have modified the sentence in lines 929 - 931 of the Fig. 2 legend as follows: Enhancer-gene regulatory interaction regions consist of 200 bins of 10 kbp between -1 Mbp and 1 Mbp region from TSS, not including promoter.

3. I think the H3K27me3 analysis was very good, but I would have liked to see also constitutive heterochromatin as well, so maybe repeat the analysis for H3K9me3.

Following the reviewer's advice, I have added the ChIP-seq data of H3K9me3 as a truck of the UCSC Genome Browser. The distribution of H3K9me3 signal was different from that of H3K27me3 in some regions. I also found the insulator-associated DNA-binding sites close to the edges of H3K9me3 regions and took some screenshots of the UCSC Genome Browser of the regions around the sites in Supplementary Fig. 3b. I have modified the following sentence on lines 974 - 976 in the legend of Fig. 4: a Distribution of histone modification marks H3K27me3 (green color) and H3K9me3 (turquoise color) and transcript levels (pink color) in upstream and downstream regions of a potential insulator site (light orange color). I have also added the following result on lines 356 - 360: The same analysis was performed using H3K9me3 marks, instead of H3K27me3 (Fig. S3b). We found that the distribution of H3K9me3 signal was different from that of H3K27me3 in some regions, and discovered the insulator-associated DNA-binding sites close to the edges of H3K9me3 regions (Fig. S3b).

4. I was not sure I understood the analysis in Figure 6. The binding site is with 500bp of the interaction site, but micro-C interactions are at best at 1Kb resolution. They say they chose the centre of the interaction site, but we don't know exactly where there is the actual interaction. Also, it is not clear what they measure. Is it the number of binding sites of a specific or multiple DBP insulator proteins at a specific distance from this midpoint that they recover in all chromatin loops? Maybe I am missing something. This analysis was not very clear.

The resolution of the Micro-C assay is considered to be 100 bp and above, as the human nucleome core particle contains 145 bp (and 193 bp with linker) of DNA. However, internucleosomal DNA is cleaved by endonuclease into fragments of multiples of 10 nucleotides (Pospelov VA et al. Nucleic Acids Research 1979). Highly nested focal interactions were observed (Goel VY et al. Nature Genetics 2023). Base pair resolution was reported using Micro Capture-C (Hua P et al. Nature 2021). Sub-kilobase (20 bp resolution) chromatin topology was reported using an MNase-based chromosome conformation capture (3C) approach (Aljahani A et al. Nature Communications 2022). On the other hand, Hi-C data was analyzed at 1 kb resolution. (Gu H et al. bioRxiv 2021). If the resolution of Micro-C interactions is at best at 1 kb, the binding sites of a DNA-binding protein will not show a peak around the center of the genomic locations of interaction edges. Each panel shows the number of binding sites of a specific DNA-binding protein at a specific distance from the midpoint of all chromatin interaction edges. I have modified and added the following sentences in lines 593-597: High-resolution chromatin interaction data from a Micro-C assay indicated that most of the predicted insulator-associated DBPs showed DNA-binding-site distribution peaks around chromatin interaction sites, suggesting that these DBPs are involved in chromatin interactions and that the chromatin interaction data has a high degree of resolution. Base pair resolution was reported using Micro Capture-C.

Minor Comments:

1. PIQ does not consider TF concentration. Other methods do that and show that TF concentration improves predictions (e.g., ____https://www.biorxiv.org/content/10.1101/2023.07.15.549134v2____or ____https://pubmed.ncbi.nlm.nih.gov/37486787____/). The authors should discuss how that would impact their results.

The directional bias of CTCF binding sites was identified by ChIA-pet interactions of CTCF binding sites. The analysis of the contribution scores of DNA-binding sites of proteins considering the binding sites of CTCF as an insulator showed the same tendency of directional bias of CTCF binding sites. In the analysis, to remove the false-positive prediction of DNA-binding sites, I used the binding sites that overlapped with a ChIP-seq peak of the DNA-binding protein. This result suggests that the DNA-binding sites of CTCF obtained by the current analysis have sufficient quality. Therefore, if the accuracy of prediction of DNA-binding sites is improved, although the number of DNA-binding sites may be different, the overall tendency of the directionality of DNA-binding sites will not change and the results of this study will not change significantly.

 As for the first reference in the reviewer's comment, chromatin interaction data from Micro-C assay does not include all chromatin interactions in a cell or tissue, because it is expensive to cover all interactions. Therefore, it would be difficult to predict all chromatin interactions based on machine learning. As for the second reference in the reviewer's comment, pioneer factors such as FOXA are known to bind to closed chromatin regions, but transcription factors and DNA-binding proteins involved in chromatin interactions and insulators generally bind to open chromatin regions. The search for the DNA-binding motifs is not required in closed chromatin regions.

2. DeepLIFT is a good approach to interpret complex structures of CNN, but is not truly explainable AI. I think the authors should acknowledge this.

In the DeepLIFT paper, the authors explain that DeepLIFT is a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input (Shrikumar A et al. ICML 2017). DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. DeepLIFT calculates a metric to measure the difference between an input and the reference of the input.

 Truly explainable AI would be able to find cause and reason, and to make choices and decisions like humans. DeepLIFT does not perform causal inferences. I did not use the term "Explainable AI" in our manuscript, but I briefly explained it in Discussion. I have added the following explanation in lines 623-628: AI (Artificial Intelligence) is considered as a black box, since the reason and cause of prediction are difficult to know. To solve this issue, tools and methods have been developed to know the reason and cause. These technologies are called Explainable AI. DeepLIFT is considered to be a tool for Explainable AI. However, DeepLIFT does not answer the reason and cause for a prediction. It calculates scores representing the contribution of the input data to the prediction.

 Furthermore, to improve the readability of the manuscript, I have included the following explanation in lines 159-165: we computed DeepLIFT scores of the input data (i.e., each binding site of the ChIP-seq data of DNA-binding proteins) in the deep leaning analysis on gene expression levels. DeepLIFT compares the importance of each input for predicting gene expression levels to its 'reference or background level' and assigns contribution scores according to the difference. DeepLIFT calculates a metric to measure the difference between an input and the reference of the input.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Osato and Hamada propose a systematic approach to identify DNA binding proteins that display directional binding. They used a modified Deep Learning method (DEcode) to investigate binding profiles of 1356 DBP from GTRD database at promoters (30 of 100bp bins around TSS) and enhancers (200 bins of 10Kb around eSNPs) and use this to predict expression of 25,071 genes in Fibroblasts, Monocytes, HMEC and NPC. This method achieves a good prediction power (Spearman correlation between predicted and actual expression of 0.74). They then use PIQ, and overlap predicted binding sites with actual ChIP-seq data to investigate the motifs of TFs that are controlling gene expression. They find 99 insulator proteins showing either a specific directional bias or minor non-directional bias, corresponding to 23 DBP previously reported to have insulator function. Of the 23 proteins they identify as regulating enhancer promoter interactions, 13 are associated with CTCF. They also show that there are significantly more insulator proteins binding sites at borders of polycomb domains, transcriptionally active or boundary regions based on chromatin interactions than other proteins.

Major Comments:

1. Some of these TFs do not have specific direct binding to DNA (P300, Cohesin). Since the authors are using binding motifs in their analysis workflow, I would remove those from the analysis.
2. I am not sure if I understood correctly, by why do the authors consider enhancers spanning 2Mb (200 bins of 10Kb around eSNPs)? This seems wrong. Enhancers are relatively small regions (100bp to 1Kb) and only a very small subset form super enhancers.
3. I think the H3K27me3 analysis was very good, but I would have liked to see also constitutive heterochromatin as well, so maybe repeat the analysis for H3K9me3.
4. I was not sure I understood the analysis in Figure 6. The binding site is with 500bp of the interaction site, but micro-C interactions are at best at 1Kb resolution. They say they chose the centre of the interaction site, but we don't know exactly where there is the actual interaction. Also, it is not clear what they measure. Is it the number of binding sites of a specific or multiple DBP insulator proteins at a specific distance from this midpoint that they recover in all chromatin loops? Maybe I am missing something. This analysis was not very clear.

Minor comments:

1. PIQ does not consider TF concentration. Other methods do that and show that TF concentration improves predictions (e.g., https://www.biorxiv.org/content/10.1101/2023.07.15.549134v2 or https://pubmed.ncbi.nlm.nih.gov/37486787/). The authors should discuss how that would impact their results.
2. DeepLIFT is a good approach to interpret complex structures of CNN, but is not truly explainable AI. I think the authors should acknowledge this.

Referee Cross-Commenting

I would like to mention that I agree with the comments of reviewers 1 and 2.

Significance

General assessment:

This is the first study to my knowledge that attempts to use Deep Learning to identify insulators and directional biases in binding. One of the limitations is that no additional methods were used to show that these DBP have directional binding bias. It is not necessarily to employ additional methods, but it would definitely strengthen the paper.

Advancements:

This is a useful catalogue of potential DNA binding proteins of interest, beyond just CTCF. Some known TFs are there, but also new ones are found.

Audience:

Basic research mainly, with particular focus on chromatin conformation and TF binding fields.

My expertise:

ML/AI methods in genomics, TF binding models, epigenetics and 3D chromatin interactions.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this work, the authors describe a deep learning computational tool to identity binding motifs of DNA binding proteins associated to insulators that led to the discovery of 99 motifs related to insulation. This is in turn related to chromatin architecture and highlight the importance of directional bias in order to form chromatin loops.

In general, there are some aspects to be clarified and better explored to make stronger conclusions. In particular, there are some aspects to clarify in the text about the Machine Learning procedure (see my points below). In addition, I have some general questions about the biological implications of the discussed findings, listed in detail in the following list.

Also, I encourage the authors to integrate the current presentation of the data with other (published) data about chromatin architecture, to make more robust the claims and go deeper into the biological implications of the current work. Se my list below.

It follows a specific list of relevant points to be addressed:

Specific points:

1. Introduction, line 95: CTCF appears two times, it seems redundant;
2. Introduction, lines 99-103: Please stress better the novelty of the work. What is the main focus? The new identified DPBs or their binding sites? What are the "novel structural and functional roles of DBPs" mentioned?
3. Results, line 111: How do the SNPs come into the procedure? From the figures it seems the input is ChIP-seq peaks of DNBPs around the TSS;
4. Again, are those SNPs coming from the different cell lines? Or are they from individuals w.r.t some reference genome? I suggest a general restructuring of this part to let the reader understand more easily. One option could be simplifying the details here or alternatively including all the necessary details;
5. Figure 1: panel a and b are misleading. Is the matrix in panel a equivalent to the matrix in panel b? If not please clarify why. Maybe in b it is included the info about the SNPs? And if yes, again, what is then difference with a.
6. Line 386-388: could the author investigate in more detail this observation? Does it mean that loops driven by other DBPs independent of the known CTCF/Cohesin? Could the author provide examples of chromatin structural data e.g. MicroC?
7. In general, how the presented results are related to some models of chromatin architecture, e.g. loop extrusion, in which it is integrated convergent CTCF binding sites?
8. Do the authors think that the identified DBPs could work in that way as well?
9. Also, can the authors comment about the mechanisms those newly identified DBPs mediate contacts by active processes or equilibrium processes?
10. Can the author provide some real examples along with published structural data (e.g. the mentioned micro-C data) to show the link between protein co-presence, directional bias and contact formation?

Significance

In this work, the authors describe a deep learning computational tool to identity binding motifs of DNA binding proteins associated to insulators that led to the discovery of 99 motifs related to insulation. This is in turn related to chromatin architecture and highlight the importance of directional bias in order to form chromatin loops.

In general, chromatin organization is an important topic in the context of a constantly expanding research field. Therefore, the work is timely and could be useful for the community. The paper appears overall well written and the figures look clear and of good quality. Nevertheless, there are some aspects to be clarified and better explored to make stronger conclusions. In particular, there are some aspects to clarify in the text about the Machine Learning procedure (see list of specific points). In addition, I have some general questions about the biological implications of the discussed findings, listed in detail in the above reported points.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The study by Osato and Hamada aims at computationally identifying a set of novel putative insulator-associated DNA binding proteins (DBPs) via estimation of their contribution to the expression of genes in the same chromosome region of their binding sites (+- 1Mbp from TSS). To achieve this, the authors leverage a deep learning architecture already published via which ChIP-seq peaks of DBPs in the TSS of a given gene are used to predict its expression level in four human cell lines.

Building on this, the authors used another tool called DeepLIFT to evaluate the weight of each DBP binding site on the final gene expression value. Hence they made the assumption that if a given DBP had an insulator function they could restrict the prediction of the gene's expression to the region included between pairs of that DBP binding sites, and evaluate the pair's motif directionality bias in the distribution of weights. They exemplify their approach's validity by the fact that they can predict the known directionality bias of CTCF/cohesin-bound sites as the highest of the lot, with the F-R orientation of the pairs the most enriched, recapitulating what already known in literature: i.e., that F-R chromatin interaction peaks are the most enriched. In addition, they find several new DBPs showing significant directionality bias; hence they could be candidates for insulation activity. They then provide correlation between these putative insulator binding sites and sites of transition between euchromatin and heterochromatin by independently using histone mark and gene expression datasets. This, of course, is not surprising because (a) there is insulation between regions with heterotypic chromatin identities, and (b) it was already known from the first papers describing insulated chromatin domains that their boundaries were well-enriched for active transcription and transcriptional regulators (e.g., Dixon et al, Nature 2012).

Finally, they use chromatin interaction (looping) sites to check the overlap between CTCF and all other DBPs and define a subset of putative insulator DBPs not overlapping CTCF peaks, suggesting potentially new insulatory mechanisms. These factors were all known transcriptional activators, but this part of the findings carry most of the novelty in the work and have the potential of opening up new directions for research in chromatin organization.

Overall, the methodology applied here is adequate, clear, and reproducible. The major issue, in our view, is that the entire manuscript's findings relies on the usage of deepLIFT, a tool which was not benchmarked previously or by the current study. In fact, deepLIFT is public as regards its code, and also appears as a preprint from 2017 on biorXiv and published in the Proceedings of Machine Learning Research conference. Also, this key tool was developed by the Kundaje lab (who produce high quality alogrithms), and not by the authors. Therefore, the manuscript is predominantly based on the execution of existing workflows to publicly-available data. This does not take anything away from the interesting question posed here, but at the same time does not provide the community with any new algorithm/workflow.

Finally, although I appreciate that the authors are purely computational and have likely no capacity for experimental validation of their claims of new DBPs having insulator roles, I would assume that there are RNA-seq and/or ChIP-seq data out there produced after knockdown of one or more of these DBPs that show directional positioning. Using this kind of data, effects on gene expression can at least be tested in regard to the authors' predictions. Moreover, in terms of validation, Figure 6 should be expanded to incorporate analysis of DBPs not overlapping CTCF/cohesin in chromatin interaction data that is important and potentially more interesting than the simple DBPs enrichment reported in the present form of the figure. Critically, I would like to see use of Micro-C/Hi-C data and ChIP-seq from these factors, where insulation scores around their directionally-bound sites show some sort of an effect like that presumed by the authors - and many such datasets are publicly-available and can be put to good use here.

As secondary issues, we would point out that:

• The suggested alternative transcripts function, also highlighted in the manuscript;s abstract, is only supported by visual inspection of a few cases for several putative DBPs. I believe this is insufficient to support what looks like one of the major claims of the paper when reading the abstract, and a more quantitative and genome-wide analysis must be adopted, although the authors mention it as just an 'observation'.
• Figure 1 serves no purpose in my opinion and can be removed, while figures can generally be improved (e.g., the browser screenshots in Figs 4 and 5) for interpretability from readers outside the immediate research field.
• Similarly, the text is rather convoluted at places and should be re-approached with more clarity for less specialized readers in mind.

Significance

The scientific novelty of the work lies primarily in the identification of a set of DBPs that are proposed to confer insulator activity genome-wide. This has been long sought after in human data (whilst it is well understood and defined in Drosophila). The authors produce a quantitative ranking of the putative insulation effect of these DBPs and, most importantly, go on to identify a smaller subset that are apparently non-overlapping with anchors of CTCF-cohesin loop anchors; the presence of strong motif orientation biases in many DBPs can also be of broad interest, especially those that cannot be trivially ascribable to the loop extrusion process.

However, although these findings open the way for speculation on multiple insulation mechanisms via proteins with multiple regulatory functions, the manuscript provide no experimental or computational means to test the proposed roles of these DBPs - and, as such, this limits the potential impact of the work and mostly targets researchers in the field of genome organization that can test these findings. Having said this, if validated, this work can significantly broaden our understanding of how chromatin is organized in 3D nuclear space.

I typically identify myself to the authors: A. Papantonis, expertise in 3D genome architecture, chromatin biology, and genomics/bioinformatics.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the reviewers for their thoughtful comments

---

Reviewer #1 (Evidence, reproducibility and clarity):

SUMMARY: The manuscript is well written, with excellent explanation and documentation of experimental approaches. All conclusions are well supported by the data. The discussion is balanced and appropriate. The data, including images and movies, are of high quality and beautifully presented. The experimental design and analysis, including quantification of parameters in the images, is rigorous. Additional rigor is provided by comparing different cell types. The rapalog and iLID dimerization strategies have been described previously, as has their use to recruit kinesin motors to membranous organelles. However, this is the first application of these strategies to recruit motors to intermediate filaments. The evidence that vimentin filaments can be redistributed locally is clear and convincing and offers appealing potential for future experimentation. The redistribution was not fully reversible in all cells, but this is not surprising given the entanglement that must result from the action of motors along the length of these long flexible polymers.

In terms of the biology of intermediate filaments, the authors show that vimentin redistribution had negligible effect on microtubule or F-actin organization, cell area, or the number of focal adhesions. Depletion of vimentin filaments locally reduced cell stiffness. Both ER and mitochondria segregated with vimentin filaments, but not lysosomes. These findings are consistent with published reports (e.g. comparing vimentin null and wildtype cell lines), but the acute and reversible nature of the motor recruitment strategy is a more elegant experimental approach, and the selectivity of the observed effects is evidence of its specificity. It is interesting that the ER network segregated with vimentin even in the absence of RNF26. While this is not explored further, it points to the potential power of this motor recruitment strategy for future studies on intermediate filament interactions.

• *

The following are some major and minor issues, which should all be easy for the authors to address.

MAJOR COMMENTS:

• • Fig. S1 shows that the Vim-mCherry-FKBP construct coassembles with endogenous vimentin, but similar data for the iLID constructs appears to be lacking. I would like to see data demonstrating the incorporation of the Vim-mCherry-SspB constructs into the vimentin filaments. This should include high magnification images of single filaments in the cytoplasm of the cells.*
• *

Response:

We have included a new Figure 2D, which illustrates the incorporation of the vimentin-mCherry-SspB construct into the vimentin network stained for endogenous vimentin.

• • The authors do not discuss the density of motor recruitment along the filaments. To address this, I'd like to see images showing the extent of recruitment of motors to the filaments using the rapalog and LID strategies. This should include high magnification images of single filaments in the cytoplasm of the cells.*
• *

Response:

We have included new Figure S1B,C and Figure S2A, which illustrate the recruitment of kinesin motors to vimentin filaments upon induction with rapalog or light, respectively, by using super-resolution imaging with an Airyscan microscope. The motors were stained with antibodies against GFP. These data are discussed in the text, lines 126-132 and 165-168.

• • For the experiments on vimentin and keratin organization, the authors do not explain that these proteins form distinct networks and do not coassemble. The authors should show this in the cell types examined. This should also be explained explicitly in the body of the manuscript, though the data could be placed in the supplementary data. This is important because many intermediate filaments can coassemble freely, and coassembled proteins would be expected to segregate together.*
• *

Response:

To address this important comment, we have now included images of vimentin and keratin in the three studied cell types using super-resolution imaging, both for cells expressing vimentin constructs (updated Figure 5) and endogenous filament staining in untransfected cells (updated Figure S4). These images illustrate that vimentin and keratin mostly form distinct filaments in HeLa cells. However, we do observe some degree of co-assembly of vimentin and keratin in COS-7 and U2OS cells. We were really surprised by this observation as, to our knowledge, it has not been clearly documented in the literature. These data help to explain why vimentin pulling causes keratin co-clustering in COS-7 and U2OS cells. We note that in a study where kinesin-1 mediated transport of vimentin and keratin has been previously investigated by the Gelfand lab in RPE1 cells, the two networks also appear to overlap quite strongly (Robert et al, 2019, FASEB J). Since no super-resolution microscopy was performed in that study, potential co-assembly of keratin and vimentin filaments was not discussed. Colocalization and coprecipitation of vimentin and keratin have been also described by Velez-delValle et al. in epithelial cells (Sci Rep 2016). Cell type-specific co-assembly of keratin and vimentin would require more investigation, and we make no strong conclusions about it, but we think that our data illustrate the usefulness of our methodology to address the co-dependence of different types of intermediate filaments.

MINOR COMMENTS:

• • The authors refer to selecting cells within an "optimized expression range" for their transiently expressed recombinant proteins. They should state the proportion of the cells that met this criterion in their transient transfection experiments as this is important information for other researchers that might wish to use this approach in their own studies*. Response:

These numbers are now included in lines 137 -142 and 173-176 of the revised paper. For the FRB-FKP system, ~50% of transfected cells could be used for analysis, for the light-induced system, ~40% were in the optimal range.

• • In Fig. 1F there should be a statistical comparison between cells transfected with the Kin14 construct and control (untransfected) cells in the absence of rapalog*
• *

Response:

This comparison has been added.

• • In Fig. 1G there should be a statistical comparison between cells expressing Kin14 and KIF5A in the absence of rapalog.*
• *

Response:

This comparison has been added.

• • The depletion of the ER network in the cell periphery is not evident in Fig. 7B, though the perinuclear accumulation is evident. Perhaps the authors could select another example or explain to the reader what exactly to look for in these images.*
• *

Response:

We note that Figure 7B is a line scan of the image shown in Figure 7A. We assume that the reviewer meant Figure 7C, which is discussed in detail below.

• • In Fig. 7C, the intensity of the mCherry declines markedly over time. This is presumably due to photobleaching but should be explained in the legend.*
• *

Response:

We have now improved Figure 7 by adding additional quantifications of ER and vimentin intensity and distribution in Figures 7D and E. We also extended the corresponding text (lines 288-297), which now reads; “Using the optogenetic tool, we observed that ER sheets and matrices, but not tubules, were pulled along with vimentin, confirming their previously described direct connections (Cremer et al., 2023) (black arrows, Figure 7C; Video S5). Most of the vimentin and ER repositioning occurred within approximately 10 minutes (Figure 7C, D, Video S5). While initially this resulted in a sparser tubular ER network at the cell periphery, over time, the network became denser, with smaller polygonal structures. This effect could also be observed in the ratio of perinuclear to peripheral intensity, where a subset of ER initially follows vimentin to the perinuclear region but then redistributes again towards the cell periphery (Figure 7D). It should be noted that while photobleaching of the ER channel was negligible, there was a 40% reduction in total Vim-mCh-SspB intensity over the course of the experiment due to photobleaching (Figure 7E).”

• *

Reviewer #1 (Significance):

SUMMARY: The authors show that chemical-induced and light-induced dimerization strategies can be used to recruit microtubule motors to vimentin filaments, allowing rapid and reversible experimental manipulation of vimentin filament organization either locally or globally in cells. These strategies provide an experimental approach for investigating the physical interaction of intermediate filaments with organelles and other cytoskeletal component, as well as a method for probing the role of intermediate filaments in cell mechanics, cytoskeletal dynamics, etc. This is a technical improvement over previous experimental strategies, which have relied largely on chronic manipulation such as global disassembly or genetic deletion of intermediate filaments, e.g. comparison of vimentin null and wild type cells.

The principal weakness of this study is that it offers limited insight into intermediate filament biology. As such, it might be most appropriate for a tools or techniques section of a journal. The dimerization strategies have been reported previously, so that is not new, but the application to intermediate filaments is novel.

• *

Response:

We agree that our paper is primarily of technical nature and thus would be most appropriate for the tools and techniques section of a journal. We also agree that we used motor recruitment strategies that we and others have employed previously. However, we would like to emphasize that the demonstration that the tools work very well for intermediate filaments is entirely novel, as are the observations that these tools can be used to very rapidly alter cell stiffness or probe the links between intermediate filaments and organelles. Most importantly, the intermediate filament field currently lacks rapid specific manipulation strategies, and our tools will allow revisiting many important pending questions in the field. For example, they will allow to distinguish short-term and direct effects of intermediate filaments on cell polarity, adhesion and migration from their function in signaling and gene expression. We also report some new biology, such as evidence of some degree of co-assembly of vimentin and keratin.

AUDIENCE: This paper will be of interest to cell biologists who study cytoskeletal interactions, particularly the interaction of intermediate filaments with other cellular organelles or cytoskeletal polymers, or the role of intermediate filaments in cellular mechanics.

REVIEWER EXPERTISE: This reviewer has expertise on the cytoskeleton, cytoskeletal dynamics, and intracellular transport including intermediate filament biology.

__ __

---

Reviewer #2 (Evidence, reproducibility and clarity):

Summary: The manuscript presents a novel methodology for acute manipulation of vimentin intermediate filaments (IFs) using chemical genetic and optogenetic tools. By recruiting microtubule-based motors to vimentin via inducible dimerization systems, the authors achieve precise temporal and spatial control over vimentin distribution. Apart from the significant advancement in terms of methods development, key findings include:

* Vimentin's role in organelle positioning: Mitochondria and ER are repositioned with vimentin, while lysosomes are less dependent on its organization.

* Cytoskeletal interactions: Vimentin clustering minimally impacts actin and microtubule networks in the short term.

* Cell stiffness: Vimentin repositioning reduces cell stiffness, indicating its significant role in cellular mechanics.

* Cell-type-specific keratin interactions: The study highlights diverse interactions between vimentin and keratin-8 across cell lines.

The study demonstrates methodological advancements enabling rapid vimentin manipulation and provides insights into vimentin's interactions with cellular structures.

A major shortcoming is the unclear narrative, what do the authors want to present? This aspect requires significant attention.

Response:

By “unclear narrative” the reviewer meant that we should have provided a more balanced discussion of the insights that could be obtained using our new method compared to previously published literature, and we have modified our narrative accordingly.

General Comments and Overall Assessment

The manuscript represents an interesting contribution to the cytoskeletal field, addressing limitations of long-term perturbation methods. The tools developed are innovative, allowing controlled and reversible vimentin reorganization with minimal off-target effects. The findings are robust and provide important insights into the role of vimentin in cellular mechanics and organelle positioning.

Strengths:

Methodological novelty with broad applicability - this is the most exciting aspect.

Comprehensive validation of the tools in multiple cell lines.

Clear differentiation between vimentin's short- and long-term roles.

Addressing gaps in understanding vimentin-organelle interactions.

Limitations:

* The manuscript is a little bit all over the place. While the method development is clear, the manuscript makes claims way beyond the method development. The message and narrative needs to be improved, and in the respect the whole structure needs an overhaul.

Response:

We have carefully modified the manuscript to avoid the impression that we make any claims that go beyond the immediate and quantifiable effects of vimentin repositioning on different cellular structures.

* Unclear how much the differences in expression levels impact results and reproducibility.

Response:

Quantifications of expression levels and their discussion are included in Figures 1G-I, 2G-H, S2B and lines 137-142 and 173-176.

* Would be good to discuss some findings that are specific to a given experimental cell line. How generalizable are these results?

Response:

Cell line-specific findings concerned mostly the co-displacement of keratin together with vimentin, which occurred in COS-7 and U2OS cells but in in HeLa cells. This interesting finding is discussed in the text, lines 246-269 and 375-383 (see also our answers on page 3 above and page 7 below).

Major Comments

Evidence and Claims:

* While the methodological aspect is very strong the balance between presenting a novel method and presenting specific cell biological findings needs to be improved. Now it is quite unclear what the manuscript wants to present.

* The abstract needs a complete overhaul. From reading the abstract, it is not clear what the manuscript wants to present.

Response:

We have modified the abstract to make it more clear that we do not make any general claims on the impact of vimentin on the interactions and functions of different organelles, but rather describe what can be directly observed after the acute displacement of vimentin and which conclusions can be made from these observations.

Regarding the research findings there are a number of things for the authors to consider. Since the methods aspect is, in the eyes of this reviewer, in focus, I have not stringently assessed the experimental findings. Hence, the comments below are things to be considered in order to make the findings related to IF research stronger:

• *

* Cell-specific keratin interactions: The manuscript could benefit from some further validation of the physical interactions between vimentin and keratin-8 across different cell types.

Response:

We have improved the images of keratin and vimentin by using super-resolution (Airyscan) microscopy to show that they indeed form distinct filaments in HeLa cells, whereas in COS-7 and U2OS cells, where their co-displacement occurs, they can also incorporate into the same filaments. This observation was very surprising but agrees with the data published by the Gelfand lab on similarity in the distribution pattern and co-transport of vimentin and keratin in RPE1 cells (Robert et al, 2019, FASEB J). Colocalization and coprecipitation of vimentin and keratin has been also described by Velez-delValle at al. in epithelial cells (Sci Rep 2016).

* Impact on microtubules: The disorganization of stable microtubules in cells expressing KIF5A was attributed to overexpression effects. It would be helpful to include additional controls, such as expressing KIF5A without vimentin constructs, to confirm this claim.

Response:

This control has been included in the new Figure S3. We note that this observation fully aligns with data published by another lab (Andreu-Carbó et al, 2024, Nat Comm).

* ER-vimentin linkages: The observation that ER-vimentin interactions persist in RNF26 knockout cells is intriguing. The manuscript would benefit from a discussion on possible candidates for alternative linkers.

Response:

We have added a short discussion (lines 394-398) about the potential involvement of nesprins, such as nesprin-3, because they can connect the nuclear envelope to intermediate filaments, and might also partly participate in ER sheet-IF connections because ER and nuclear membranes are continuous and show some overlap in proteome.

* Construct variability: Do the authors have some data on how much Expression level differences significantly affect the outcomes (e.g., incomplete recovery)?

Response:

We have added a figure (Figure S2B), which shows that incomplete recovery of vimentin clustering does not correlate with protein expression levels and likely depends on other factors, which could possibly be the cell cycle phase or degree of vimentin entanglement after repositioning. This point is discussed in revised text, lines 194-197.

---

Reviewer #2 (Significance):

Significance

General Assessment: The study represents a significant technical advance in the study of cytoskeletal dynamics. The tools developed address critical limitations of traditional vimentin perturbation methods, allowing for spatiotemporally precise manipulation without long-term effects on gene expression or signaling pathways.

Novelty:

This is, to my knowledge, the first demonstration of reversible and acute vimentin repositioning using optogenetics. The study extends understanding of vimentin's short-term mechanical and organizational roles, distinguishing them from compensatory effects observed in knockdown models.

Audience and Impact: The manuscript will appeal to researchers in cytoskeletal dynamics, cell mechanics, and organelle biology. The tools have broader applicability in studying other cytoskeletal systems and could inspire translational applications, such as investigating the role of vimentin in cancer or fibrosis.

The reference list provide a relatively representative selection of articles relevant for the article. However, the authors may consider whether there could be relevant information in the relatively recent special edition of Current Opinion in Cell Biology, which focused on IFs, specially featuring vimentin https://www.sciencedirect.com/special-issue/10TFHK2QCKW

Response:

We thank the reviewer for this excellent suggestion, and we have included some additional references from this issue.

Field of Expertise

I specialize in cell biology, intermediate filaments, post-translational modifications, cytoskeletal dynamics, and advanced microscopy techniques.

Reviewer #3 (Evidence, reproducibility and clarity):

Summary:

This is an excellent paper describing the use of chemical and light-induced heterodimerization of microtubule-based motors to rapidly disrupt the distribution of the vimentin cytoskeletal network. Rapid clustering of vimentin did not significantly affect the microtubule or actin networks, cell spreading or focal adhesions. Other organelles were repositioned together with vimentin. Interestingly, in some cell lines, keratin networks were displaced along with vimentin while in other cells they were not.

Major comments:

The conclusions are well supported by the data presented and appropriate controls are included.

Optional comments:

1. • The authors should expand on why they think the plus end directed KIF5A gives such a strong localization of vimentin to the perinuclear area.* Response:

We think that two factors can contribute to this counterintuitive effect. First, vimentin is strongly concentrated and entangled in the perinuclear region, and displacement of some vimentin filaments to the cell periphery can cause the collapse of the rest to the cell center, with kinesins being unable to pull the perinuclear network apart. Second, kinesin-1 KIF5A is a motor that strongly prefers stable, post-translationally modified microtubules, and our previous study has shown that a significant proportion of such microtubules are located with their minus ends facing towards the cell periphery (Chen et al., Elife 2016). This could contribute to the accumulation of vimentin in the cell center upon KIF5A recruitment. These considerations were added to the revised text, lines 344-347.

• Consideration should be given to the idea that the pulling of ER and mitochondria along with the vimentin could be due to trapping of these organelles within the vimentin matrix and not necessarily due to direct interactions. Such reasoning could explain the transient localization of lysosomes with the center aggregate since lysosomes are generally not thought to significantly bind to vimentin networks.*

Response:

This is an excellent point, and we have included it in the revised article, lines 333-335 and 405.

Reviewer #3 (Significance):

This study describes some valuable tools that should be useful to cell biologists interested in determining the role of the cytoskeleton and possibly other organelles in a variety of cellular contexts. It overcomes some of the existing shortcomings of the pharmacological reagents currently available for studying intermediate filament biology and will provide a useful adjunct to other more long-term manipulations of the cytoskeleton. While much of the data presented confirm results obtained by other methods, this is a significant technical advance as it provides a short time scale, and in one instance, reversible manipulation of the cytoskeleton.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

This is an excellent paper describing the use of chemical and light-induced heterodimerization of microtubule-based motors to rapidly disrupt the distribution of the vimentin cytoskeletal network. Rapid clustering of vimentin did not significantly affect the microtubule or actin networks, cell spreading or focal adhesions. Other organelles were repositioned together with vimentin. Interestingly, in some cell lines, keratin networks were displaced along with vimentin while in other cells they were not.

Major comments:

The conclusions are well supported by the data presented and appropriate controls are included.

Optional comments:

1. The authors should expand on why they think the plus end directed KIF5A gives such a strong localization of vimentin to the perinuclear area.
2. Consideration should be given to the idea that the pulling of ER and mitochondria along with the vimentin could be due to trapping of these organelles within the vimentin matrix and not necessarily due to direct interactions. Such reasoning could explain the transient localization of lysosomes with the center aggregate since lysosomes are generally not thought to significantly bind to vimentin networks.

Significance

This study describes some valuable tools that should be useful to cell biologists interested in determining the role of the cytoskeleton and possibly other organelles in a variety of cellular contexts. It overcomes some of the existing shortcomings of the pharmacological reagents currently available for studying intermediate filament biology and will provide a useful adjunct to other more long-term manipulations of the cytoskeleton. While much of the data presented confirm results obtained by other methods, this is a significant technical advance as it provides a short time scale, and in one instance, reversible manipulation of the cytoskeleton.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary: The manuscript presents a novel methodology for acute manipulation of vimentin intermediate filaments (IFs) using chemical genetic and optogenetic tools. By recruiting microtubule-based motors to vimentin via inducible dimerization systems, the authors achieve precise temporal and spatial control over vimentin distribution. Apart from the significant advancement in terms of methods development, key findings include:

• Vimentin's role in organelle positioning: Mitochondria and ER are repositioned with vimentin, while lysosomes are less dependent on its organization.
• Cytoskeletal interactions: Vimentin clustering minimally impacts actin and microtubule networks in the short term.
• Cell stiffness: Vimentin repositioning reduces cell stiffness, indicating its significant role in cellular mechanics.
• Cell-type-specific keratin interactions: The study highlights diverse interactions between vimentin and keratin-8 across cell lines.

The study demonstrates methodological advancements enabling rapid vimentin manipulation and provides insights into vimentin's interactions with cellular structures. A major shortcoming is the unclear narrative, what do the authors want to present? This aspect requires significant attention.

General Comments and Overall Assessment

The manuscript represents an interesting contribution to the cytoskeletal field, addressing limitations of long-term perturbation methods. The tools developed are innovative, allowing controlled and reversible vimentin reorganization with minimal off-target effects. The findings are robust and provide important insights into the role of vimentin in cellular mechanics and organelle positioning.

Strengths:

Methodological novelty with broad applicability - this is the most exciting aspect. Comprehensive validation of the tools in multiple cell lines. Clear differentiation between vimentin's short- and long-term roles. Addressing gaps in understanding vimentin-organelle interactions.

Limitations:

• The manuscript is a little bit all over the place. While the method development is clear, the manuscript makes claims way beyond the method development. The message and narrative needs to be improved, and in the respect the whole structure needs an overhaul.
• Unclear how much the differences in expression levels impact results and reproducibility.
• Would be good to discuss some findings that are specific to a given experimental cell line. How generalizable are these results?

Major Comments

Evidence and Claims:

• While the methodological aspect is very strong the balance between presenting a novel method and presenting specific cell biological findings needs to be improved. Now it is quite unclear what the manuscript wants to present.
• The abstract needs a complete overhaul. From reading the abstract, it is not clear what the manuscript wants to present.

Regarding the research findings there are a number of things for the authors to consider. Since the methods aspect is, in the eyes of this reviewer, in focus, I have not stringently assessed the experimental findings. Hence, the comments below are things to be considered in order to make the findings related to IF research stronger:

• Cell-specific keratin interactions: The manuscript could benefit from some further validation of the physical interactions between vimentin and keratin-8 across different cell types.
• Impact on microtubules: The disorganization of stable microtubules in cells expressing KIF5A was attributed to overexpression effects. It would be helpful to include additional controls, such as expressing KIF5A without vimentin constructs, to confirm this claim.
• ER-vimentin linkages: The observation that ER-vimentin interactions persist in RNF26 knockout cells is intriguing. The manuscript would benefit from a discussion on possible candidates for alternative linkers.
• Construct variability: Do the authors have some data on how much Expression level differences significantly affect the outcomes (e.g., incomplete recovery)?

Significance

General Assessment: The study represents a significant technical advance in the study of cytoskeletal dynamics. The tools developed address critical limitations of traditional vimentin perturbation methods, allowing for spatiotemporally precise manipulation without long-term effects on gene expression or signaling pathways.

Novelty:

This is, to my knowledge, the first demonstration of reversible and acute vimentin repositioning using optogenetics. The study extends understanding of vimentin's short-term mechanical and organizational roles, distinguishing them from compensatory effects observed in knockdown models.

Audience and Impact: The manuscript will appeal to researchers in cytoskeletal dynamics, cell mechanics, and organelle biology. The tools have broader applicability in studying other cytoskeletal systems and could inspire translational applications, such as investigating the role of vimentin in cancer or fibrosis.

The reference list provide a relatively representative selection of articles relevant for the article. However, the authors may consider whether there could be relevant information in the relatively recent special edition of Current Opinion in Cell Biology, which focused on IFs, specially featuring vimentin https://www.sciencedirect.com/special-issue/10TFHK2QCKW

Field of Expertise

I specialize in cell biology, intermediate filaments, post-translational modifications, cytoskeletal dynamics, and advanced microscopy techniques.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary The manuscript is well written, with excellent explanation and documentation of experimental approaches. All conclusions are well supported by the data. The discussion is balanced and appropriate. The data, including images and movies, are of high quality and beautifully presented. The experimental design and analysis, including quantification of parameters in the images, is rigorous. Additional rigor is provided by comparing different cell types. The rapalog and iLID dimerization strategies have been described previously, as has their use to recruit kinesin motors to membranous organelles. However, this is the first application of these strategies to recruit motors to intermediate filaments. The evidence that vimentin filaments can be redistributed locally is clear and convincing and offers appealing potential for future experimentation. The redistribution was not fully reversible in all cells, but this is not surprising given the entanglement that must result from the action of motors along the length of these long flexible polymers.

In terms of the biology of intermediate filaments, the authors show that vimentin redistribution had negligible effect on microtubule or F-actin organization, cell area, or the number of focal adhesions. Depletion of vimentin filaments locally reduced cell stiffness. Both ER and mitochondria segregated with vimentin filaments, but not lysosomes. These findings are consistent with published reports (e.g. comparing vimentin null and wildtype cell lines), but the acute and reversible nature of the motor recruitment strategy is a more elegant experimental approach, and the selectivity of the observed effects is evidence of its specificity. It is interesting that the ER network segregated with vimentin even in the absence of RNF26. While this is not explored further, it points to the potential power of this motor recruitment strategy for future studies on intermediate filament interactions.

The following are some major and minor issues, which should all be easy for the authors to address.

Major Comments:

• Fig. S1 shows that the Vim-mCherry-FKBP construct coassembles with endogenous vimentin, but similar data for the iLID constructs appears to be lacking. I would like to see data demonstrating the incorporation of the Vim-mCherry-SspB constructs into the vimentin filaments. This should include high magnification images of single filaments in the cytoplasm of the cells.
• The authors do not discuss the density of motor recruitment along the filaments. To address this, I'd like to see images showing the extent of recruitment of motors to the filaments using the rapalog and LID strategies. This should include high magnification images of single filaments in the cytoplasm of the cells.
• For the experiments on vimentin and keratin organization, the authors do not explain that these proteins form distinct networks and do not coassemble. The authors should show this in the cell types examined. This should also be explained explicitly in the body of the manuscript, though the data could be placed in the supplementary data. This is important because many intermediate filaments can coassemble freely, and coassembled proteins would be expected to segregate together.

Minor Comments:

• The authors refer to selecting cells within an "optimized expression range" for their transiently expressed recombinant proteins. They should state the proportion of the cells that met this criterion in their transient transfection experiments as this is important information for other researchers that might wish to use this approach in their own studies.
• In Fig. 1F there should be a statistical comparison between cells transfected with the Kin14 construct and control (untransfected) cells in the absence of rapalog
• In Fig. 1G there should be a statistical comparison between cells expressing Kin14 and KIF5A in the absence of rapalog
• The depletion of the ER network in the cell periphery is not evident in Fig. 7B, though the perinuclear accumulation is evident. Perhaps the authors could select another example or explain to the reader what exactly to look for in these images.
• In Fig. 7C, the intensity of the mCherry declines markedly over time. This is presumably due to photobleaching but should be explained in the legend.

Referees cross-commenting

This session contains comments of Reviewer 1 and Reviewer 2

Reviewer 1:

I don't understand what Reviewer 2 means by "A major shortcoming is the unclear narrative, what do the authors want to present? This aspect requires significant attention." I found the narrative, purpose and conclusions of this study very clear to me. I also do not understand Reviewer 2's concern with the abstract. I re-read it and it still seems very clear and appropriate to me. For example, the authors state "Here, we present tools that allow rapid manipulation of vimentin IFs in the whole cytoplasm or within specific subcellular regions by inducibly coupling them to microtubule motors, either pharmacologically or using light". This seems clear and correct to me. It would be helpful if Reviewer 2 could point to specific language and explain why it is problematic.

Reviewer 2:

The strength of this paper is clearly the strong methods development and I find this aspect very intriguing and attractive. There is an imbalance in the narrative presenting on one hand the method and on the other hand presenting concrete research results. In my view, although interesting, the different experimental results serve more as proof-of-concept and they should not be presented as bona fide evidence of an existing or lacking bilateral interrealtionship.

Indeed, the cited sentence makes sense: "Here, we present tools that allow rapid manipulation of vimentin IFs in the whole cytoplasm or within specific subcellular regions by inducibly coupling them to microtubule motors, either pharmacologically or using light." as it features the methods aspect of the paper. However, the following sentences: "Perinuclear clustering of vimentin had no strong effect on the actin or microtubule organization, cell spreading, and focal adhesions, but reduced cell stiffness. Mitochondria and endoplasmic reticulum sheets were repositioned together with vimentin, whereas lysosomes were only briefly repositioned and rapidly regained their normal distribution. Keratin was displaced along with vimentin in some cell lines but remained intact in others. " embraces everything from actin to microtubules to cell spreading to focal ahdesions to cell stiffness to mitochondrial function to lysosomes to interactions with other IF family members etc. This gives the impression that the authors want to make claims on how vimentin affects or does not affect these cellular functions and structures and once just cannot make such sweeping claims with so little evidence. With the experimental setting included, non of these claims can be really made without rigidly examining each and every interaction (which has been done separately for many of these bilateral interactions during the past 20 years or so).

Hence, it should be made clear that these observations are used and mentioned as proof of concept that the tool is working, not as evidence that this or that interaction takes place or does not take place. As I indicated in my review, such claims on any of these bilateral interactions would require a lot more evidence to be properly substantiated.

My comment is to be regarded as a positive one. If I would judge the paper based on how one could interpret the abstract and the text regarding, for example, that vimentin does not affect focal adhesions but changes cellular stiffness, my review would be significantly more stringent. However, I would really like to see this paper being published, but the claims on revealing new vimentin functions or disproving earlier observations based on these very limited data are just not sufficiently substantiated to be acceptable. Hence, I urge the authors to adjust the narrative to be clear on the methods development, which is also the focus of the title. I believe this is a justified recommendation and also, overall, a fair shake of the study and a constructive approach on how to publish this manuscript without extensive experiments.

Reviewer 1:

I thank Reviewer 2 for this explanation. I do understand their point. However, while not the end of the story, I do feel the authors' data are a bit more than just a proof of principle and do offer important insights into the biology which the field will need to grapple with. Each graph includes measurements on dozens of cells from multiple experiments and there is clearly selectivity to what segregates with the vimentin filaments and what does not. I would just ask the authors to be a bit more nuanced in their interpretation and conclusions about the biology to address Reviewer 2's concerns. Reviewer 2:

That sounds like a fair assessment. Main thing is that this data is presented in a balanced way, with emphasis on the model development. Some of the presented data are in contradiction with quite established concepts by several researchers and the data presented here does not substantiate a paradigm shift. Regardless of this, some pieces of the data are intriguing, for example, the live cell imaging.

Significance

Summary: The authors show that chemical-induced and light-induced dimerization strategies can be used to recruit microtubule motors to vimentin filaments, allowing rapid and reversible experimental manipulation of vimentin filament organization either locally or globally in cells. These strategies provide an experimental approach for investigating the physical interaction of intermediate filaments with organelles and other cytoskeletal component, as well as a method for probing the role of intermediate filaments in cell mechanics, cytoskeletal dynamics, etc. This is a technical improvement over previous experimental strategies, which have relied largely on chronic manipulation such as global disassembly or genetic deletion of intermediate filaments, e.g. comparison of vimentin null and wild type cells.

The principal weakness of this study is that it offers limited insight into intermediate filament biology. As such, it might be most appropriate for a tools or techniques section of a journal. The dimerization strategies have been reported previously, so that is not new, but the application to intermediate filaments is novel.

Audience: This paper will be of interest to cell biologists who study cytoskeletal interactions, particularly the interaction of intermediate filaments with other cellular organelles or cytoskeletal polymers, or the role of intermediate filaments in cellular mechanics.

Reviewer Expertise This reviewer has expertise on the cytoskeleton, cytoskeletal dynamics, and intracellular transport including intermediate filament biology.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

__Reviewer #1 __

(Evidence, reproducibility and clarity (Required)):

The manuscript identified a novel role of Intraflagellar Transport Protein 20 (IFT20) in the function and homeostasis of lymphatic endothelial junctions. The authors showed that IFT20 regulates VE-cadherin localization at adherens junctions in lymphatic endothelial cells. The authors performed impressive in vivo work that shows the requirement for IFT20 for the homeostasis of intercellular junctions, lymphangiogenesis, and drainage function of lymphatic vessels. In contrast, the cell biology part of the paper was underwhelming and will need significant revisions to support the proposed model. In the result section, several conclusions have to be toned down to match the actual results. The study employs in vivo mouse models, immunofluorescence, biochemical assays, and loss-of-function experiments to support their conclusions.

Major comments - The authors present disrupted localization of VE-cadherin. Is this a mislocalization and/or protein stability issue in IFT20 KD cells? A western blot can help assess protein levels, and a phase-chase endocytosis assay of VE-cadherin can strengthen evidence. The authors did not confirm the permeability phenotype seen in vivo.

We thank the reviewer for this helpful suggestion.

Planned 1: Western blot to assess total VE-cadherin protein levels in IFT20 WT and KD cells.

Planned 2: Immunofluorescence staining for cell-surface VE-cadherin using permeabilized and non-permeabilized IFT20 WT and KD cells during VEGF-C stimulation and washout.

Together, these two experiments will assess VE-cadherin stability and more directly test the hypothesis that VE-cadherin does not recycle effectively back to the cell surface in the absence of IFT20.

• While the authors focused on IFT20 and rab5, we do not have a clear idea about the vesicular dynamics as well as the status of early, late, and recycling endosomes in IFT20 KD cells. Is IFT 20 localized to non-rab5+ endosomes, and if yes, what are the species? A more general endosomal profiling would help strengthen the authors' message. For example, in Fig. 4-5, the authors will have to stain for other early endosomal markers as well as late, and recycling endosomal markers in control and IFT20 KD cells.

Thank you for this helpful suggestion.

Planned 3: Immunofluorescence staining for EEA1 (early endosome), RAB7 (late endosome), RAB4 (fast recycling), RAB11 (recycling endosome) along with IFT20 to determine its localization pattern.

This experiment will determine the localization of IFT20 relative to various endosomal compartments.

• In fig. 6C, a majority of VE-cadherin is not associated with Rab5. Staining with additional endosomal markers might help identify other endosomal species colocalizing with VE-cadherin. It will be critical to add to Fig. 6c the intensity profiles depicting colocalizations. The authors can also live image a fluorescently (f)-tagged VE-cadherin (maybe with another f-tagged rab5) and assess their association dynamics in IFT20 KD cells (similar to fig6C).

Thank you for this helpful suggestion.

Planned 4: Immunofluorescence staining for EEA1 (early endosome), RAB7 (late endosome), RAB4 (fast recycling), RAB11 (recycling endosome) along with VE-cadherin in IFT20 WT and KD cells to determine its localization pattern.

Planned 5: Additional colocalization analysis such as adding intensity profiles and possibly proximity ligation assay.

Beyond the scope of this manuscript 1: While we agree that imaging the dynamics of FP-tagged VE-cadherin in live cells would provide more detail about its localization, we feel that this is beyond the scope of the current manuscript.

These experiments will determine the localization of VE-cadherin across various endosomal compartments and strengthen the current colocalization data.

• Primary cilia do seem to regulate vascular plexus in the mouse retina as well as endothelial permeability through mediating subcellular localization of junction proteins. The authors do not clearly exclude the ciliary function of IFT20 in mediating lymphatic endothelial cell-cell junctions. A rescue experiment can help settle this question by targeting IFT20 exclusively to cilia (or not) and assessing, for example, VE-cadherin localization. The following is optional: It is also unclear whether the described regulation is specific to IFT20 or can be phenocopied by the ablation of another IFT subunit and/or cilia ablation through the depletion of a non-IFT cilia assembly regulator.

Thank you for this helpful suggestion. We propose an alternative strategy.

Planned 6: To determine the role of ciliary vs. nonciliary functions, we will knockdown IFT74, an IFT protein in the same IFT complex B as IFT20 that is required for cilia assembly and function but is not known to participate in vesicular trafficking. We will assess VE-cadherin localization in IFT74 WT and KD cells by immunofluorescence.

Beyond the scope of the manuscript 2: We have not optimized reagents for targeting IFT20 to the cilium (e.g. ciliary targeting sequence) and believe that assessing the effects of a protein from the same IFT complex (IFT74) without known nonciliary functions will alleviate the reviewer’s concern.

• Figs. 7A and B do not seem very convincing. The control vs. IFT20 KD western blot levels look mostly similar between the two conditions. The result section does not translate the actual data in Fig. 7A and B. Additionally, there are no statistical comparisons between control and KD conditions in the graphs. Except for a potential pVEGFR-3 increase at 30 min VEGF-C in IFT20 KD cells, but after washout the level is similar to control. This figure does not support well the model presented in fig. 8. The conclusion in lines 456-459 has to be toned down.

Thank you for this helpful suggestion.

Planned 7: We will remove these western blot data with the exception of pVEGFR-3 and add phospho-tyrosine immunofluorescence. We will use immunofluorescence to quantify phosphorylated tyrosine levels and repeat western blots for pVEGFR-3 at different concentrations and time points of VEGF-C stimulation in IFT20 WT and KD cells. We will remove the other western blot data and revise the text accordingly. We will also attempt to pull down total VEGFR-3 and then blot for pVEGFR-3 to improve sensitivity of this assay.

These experiments will focus our analysis on the activation of VEGFR-3.

• The authors were not able to stain for pVEGFR3. It would still be helpful to see a colocalization between total VEGFR3, IFT20, and VE-cadherin in control cells and IFT20 KD cells (VEGFR3 and VE-cadherin).

Thank you for this helpful suggestion.

Planned 8: We will perform immunofluorescence for VEGFR-3, IFT20, and VECAD and assess their localization.

Minor comments - The control used in Figures 1 and 2 does not seem ideal. The proper control would be IFT20fl/fl cre neg. Is there a reason why the authors excluded a lox allele in control? Also, the authors have to provide the mice age used in these figures and when the Cre kicks in in the result section.

Thank you for this helpful suggestion.

Planned 9: We will clarify the use of control genotypes, and add mouse ages and Cre details to results/methods. This is a constitutive LYVE-1 Cre.

• Please describe the overall mouse phenotype(s) of the LYVE1 CRE-IFT20 flox.

Thank you for pointing out this oversight.

Planned 10: We will include a description of the overall phenotypes of LYVE1 Cre IFT20 KOs in the text. One notable phenotype is abdominal ascites.

• Line 109:'By expression, the authors probably mean immunostained.

Thank you for pointing out this oversight.

Planned 11: We will change to “immunostaining for”.

• Many graphs exhibit undefined Y-axis labels and units. Please clarify these as well as the way they were quantified. Include such information in figure legends and/or in the materials and methods section. The figures in question are fig1C, E and F, fig2E, fig3E, fig4b and D, fig6b and D, fig7B and C.

Thank you for this helpful suggestion.

Planned 12: We will clarify the quantification strategies and units in the text and figure legends and make sure the axes are clearly labeled.

• Line 295:'homeostasis"-the authors probably mean in a serum-rich condition.

Thank you for this helpful suggestion.

Planned 13: That is indeed what we meant. We will merge this sentence and the next sentence to be clearer.

• Fig4C specifically the lower two images on the right side: the images do not seem to represent the corresponding graphs.

Thank you for this helpful suggestion.

Planned 14: We will double check these images and adjust if necessary.

• Please add the statistical tests used to evaluate significance in all figure legends.

Thank you for pointing out this oversight.

Planned 15: We will be sure statistical tests are named in all figure legends.

Reviewer #1 (Significance (Required)):

This study provides novel insights into IFT20's role in VE-cadherin trafficking and endothelial junction stability, with its strongest aspect being the in vivo data in Figures 1 and 2, demonstrating lymphatic defects upon IFT20 loss. This represents a conceptual advance by extending IFT protein function beyond cilia (if one of the major comments is addressed) to vascular integrity. However, mechanistic depth is lacking, and ciliary role was not tested-additional rescue and colocalization experiments are needed to confirm the model. The study will interest vascular and lymphatic biologists, as well as cell biologists studying intracellular trafficking and cilia.

Expertise: cilia and mouse genetics

__Reviewer #2 __

(Evidence, reproducibility and clarity (Required)):

Paulson et al. use an in vivo model of IFT20 deletion (Lyve1-Cre) and primary lymphatic endothelial cell (LEC) cultures to investigate the role of IFT20 in controlling LEC-LEC junction dynamics. The key findings/suggestions include: i) Authors show alterations in the VE-cadherin (or ZO-1) staining at the LEC junctions upon IFT20 deletion or silencing. ii) They also show evidence of the IFT20 localization to RAB5 endosomes and alteration of RAB5 endosome dynamics upon IFT20 silencing.

In the current manuscript, some of the key data are not convincing. Further experimentation and analysis (also of the existing data) are needed to solidify the authors' statements as detailed below. I expect that the suggested experiments can be executed in 3-to-6 months and require, at least, antibodies, which have not been used in the current manuscript.

Major comments

1. The data information, presented in the figure legends, is difficult to understand. The authors should always indicate how many biological replicates and independent experiments the data is derived from. This holds also for the representative images. Now, it seems that some of the quantified data are derived from only 1 experiment (see, for example, rows 423-425: "Graphs show one representative biological replicate of two, each comprising two technical replicates with 100+ cells per condition"). The quantifications should be based on data from at least three independent experiments.

Often data points represent the field of views from a single sample, thus, biasing the statistical testing. The data points should represent biological replicates or independent experiments to allow the reader to make conclusions, about whether the findings are statistically significant and can be repeated.

Thank you for this helpful critique.

Planned 16: We will be sure to indicate biological and technical replicates and ensure that quantifications are representative of at least three independent experiments. We will also ensure that quantifications are statistically robust.

The Lyve1-Cre is not specific for lymphatic vasculature (for example https://www.jax.org/strain/012601# and Lee LK et al. 2020, Cell Reports), as also stated by the authors (row 112). However, this is not shown in the data and complicates the interpretation of the data. Here, authors can stain the IFT20 with their existing mouse IFT20-specific antibody to show the loss in the lymphatic and/or blood vasculature. If IFT20 is lost in both vasculature types, it is not possible to say "lymphatic specific" (for example, row 143) and draw conclusions that the observed phenotypes would be primary to IFT20 loss in the lymphatic vasculature.

Thank you for this helpful suggestion.

Planned 17: We will assess IFT20 KO in blood vasculature and tone down lymphatic-specific language in the text.

The authors write (rows 164-168) "Lymphatic vessels in the IFT20 KO or VE-cadherin KO embryonic dorsal skin exhibited increased and variable lumen size and excessive branching, suggesting that impaired lymphatic organization and function contributed to the fluid homeostasis defect. Here, immunofluorescence staining for LYVE-1 in the ear skin revealed similar patterning defects in adult IFT20 KO lymphatic vessels (Figure 2A), that have also been described in VE-cadherin KO mice (Hägerling et al., 2018)." However, based on Figure 2A, it is not obvious that there would be excessive lymphatic vessel branching, impaired organization or similarities to VE-cadherin deleted lymphatic vessels. To justify their statement, the authors should provide quantification of the branching (at least 3 mice/genotype).

Thank you for this helpful critique.

Planned 18: Based on the suggestion from Reviewer 3, we will remove these morphological and skin drainage data.

IFT20 deletion or silencing causes alterations in the cell junction pattern/VE-cadherin intensity. The authors' interpretation that IFT20 deletion/silencing would cause discontinuous or "button-like" junctions is not supported by the provided images (Figures 1E, 3F, 6A, 6C). Rather, it seems that the levels of VE-cadherin in vivo are decreased, whereas the "continuity" of the junction is not altered. In cell culture, IFT20 silencing seems to cause wider and, to some extent, overlapping VE-cadherin junctions and not "discontinuous". These junctions may represent a more immature state. The authors should change the nomenclature accordingly or provide additional data. Using the existing cell culture experiment images, it would be more appropriate to analyze the width of the VE-cadherin junctions, instead of the "granularity".

Thank you for this helpful suggestion.

To assess VE-cadherin levels in vitro, we will perform western blots as described in Planned 1 above.

Planned 19: We will measure widths of junctions from IFT20 KD and WT images and adjust the language in the text.

Paulson et al. show images of IFT20 and RAB5 double-stained samples. The co-localization seems to happen mostly at the weakly IFT20 positive puncta (Figure 3A-B). Authors should show the disappearance of the signal in the siIFT20 treated samples (in comparison to siControl samples) to highlight the specificity of the weak signal.

Thank you for this helpful suggestion.

Planned 20: We will add data showing the IFT20 KD more clearly at high magnification.

1. The Authors analyze the co-localization of VE-cadherin and RAB5 as co-localization area (Figure 6C-D). The images show that the co-localization is stated to happen at LEC periphery/junctions. LEC periphery is notoriously thin and microscope Z-resolution does not allow distinction of truly co-localizing or "on top of each other" signal. Based on row 607 co-localization would be expected to happen at least in EEA1+ vesicles, which are located perinuclearly (not at the junctions) in LECs (Korhonen et al. 2022, JCI). Authors could use EEA1, RAB5, and VE-cadherin triple staining for the quantification.

Thank you for this helpful suggestion.

Please see Planned 3 and Planned 4 above where we propose experiments to address this concern.

In the current experiments, authors cannot conclude whether the VE-cadherin signal is at the cell junction (non-internalized), in endosomes (internalized during the experiment), or newly produced VE-cadherin on its way to the plasma membrane. To allow conclusions about the internalized VE-cadherin, and its localization in RAB5 vesicles, authors should conduct, for example, a classical endocytosis assay: incubation of live cells with non-blocking anti-VE-cadherin antibody, followed by acid wash to remove the non-internalized antibody, fixation and staining for RAB5. Also, shorter VEGF-C treatment would allow conclusions about the VE-cadherin dynamics.

Thank you for this helpful suggestion.

In Planned 2 above, we will perform immunofluorescence staining for cell-surface VE-cadherin using permeabilized and non-permeabilized IFT20 WT and KD cells during VEGF-C stimulation at various timepoints and washout to address this concern.

siRNAs can have off-target effects and, thus, the use of at least two independent methods/oligos for silencing is needed. Paulson et al. use a pool of 4 oligos for silencing. They should rather test the efficacy of the single oligos and then use the two best oligos (1/sample) to show and quantify the same phenotype. This is needed at least for the key experiments shown in Figures 4C-D, Figure 6A-B (see also comment #3), Figure 7A-B

Thank you for this helpful suggestion. We chose these reagents based on pooled siRNAs at low concentration minimizing off-target effects while still achieving strong KD vs. single siRNAs at higher concentration. Please see this technical note for further information about minimizing off-target effects by the use of pooled siRNAs vs. single siRNAs: https://horizondiscovery.com/-/media/Files/Horizon/resources/Application-notes/off-target-tech-review-technote.pdf?sc_lang=en

1. “SMARTpool siRNA reagents pool four highly functional SMARTselection designed siRNAs targeting the same gene. Studies show that strong on-target gene knockdown can be achieved with minimal off-target effects if a pool consisting of highly functional multiple siRNA is subsituted for individual duplexes. This finding is in contrast to speculation that mixtrues of siRNAs can compound off-target effects. … [Their data show that] while individual duplexes delivered at 100 nM can induce varying numbers of off-targeted genes, transfection of the corresponding SMARTpool siRNA (100 nM total concentration) induces only a fraction of the total off-target profile.”
2. “Our scientists have identified a unique combination of [chemical] modifications that eliminate as much as 80% of off-target effects.”
3. “The ON-TARGETplus product line is comprised of four individual siRNAs, and SMARTpool reagents which are chemically modified and rationally designed to minimize off-target effects.”

   OPTIONAL: Paulson et al stated in the first article (2021, Front. Cell Dev. Biol.) that IFT20 deletion/silencing causes lymphatic endothelial phenotypes due to its role in primary cilia, whereas here the authors conclude that IFT20 controls VE-cadherin dynamics at the RAB5 vesicles. However, the current experiments cannot dissect the role of IFT20 in these two distinct locations. For this, authors could delete/silence another gene required for primary cilia or RAB5 endosomes and then analyze, which IFT20 phenotypes are recapitulated.

Thank you for this helpful suggestion. Please see Planned 6 above where we propose to determine the role of ciliary vs. nonciliary IFT functions by knocking down IFT74, an IFT protein in the same IFT complex B as IFT20 that is required for cilia assembly and function but is not known to participate in vesicular trafficking. We will assess VE-cadherin localization in IFT74 WT and KD cells by immunofluorescence.

The data shown in Figure 2 B-E (Lymphatic drainage) is not necessary for the current manuscript ("IFT20 regulates VE-cadherin traffic in LECs") and can be removed. As the authors state in the manuscript, the drainage phenotype may be due to lymphatic vessel valve defects (rows 584-585) rather than primary for LEC-LEC junction defects. The data does not justify the abstract sentence "and lymph transport is impaired by intracellular sequestration of VE-cadherin" (row 42).

Thank you for this helpful suggestion. Please see Planned 18 above, where we propose to remove these data.

Minor comments

1. For some of the images, the signal should be enhanced to allow visual inspection also in the paper version (Figures 5A-B and 6C, magenta).

Thank you for this helpful suggestion.

Planned 21: We will enhance the signal in the indicated figures.

Authors show representative Western Blots and quantification of several biological replicates/sample types to investigate signaling responses upon VEGF-C treatment of control and siIFT20 cells. The authors state that the P-levels of VEGFR3, ERK, VE-cadherin, and AKT have different dynamics in control and IFT20-silenced cells. To justify this conclusion, authors should test the statistical significance between the siControl and siIFT20 samples at each time point. The current quantification (Figure 7B) shows that there is, at least, a trend of increased p-VEGFR3, p-VE-cadherin, p-ERK, and p-AKT in IFT20 silenced cells. However, the representative Western Blot image does not display a clear difference (Figure 7A). Authors should include the original western blots, used for quantification, as supplements.

Thank you for this helpful suggestion. Please see Planned 7 above where we propose to remove these data with the exception of pVEGFR-3 and add corresponding immunofluorescence data. We will ensure blots are included as supplemental figures.

The authors use western blot quantification to show that the altered LEC junctions affect VEGFR3 signaling. They further hypothesize that the increased VEGFR3 signaling may be a consequence of VEGFR3 localization in endosomes. The authors did not detect any signal using the phospho-specific VEGFR3 antibody (rows 441-442). To analyze the location of VEGFR3 upon VEGF-C treatment in siControl and siIFT20 LECs, the authors should use anti-VEGFR3 (total) antibodies that have been shown to detect VEGFR3 in similar assays.

Thank you for this helpful suggestion.

Please see Planned 8 above where we will perform immunofluorescence for VEGFR-3, IFT20, and VECAD and assess their localization.

The normality of the data should be tested before the selection of the statistical test. If this has been done, please, indicate it in the materials and methods or re-run the statistical analysis, if some of the data is not normally distributed.

Thank you for this helpful suggestion.

Planned 22: We will double check the statistics and normality for all quantifications.

The authors should use arrows, arrowheads, etc. to highlight examples of relevant features in the images. For example, in Figure 3C, the increased stress fiber formation is not obvious to the reader.

Thank you for this helpful suggestion.

Planned 23: We will add arrows etc. where appropriate.

Reviewer #2 (Significance (Required)):

Lymphatics are essential for fluid, leukocyte, and lipid trafficking to lymph nodes and/or systemic circulation. Recent findings have promoted lymphatics as a potential target to control the level of adaptive immunity in inflammation-associated diseases, including tumorigenesis (for example Song et al 2020, Nature). Early work on lymphatic endothelium in vivo, highlighted the dynamics of lymphatic endothelial junction, which, reversibly, can alter between continuous and discontinuous ("button-like") states (Baluk 2007, Am. Jour. Pathol.; Yao 2012, Am J. Pathol.). These changes may have an effect on fluid drainage capacity, lymphatic vessel growth, and prevention of pathogen dissemination to the systemic circulation. Recently, lymphatic junctions have been shown to present hubs of VEGFR3 signaling, VEGFR3 and VE-cadherin dynamics, and leukocyte transmigration (Sung et al. 2022, Nat. Cardiovasc. Res.; Hagerling et al. 2018, EMBO J.; Liaqat et al. 2024, EMBO J.). Thus, the manuscript by Paulson et al. investigates a topical subject.

The authors suggest a role for IFT20 in the control of VE-cadherin dynamics. Based on my expertise in lymphatic endothelial biology, I envision that the manuscript can potentially increase knowledge on the regulators of the lymphatic endothelial junctions, which might have physiological, and in the long term, translational significance. However, in the current manuscript, the exact mechanisms of how IFT20 controls lymphatic endothelial junctions are left open. In addition to the lymphatic research field, the study is, potentially of interest to researchers working on blood vasculature or, even, epithelium, i.e. tissues where junctional dynamics play a major role in health and disease.

Furter controls, analysis, and experimentation are needed to warrant the authors' statements. In their future work, the authors should also consider means to rigorously dissect the IFT20 functions in primary cilia and endosomes.

__Reviewer #3 __

(Evidence, reproducibility and clarity (Required)):

In this manuscript, the group of Fink and coworkers investigates mechanistic aspects of the intraflagellar transport protein 20 (IFT20) function in lymphatic endothelial cells (LECs). In a previous study, this group had demonstrated the presence of primary cilia on LECs and shown that loss of IFT20 during development resulted in edema, lymphatic vessel dilation and altered branching. Lymphatic-specific deletion of IFT20 cell-autonomously exacerbated acute lymphangiogenesis after corneal suture. In this manuscript, Paulson et al. recapitulate the suture-induced hyper-lymphangiogenesis after lymphatic-specific IFT20 KO using a LYVE1-Cre delete strain and demonstrate a reduced, more discontinuous VE-Cadherin (VECad) staining in newly formed lymphatic vessels (LVs). Prompted by distended and hyperbranching dermal vessels, the performed functional tracer injection experiments and demonstrate increased lymphatic backflow and leakage into the interstitium. To gain further mechanistic insights the authors turned to reductionist cell culture models, starting with a mouse LEC line, in which IFT20 had been deleted using CRISPR/Cas9 resulting in loss of primary cilia, increased stress fibre formation and impaired junctional integrity. More importantly, similar effects were detected in human dermal (HD)LECs after IFT20 KD. Further IFT20 KD HDLECs showed accumulation of RAB5+ vesicles indicating defective endosome maturation. Indistinguishable formation of RAB5+ endosomes after VEGF-C stimulation in HDLECs and IFT20KD HDLECs indicated that endocytosis and formation of early endosomes occur independent if IFT20. Through starvation, stimulation and wash-out experiments the authors provide colocalization data suggesting that after VEGF-C stimulation IFT20 is recruited to endosomes where it contributes to VECad recycling. Finally, the authors addressed if the increase in RAB5+ endosomes following VEGF-C stimulation resulted in prolonged retention of signaling-active VEGFR-3 in endosomes. Western blotting for phosphorylation of VEGFR-3 and its downstream signaling components after activation of starved HDLECs or IFT20KD HDLECs and subsequent factor wash out provided evidence towards this model.

Subsequently open question and potential suggestions for improvement are listed: The authors describe a slight leakiness of the LYVE1-Cre deleter strain to result in massive hemangiogenesis (line112). How extensive is the resulting deletion in blood endothelial cells? What are the consequences for VECad distribution in BEC junctions i.e. for blood vessels and vascular permeability? Are the defects described specific for LECs or are the manifestation of generic defects in LECs?

Thank you for these helpful suggestions.

Please see Planned 17 above where we will assess IFT20 KO in blood vasculature and tone down lymphatic-specific language in the text.

Fig. 1 E, what is the distribution of LYVE1 in IFT20 KO LECs at higher magnification, is LYVE1 excluded from the VECad expression domain?

Thank you for this helpful suggestion.

Planned 24: We will review our corneal confocal data to address this question.

Fig.1 F, what does VECad-positive LV (%) area (line 154 - 155) refer to, given that all LECs are VECad+ but the junctional distribution of the protein is distinctly different?

Thank you for pointing out our need to clarify. This quantification measures the overlap of VE-cadherin with LYVE-1 as a way to measure the area covered by adherens junctions between lymphatic endothelial cells. Where junctions are punctate, they have smaller area vs. long continuous junctions.

Planned 25: We will update the text to clarify this measurement.

In the discussion, the authors speculate that the development of valves could be potentially impaired in IFT20 LEC KO mice. Ear skin would be an excellent tissue to stain the valves and analyse their structure in collecting LVs. Of particular interest in this context are Int a9, VECad, FOXC2 and PROX1 expression. The later two are required for valve formation and upregulated in valve forming areas in response to oscillatory shear stress (Sabine A et al. (2012) Dev Cell 22 (2):430-445. doi:10.1016/j.devcel.2011.12.020).

Thank you for this helpful suggestion. Based on the suggestion from Reviewer 2, we will remove the ear lymph drainage data and focus on the cell biology in this manuscript. Our current experiments focus more on lymphatic valve formation in this context and these data can be moved to a separate manuscript.

Planned 26: We will revise the text to remove speculation about valve development in this model and address this in a later manuscript.

Does IFT20 KO and loss of the primary cilium impair OSS sensing and result in a failure to express sufficient levels of PROX1 for valve formation (Fig. 7 C).

Thank you for this helpful comment. We will address the role of cilia in OSS sensing and valve formation in a forthcoming manuscript.

A larger area view including pre-collectors and collectors would be informative and reveal changes in the overall structure of the lymphatic vessel bed in absence of IFT20.

Based on the suggestion from Reviewer 2, we will remove these data.

Fig. 2 A, (line 187 - 190) please indicate the age of analysed animals.

Planned 27: We will add the ages of mice used.

With respect to Fig.1, LVs in the area are mainly capillaries, what is the distribution of VECad? Are the LVs comprised of oak-leave shaped LECs, higher magn. pictures would be required.

Thank you for this helpful suggestion.

Planned 28: We will include higher magnification images of capillaries.

Fig. 2 (C - E) Line 201 - 203 the description of retrograde flow using a clock terminology is unusual and not clear to the reader. Is this meant relative to the point of injection with 12 being at the top or relative to the injection axis (i.e. forward / backward direction)? It would seem that indication of the angle in combination with a sketch of the analysis would help the reader to interpret these data.

Thank you for this helpful critique. We will remove these data based on this suggestion and that of Reviewer 2.

The application of cell culture models is appropriate, however, the value of the mLEC model is questionable given that VECad is not detectable in these cells and PROX1 and VEGFR-3 staining are not shown. Therefore, the HDLEC model bears significantly more relevance. In Fig. 3D, were mLECS mitotically arrested during the 24hrs transwell migration, to exclude division and crowding effects during the observation time?

Thank you for this helpful critique.

Planned 29: We will clarify the methods for this experiment in the text.

Fig. 6 It is commendable that the authors report their lack of success to directly visualize VEGFR-3 endocytosis by IF and attempt a WB analysis instead. However given the spread of the results normalization to ß-actin as a loading control appears inappropriate. Phosphorylated forms of VEGFR-3 and VECad should be normalized to the expression of the total protein as measured with a non-phospospecific antibody, exactly the way done here for ERK1/2 and AKT. Generally, IP-WB experiments provide superior data in this type of setting.

Thank you for this helpful suggestion. Based on suggestions from the other reviewers, we will remove these WB data with the exception of pVEGFR-3 and add corresponding immunofluorescence. We will include additional time points and include blots used for quantifications as supplements.

Line 597 - 599: "VEGFR-3 signaling is required for the establishment of VE-cadherin button junctions as lymphatic collecting vessels mature but is not required for their maintenance (Jannaway et al., 2023)." Collecting LVs are characterized by zipper junctions, but not button junctions. Therefore, this sentence needs clarification.

Thank you for this helpful suggestion.

Planned 30: We will clarify this text.

Reviewer #3 (Significance (Required)):

The role of IFT20 in formation of the primary cilium and endocytic vesicle transport warrants its investigation in lymphatic endothelial cells. Therefore, this study addresses relevant questions and provides important first insights into the cell biological function of IFT20 in this cell type. IFT20 has so far not been implicated in endocytosis and recycling of VECad and VEGFR-3 and the model suggested by the authors is compelling and adds to the mechanistic understanding of previous studies on the role of VECad in LECs. In particular, it could be of relevance for the enigmatic formation of button junction in lymphatic capillaries and the mechano-response of LECS underlying valve formation. At this point, the picture obtained from the endocytosis assays is more conclusive compared to the analysis of the impact of IFT20 loss on button junction formation. Clearly the study is of interest for a general cell biological audience as well as vascular biologists.

• *

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, the group of Fink and coworkers investigates mechanistic aspects of the intraflagellar transport protein 20 (IFT20) function in lymphatic endothelial cells (LECs). In a previous study, this group had demonstrated the presence of primary cilia on LECs and shown that loss of IFT20 during development resulted in edema, lymphatic vessel dilation and altered branching. Lymphatic-specific deletion of IFT20 cell-autonomously exacerbated acute lymphangiogenesis after corneal suture. In this manuscript, Paulson et al. recapitulate the suture-induced hyper-lymphangiogenesis after lymphatic-specific IFT20 KO using a LYVE1-Cre delete strain and demonstrate a reduced, more discontinuous VE-Cadherin (VECad) staining in newly formed lymphatic vessels (LVs). Prompted by distended and hyperbranching dermal vessels, the performed functional tracer injection experiments and demonstrate increased lymphatic backflow and leakage into the interstitium. To gain further mechanistic insights the authors turned to reductionist cell culture models, starting with a mouse LEC line, in which IFT20 had been deleted using CRISPR/Cas9 resulting in loss of primary cilia, increased stress fibre formation and impaired junctional integrity. More importantly, similar effects were detected in human dermal (HD)LECs after IFT20 KD. Further IFT20 KD HDLECs showed accumulation of RAB5+ vesicles indicating defective endosome maturation. Indistinguishable formation of RAB5+ endosomes after VEGF-C stimulation in HDLECs and IFT20KD HDLECs indicated that endocytosis and formation of early endosomes occur independent if IFT20. Through starvation, stimulation and wash-out experiments the authors provide colocalization data suggesting that after VEGF-C stimulation IFT20 is recruited to endosomes where it contributes to VECad recycling. Finally, the authors addressed if the increase in RAB5+ endosomes following VEGF-C stimulation resulted in prolonged retention of signaling-active VEGFR-3 in endosomes. Western blotting for phosphorylation of VEGFR-3 and its downstream signaling components after activation of starved HDLECs or IFT20KD HDLECs and subsequent factor wash out provided evidence towards this model.

Subsequently open question and potential suggestions for improvement are listed: The authors describe a slight leakiness of the LYVE1-Cre deleter strain to result in massive hemangiogenesis (line112). How extensive is the resulting deletion in blood endothelial cells? What are the consequences for VECad distribution in BEC junctions i.e. for blood vessels and vascular permeability? Are the defects described specific for LECs or are the manifestation of generic defects in LECs? Fig. 1 E, what is the distribution of LYVE1 in IFT20 KO LECs at higher magnification, is LYVE1 excluded from the VECad expression domain? Fig.1 F, what does VECad-positive LV (%) area (line 154 - 155) refer to, given that all LECs are VECad+ but the junctional distribution of the protein is distinctly different?

In the discussion, the authors speculate that the development of valves could be potentially impaired in IFT20 LEC KO mice. Ear skin would be an excellent tissue to stain the valves and analyse their structure in collecting LVs. Of particular interest in this context are Int a9, VECad, FOXC2 and PROX1 expression. The later two are required for valve formation and upregulated in valve forming areas in response to oscillatory shear stress (Sabine A et al. (2012) Dev Cell 22 (2):430-445. doi:10.1016/j.devcel.2011.12.020). Does IFT20 KO and loss of the primary cilium impair OSS sensing and result in a failure to express sufficient levels of PROX1 for valve formation (Fig. 7 C). A larger area view including pre-collectors and collectors would be informative and reveal changes in the overall structure of the lymphatic vessel bed in absence of IFT20. Fig. 2 A, (line 187 - 190) please indicate the age of analysed animals. With respect to Fig.1, LVs in the area are mainly capillaries, what is the distribution of VECad? Are the LVs comprised of oak-leave shaped LECs, higher magn. pictures would be required. Fig. 2 (C - E) Line 201 - 203 the description of retrograde flow using a clock terminology is unusual and not clear to the reader. Is this meant relative to the point of injection with 12 being at the top or relative to the injection axis (i.e. forward / backward direction)? It would seem that indication of the angle in combination with a sketch of the analysis would help the reader to interpret these data.

The application of cell culture models is appropriate, however, the value of the mLEC model is questionable given that VECad is not detectable in these cells and PROX1 and VEGFR-3 staining are not shown. Therefore, the HDLEC model bears significantly more relevance. In Fig. 3D, were mLECS mitotically arrested during the 24hrs transwell migration, to exclude division and crowding effects during the observation time?

Fig. 6 It is commendable that the authors report their lack of success to directly visualize VEGFR-3 endocytosis by IF and attempt a WB analysis instead. However given the spread of the results normalization to ß-actin as a loading control appears inappropriate. Phosphorylated forms of VEGFR-3 and VECad should be normalized to the expression of the total protein as measured with a non-phospospecific antibody, exactly the way done here for ERK1/2 and AKT. Generally, IP-WB experiments provide superior data in this type of setting. Line 597 - 599: "VEGFR-3 signaling is required for the establishment of VE-cadherin button junctions as lymphatic collecting vessels mature but is not required for their maintenance (Jannaway et al., 2023)." Collecting LVs are characterized by zipper junctions, but not button junctions. Therefore, this sentence needs clarification.

Significance

The role of IFT20 in formation of the primary cilium and endocytic vesicle transport warrants its investigation in lymphatic endothelial cells. Therefore, this study addresses relevant questions and provides important first insights into the cell biological function of IFT20 in this cell type. IFT20 has so far not been implicated in endocytosis and recycling of VECad and VEGFR-3 and the model suggested by the authors is compelling and adds to the mechanistic understanding of previous studies on the role of VECad in LECs. In particular, it could be of relevance for the enigmatic formation of button junction in lymphatic capillaries and the mechano-response of LECS underlying valve formation. At this point, the picture obtained from the endocytosis assays is more conclusive compared to the analysis of the impact of IFT20 loss on button junction formation. Clearly the study is of interest for a general cell biological audience as well as vascular biologists.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Paulson et al. use an in vivo model of IFT20 deletion (Lyve1-Cre) and primary lymphatic endothelial cell (LEC) cultures to investigate the role of IFT20 in controlling LEC-LEC junction dynamics. The key findings/suggestions include: i) Authors show alterations in the VE-cadherin (or ZO-1) staining at the LEC junctions upon IFT20 deletion or silencing. ii) They also show evidence of the IFT20 localization to RAB5 endosomes and alteration of RAB5 endosome dynamics upon IFT20 silencing.

In the current manuscript, some of the key data are not convincing. Further experimentation and analysis (also of the existing data) are needed to solidify the authors' statements as detailed below. I expect that the suggested experiments can be executed in 3-to-6 months and require, at least, antibodies, which have not been used in the current manuscript.

Major comments

1. The data information, presented in the figure legends, is difficult to understand. The authors should always indicate how many biological replicates and independent experiments the data is derived from. This holds also for the representative images. Now, it seems that some of the quantified data are derived from only 1 experiment (see, for example, rows 423-425: "Graphs show one representative biological replicate of two, each comprising two technical replicates with 100+ cells per condition"). The quantifications should be based on data from at least three independent experiments.

Often data points represent the field of views from a single sample, thus, biasing the statistical testing. The data points should represent biological replicates or independent experiments to allow the reader to make conclusions, about whether the findings are statistically significant and can be repeated. 2. The Lyve1-Cre is not specific for lymphatic vasculature (for example https://www.jax.org/strain/012601# and Lee LK et al. 2020, Cell Reports), as also stated by the authors (row 112). However, this is not shown in the data and complicates the interpretation of the data. Here, authors can stain the IFT20 with their existing mouse IFT20-specific antibody to show the loss in the lymphatic and/or blood vasculature. If IFT20 is lost in both vasculature types, it is not possible to say "lymphatic specific" (for example, row 143) and draw conclusions that the observed phenotypes would be primary to IFT20 loss in the lymphatic vasculature. 3. The authors write (rows 164-168) "Lymphatic vessels in the IFT20 KO or VE-cadherin KO embryonic dorsal skin exhibited increased and variable lumen size and excessive branching, suggesting that impaired lymphatic organization and function contributed to the fluid homeostasis defect. Here, immunofluorescence staining for LYVE-1 in the ear skin revealed similar patterning defects in adult IFT20 KO lymphatic vessels (Figure 2A), that have also been described in VE-cadherin KO mice (Hägerling et al., 2018)." However, based on Figure 2A, it is not obvious that there would be excessive lymphatic vessel branching, impaired organization or similarities to VE-cadherin deleted lymphatic vessels. To justify their statement, the authors should provide quantification of the branching (at least 3 mice/genotype). 4. IFT20 deletion or silencing causes alterations in the cell junction pattern/VE-cadherin intensity. The authors' interpretation that IFT20 deletion/silencing would cause discontinuous or "button-like" junctions is not supported by the provided images (Figures 1E, 3F, 6A, 6C). Rather, it seems that the levels of VE-cadherin in vivo are decreased, whereas the "continuity" of the junction is not altered. In cell culture, IFT20 silencing seems to cause wider and, to some extent, overlapping VE-cadherin junctions and not "discontinuous". These junctions may represent a more immature state. The authors should change the nomenclature accordingly or provide additional data. Using the existing cell culture experiment images, it would be more appropriate to analyze the width of the VE-cadherin junctions, instead of the "granularity". 5. Paulson et al. show images of IFT20 and RAB5 double-stained samples. The co-localization seems to happen mostly at the weakly IFT20 positive puncta (Figure 3A-B). Authors should show the disappearance of the signal in the siIFT20 treated samples (in comparison to siControl samples) to highlight the specificity of the weak signal. 6. The Authors analyze the co-localization of VE-cadherin and RAB5 as co-localization area (Figure 6C-D). The images show that the co-localization is stated to happen at LEC periphery/junctions. LEC periphery is notoriously thin and microscope Z-resolution does not allow distinction of truly co-localizing or "on top of each other" signal. Based on row 607 co-localization would be expected to happen at least in EEA1+ vesicles, which are located perinuclearly (not at the junctions) in LECs (Korhonen et al. 2022, JCI). Authors could use EEA1, RAB5, and VE-cadherin triple staining for the quantification.

In the current experiments, authors cannot conclude whether the VE-cadherin signal is at the cell junction (non-internalized), in endosomes (internalized during the experiment), or newly produced VE-cadherin on its way to the plasma membrane. To allow conclusions about the internalized VE-cadherin, and its localization in RAB5 vesicles, authors should conduct, for example, a classical endocytosis assay: incubation of live cells with non-blocking anti-VE-cadherin antibody, followed by acid wash to remove the non-internalized antibody, fixation and staining for RAB5. Also, shorter VEGF-C treatment would allow conclusions about the VE-cadherin dynamics. 7. siRNAs can have off-target effects and, thus, the use of at least two independent methods/oligos for silencing is needed. Paulson et al. use a pool of 4 oligos for silencing. They should rather test the efficacy of the single oligos and then use the two best oligos (1/sample) to show and quantify the same phenotype. This is needed at least for the key experiments shown in Figures 4C-D, Figure 6A-B (see also comment #3), Figure 7A-B 8. OPTIONAL: Paulson et al stated in the first article (2021, Front. Cell Dev. Biol.) that IFT20 deletion/silencing causes lymphatic endothelial phenotypes due to its role in primary cilia, whereas here the authors conclude that IFT20 controls VE-cadherin dynamics at the RAB5 vesicles. However, the current experiments cannot dissect the role of IFT20 in these two distinct locations. For this, authors could delete/silence another gene required for primary cilia or RAB5 endosomes and then analyze, which IFT20 phenotypes are recapitulated. 9. The data shown in Figure 2 B-E (Lymphatic drainage) is not necessary for the current manuscript ("IFT20 regulates VE-cadherin traffic in LECs") and can be removed. As the authors state in the manuscript, the drainage phenotype may be due to lymphatic vessel valve defects (rows 584-585) rather than primary for LEC-LEC junction defects. The data does not justify the abstract sentence "and lymph transport is impaired by intracellular sequestration of VE-cadherin" (row 42).

Minor comments

1. For some of the images, the signal should be enhanced to allow visual inspection also in the paper version (Figures 5A-B and 6C, magenta).
2. Authors show representative Western Blots and quantification of several biological replicates/sample types to investigate signaling responses upon VEGF-C treatment of control and siIFT20 cells. The authors state that the P-levels of VEGFR3, ERK, VE-cadherin, and AKT have different dynamics in control and IFT20-silenced cells. To justify this conclusion, authors should test the statistical significance between the siControl and siIFT20 samples at each time point.

The current quantification (Figure 7B) shows that there is, at least, a trend of increased p-VEGFR3, p-VE-cadherin, p-ERK, and p-AKT in IFT20 silenced cells. However, the representative Western Blot image does not display a clear difference (Figure 7A). Authors should include the original western blots, used for quantification, as supplements. 12. The authors use western blot quantification to show that the altered LEC junctions affect VEGFR3 signaling. They further hypothesize that the increased VEGFR3 signaling may be a consequence of VEGFR3 localization in endosomes. The authors did not detect any signal using the phospho-specific VEGFR3 antibody (rows 441-442). To analyze the location of VEGFR3 upon VEGF-C treatment in siControl and siIFT20 LECs, the authors should use anti-VEGFR3 (total) antibodies that have been shown to detect VEGFR3 in similar assays. 13. The normality of the data should be tested before the selection of the statistical test. If this has been done, please, indicate it in the materials and methods or re-run the statistical analysis, if some of the data is not normally distributed. 14. The authors should use arrows, arrowheads, etc. to highlight examples of relevant features in the images. For example, in Figure 3C, the increased stress fiber formation is not obvious to the reader.

Significance

Lymphatics are essential for fluid, leukocyte, and lipid trafficking to lymph nodes and/or systemic circulation. Recent findings have promoted lymphatics as a potential target to control the level of adaptive immunity in inflammation-associated diseases, including tumorigenesis (for example Song et al 2020, Nature). Early work on lymphatic endothelium in vivo, highlighted the dynamics of lymphatic endothelial junction, which, reversibly, can alter between continuous and discontinuous ("button-like") states (Baluk 2007, Am. Jour. Pathol.; Yao 2012, Am J. Pathol.). These changes may have an effect on fluid drainage capacity, lymphatic vessel growth, and prevention of pathogen dissemination to the systemic circulation. Recently, lymphatic junctions have been shown to present hubs of VEGFR3 signaling, VEGFR3 and VE-cadherin dynamics, and leukocyte transmigration (Sung et al. 2022, Nat. Cardiovasc. Res.; Hagerling et al. 2018, EMBO J.; Liaqat et al. 2024, EMBO J.). Thus, the manuscript by Paulson et al. investigates a topical subject.

The authors suggest a role for IFT20 in the control of VE-cadherin dynamics. Based on my expertise in lymphatic endothelial biology, I envision that the manuscript can potentially increase knowledge on the regulators of the lymphatic endothelial junctions, which might have physiological, and in the long term, translational significance. However, in the current manuscript, the exact mechanisms of how IFT20 controls lymphatic endothelial junctions are left open. In addition to the lymphatic research field, the study is, potentially of interest to researchers working on blood vasculature or, even, epithelium, i.e. tissues where junctional dynamics play a major role in health and disease.

Furter controls, analysis, and experimentation are needed to warrant the authors' statements. In their future work, the authors should also consider means to rigorously dissect the IFT20 functions in primary cilia and endosomes.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript identified a novel role of Intraflagellar Transport Protein 20 (IFT20) in the function and homeostasis of lymphatic endothelial junctions. The authors showed that IFT20 regulates VE-cadherin localization at adherens junctions in lymphatic endothelial cells. The authors performed impressive in vivo work that shows the requirement for IFT20 for the homeostasis of intercellular junctions, lymphangiogenesis, and drainage function of lymphatic vessels. In contrast, the cell biology part of the paper was underwhelming and will need significant revisions to support the proposed model. In the result section, several conclusions have to be toned down to match the actual results. The study employs in vivo mouse models, immunofluorescence, biochemical assays, and loss-of-function experiments to support their conclusions.

Major comments

• The authors present disrupted localization of VE-cadherin. Is this a mislocalization and/or protein stability issue in IFT20 KD cells? A western blot can help assess protein levels, and a phase-chase endocytosis assay of VE-cadherin can strengthen evidence. The authors did not confirm the permeability phenotype seen in vivo.
• While the authors focused on IFT20 and rab5, we do not have a clear idea about the vesicular dynamics as well as the status of early, late, and recycling endosomes in IFT20 KD cells. Is IFT 20 localized to non-rab5+ endosomes, and if yes, what are the species? A more general endosomal profiling would help strengthen the authors' message. For example, in Fig. 4-5, the authors will have to stain for other early endosomal markers as well as late, and recycling endosomal markers in control and IFT20 KD cells.
• In fig. 6C, a majority of VE-cadherin is not associated with Rab5. Staining with additional endosomal markers might help identify other endosomal species colocalizing with VE-cadherin. It will be critical to add to Fig. 6c the intensity profiles depicting colocalizations. The authors can also live image a fluorescently (f)-tagged VE-cadherin (maybe with another f-tagged rab5) and assess their association dynamics in IFT20 KD cells (similar to fig6C).
• Primary cilia do seem to regulate vascular plexus in the mouse retina as well as endothelial permeability through mediating subcellular localization of junction proteins. The authors do not clearly exclude the ciliary function of IFT20 in mediating lymphatic endothelial cell-cell junctions. A rescue experiment can help settle this question by targeting IFT20 exclusively to cilia (or not) and assessing, for example, VE-cadherin localization. The following is optional: It is also unclear whether the described regulation is specific to IFT20 or can be phenocopied by the ablation of another IFT subunit and/or cilia ablation through the depletion of a non-IFT cilia assembly regulator.
• Figs. 7A and B do not seem very convincing. The control vs. IFT20 KD western blot levels look mostly similar between the two conditions. The result section does not translate the actual data in Fig. 7A and B. Additionally, there are no statistical comparisons between control and KD conditions in the graphs. Except for a potential pVEGFR-3 increase at 30 min VEGF-C in IFT20 KD cells, but after washout the level is similar to control. This figure does not support well the model presented in fig. 8. The conclusion in lines 456-459 has to be toned down.
• The authors were not able to stain for pVEGFR3. It would still be helpful to see a colocalization between total VEGFR3, IFT20, and VE-cadherin in control cells and IFT20 KD cells (VEGFR3 and VE-cadherin).

Minor comments

• The control used in Figures 1 and 2 does not seem ideal. The proper control would be IFT20fl/fl cre neg. Is there a reason why the authors excluded a lox allele in control? Also, the authors have to provide the mice age used in these figures and when the Cre kicks in in the result section.
• Please describe the overall mouse phenotype(s) of the LYVE1 CRE-IFT20 flox.
• Line 109:'By expression, the authors probably mean immunostained.
• Many graphs exhibit undefined Y-axis labels and units. Please clarify these as well as the way they were quantified. Include such information in figure legends and/or in the materials and methods section. The figures in question are fig1C, E and F, fig2E, fig3E, fig4b and D, fig6b and D, fig7B and C.
• Line 295:'homeostasis"-the authors probably mean in a serum-rich condition.
• Fig4C specifically the lower two images on the right side: the images do not seem to represent the corresponding graphs.
• Please add the statistical tests used to evaluate significance in all figure legends.

Significance

This study provides novel insights into IFT20's role in VE-cadherin trafficking and endothelial junction stability, with its strongest aspect being the in vivo data in Figures 1 and 2, demonstrating lymphatic defects upon IFT20 loss. This represents a conceptual advance by extending IFT protein function beyond cilia (if one of the major comments is addressed) to vascular integrity. However, mechanistic depth is lacking, and ciliary role was not tested-additional rescue and colocalization experiments are needed to confirm the model. The study will interest vascular and lymphatic biologists, as well as cell biologists studying intracellular trafficking and cilia.

Expertise: cilia and mouse genetics

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Reply to the Reviewers

I thank the Referees for their...

Referee #1

1. The authors should provide more information when...

Responses + The typical domed appearance of a hydrocephalus-harboring skull is apparent as early as P4, as shown in a new side-by-side comparison of pups at that age (Fig. 1A). + Though this is not stated in the MS 2. Figure 6: Why has only...

Response: We expanded the comparison

Minor comments:

1. The text contains several...

Response: We added...

Referee #2

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Kemp et al. aimed to explore the transcriptional cell cycle regulation of replication-dependent (RD) histone genes at histone locus body (HLB) in Drosophila. They evaluate the accumulation of RNA pol II and RD histone transcripts at HLB during the cell cycle using live and fixed imaging of Drosophila tissues at different stages of development. They find that RNA pol II is enriched at HLB, not only during S phase when RD histone genes are transcribed but throughout the cell cycle. Outside of S phase, they detect short but not full-length RD histone transcripts suggesting a mechanism of RNA pol II pausing. Full length RD transcripts are only produced upon cyclin E/Cdk2 activation when cells enter S phase, arguing that Cyclin E/Cdk2 can activate transcription elongation. They propose that the elongation release triggered by Cyclin E/Cdk2 is the critical step linking RD histone gene expression and cell cycle progression rather than the recruitment of RNA pol II to HLB. The data are interesting and robust, however, additional experiments could reinforce the findings and the proposed model.

Specific comments/concerns are listed below.

1. In Figure 3, quantifications of the fluorescence at HLBs for mCherry-RBP1 and MXC-mScarlet should be provided.
2. In Figure 5C, both 5' and 3' transcripts are observed in G214 cells. However, their accumulation in the cytoplasm is not visible. How do the authors explain this result? What happens in S14 cells?
3. In Figure 6, the authors observed RD histone 3' transcripts only in replicating cells (EdU positive) while they detected 5' transcripts in both replicating and non-replicating cells. They argue that the appearance of 3' transcripts is due to the release from transcriptional pausing. To further support particular states in the transcriptional arrest, data by immunofluorescence using specific antibodies recognizing either RNA pol II ser5P or ser2P would determine whether the presence of 3' transcripts is associated with the accumulation at HLB of RNA pol II ser2P (elongating polymerase). Moreover, is there a correlation between P-MXC and RNA pol II ser2P?
4. In Figure 7 panels C and D, the 5' transcripts should be shown. Although RD histone 3' transcripts accumulate in CyE+ embryonic cells, unfortunately, their presence at HLBs (pointed by arrows) is not visible in the image of panel E. To firm up conclusions quantifications of the 3' and 5' transcripts should be provided for CycE+ and CycEnull cells. In Hur et al., 2020, the authors looked at RD histone transcripts in WT embryo and CycE+/-/Cdk2+/- mutant. They found that the amount of H3 transcripts using a probe corresponding to the coding sequence is not changed in the mutant as compared to the WT. In contrast, they found that there is an increase of transcripts that are not correctly processed using probes downstream the stem-loop region. This seems inconsistent with the results presented here where a decrease of 3' transcripts is observed. This needs an explanation/discussion. Are such incorrectly processed transcripts observed in CycEnull mutant?
5. The authors suggest that active Cyclin E/Cdk2 triggers the release of RNA pol II promoter-proximal pausing and therefore induces transcriptional elongation at RD histone genes when cells enter S phase. To further support this hypothesis, determining whether there is an enrichment of the elongation factor p-TEFb at HLB when Cyclin E/Cdk2 is active would help.
6. Instead of using cycling E mutants, to determine whether it is the phosphorylation of MXC which directly impacts the elongation of RD histone genes, it would be interesting to generate phospho-null or phospho-mimetic mutant of MXC.
7. In Suzuki et al., 2022, the authors described 3' RNA pol II pausing at RD histone genes. Although this study used human cells, it would be interesting to discuss that in addition to a promoter-proximal pausing that regulates transcription elongation, a 3' pausing could also regulate the transcription termination and 3' processing.
8. In the discussion, the authors should point out some limitations of their studies linked to the method and could propose for the future that a more precise and molecular view of the pausing mechanism could be carried out using sequencing methods such as ChIP-seq of various isoforms of the RNA pol II (total, ser2P, ser5P) and elongation regulators (p-TEFb.....) and PRO-seq.

Minor points:

1. In Figure 1, for panels B and D as well as for panels C and E, to falicitate comparison of the localization of the different proteins, it would help to show the same developmental stages and the same image scales.
2. In Figures 3 and 7 (C-F), the developmental stages should be indicated on the images, as it is done in the other figures.
3. In the legend of Figure 7, it is indicated (D) and (E) instead of (C) and (D) in the sentence: "Endocycling midgut cells in (D) contain cytoplasmic histone mRNA which is absent in (E) (boxed regions)."

Significance

Kemp et al. aimed to explore the transcriptional cell cycle regulation of replication-dependent (RD) histone genes at histone locus body (HLB) in Drosophila. They evaluate the accumulation of RNA pol II and RD histone transcripts at HLB during the cell cycle using live and fixed imaging of Drosophila tissues at different stages of development. They find that RNA pol II is enriched at HLB, not only during S phase when RD histone genes are transcribed but throughout the cell cycle. Outside of S phase, they detect short but not full-length RD histone transcripts suggesting a mechanism of RNA pol II pausing. Full length RD transcripts are only produced upon cyclin E/Cdk2 activation when cells enter S phase, arguing that Cyclin E/Cdk2 can activate transcription elongation. They propose that the elongation release triggered by Cyclin E/Cdk2 is the critical step linking RD histone gene expression and cell cycle progression rather than the recruitment of RNA pol II to HLB.

The data are interesting and robust, however, additional experiments could reinforce the findings and the proposed model.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Kemp et al. seek to define the molecular interactions that limit replication-dependent histone gene transcription to S-phase of the cell cycle. They use the Drosophila model system and leverage live-imaging tools, such as tagged proteins and Jabba trap, and RNA FISH in several tissues to determine that RNA Pol II is enriched at the locus throughout the cell cycle and is paused outside of S-phase. Therefore, they conclude that it is not Pol II recruitment to the locus that couples histone transcription to S-phase, but release of Pol II pausing.

Major comments:

The data presented are clean and well-presented. The claims are supported by the data without exaggeration. It would be helpful to provide -omics support for this entirely image-based analysis (e.g. PRO- or GRO-seq data from synchronized, sorted Drosophila cells may already exist- OPTIONAL).

A major requirement is that the authors make clear in Introduction and Discussion that the observation of Pol Ii pausing at RD histone genes is not novel. This requires, at minimum, a discussion of Liu (2024) and Suzuki (2022). This allows readers to focus on the advance novel to this work, which is specifically the cell cycle coupling of Pol II pausing.

As the authors are claiming different dynamics between Spt6 and RPB1 in Figure 1, they should provide similarly-staged embryos for comparison. For example, the authors should show RPB1 in early/mid S of NC 14, as this is when they see Spt6 variability. In theory, this should be relatively easy as these are stills from the live videos.

Minor comments:

The use of Spt6 live imaging early on was slightly confusing. The authors should consider moving this data later in the results or providing more written justification for why they investigated Spt6 (further than "to further explore the regulation of RNA pol II dynamics... p6). Similarly, Spt6 is included in the model figure, which might be a stretch given the only Spt6 data involves the timing of Spt6 colocalization with Mxc during the cell cycle.

Misleading language/missed citations:

p3: "600 kB array" is misleading. The whole locus is ~ 600 kB.

p3: Mxc may remain at the locus throughout the cell cycle, so the whole HLB does not disassemble (Terzo, 2015).

p4: H1-specific factors include cramped (Gibert and Karch, 2011; Bodner et al. 2024 bioRxiv)

p4: Hodkinson, 2023 is not the correct reference. The correct reference is Hodkinson, 2024, Genetics.

p5: The Drosophila HLB is detectable at NC 10 (White, 2011; Terzo, 2015) not White, 2007

p5: A typo: "imagining"

p7: The section title "RNA pol II is necessary for HLB assembly" is incorrect, as Figure 3 shows that pol II is NOT necessary for Mxc recruitment, but for HLB growth. Mxc, however, is necessary for pol II recruitment.

p9: The authors should clarify what "HisC" means as this is the first usage.

Figures/experiments:

Fig 2: The authors should show the gating in Figure I that led to the three categories in Figure J. The legend/colors in Figure J are not necessary.

An "easy" experiment would be to use the FUCCI cell lines and 5'/3' RNA FISH in combination (assuming fluorophores allow) - OPTIONAL

Discussion:

p13: The reference to the work of Gugliemlmi, 2013 should first come up in the Introduction, as it provides rationale.

p13: "without engaging in transcription" is misleading, as pol II is transcribing, but paused.

p15: It makes sense for pol II to pause at histone genes in G1, as they are preparing for the rapid burst of histone transcription needed in S phase. But what might be the functional rationale for pol II pausing in G2, if the HLB disassembles in M?

Methods:

It should be made clear how embryos were staged for live imaging, as it is likely by timing after cell cycle events. What is this timing? It would be best if this detail is not just mentioned in the methods, but also in the main text. This is especially important for readers not familiar with Drosophila embryogenesis. Please cite/acknowledge DGRC for Fly-FUCCI line (if appropriate)

Significance

This study provides convincing evidence that pol II is enriched at the histone locus and paused outside of S-phase. What limits the significance is that several prior studies concluded that Pol II is paused at the histone locus:

Lu et al. bioRxiv 2024, "Integrator-mediated clustering of poised RNA polymerase II synchronizes histone transcription"

Suzuki et al. Nat Comm 2022, "The 3' Pol II pausing at replication-dependent histone genes is regulated by Mediator through Cajal bodies' association with histone locus bodies"

Neither of these studies is discussed or even cited in the manuscript, which is disappointing. Therefore, the advance is limited to the cell-cycle coupling of pausing. This is still important, as a major knowledge gap as outlined by the authors is that it is not clear how histone transcription is coupled to S-phase and they rule out Pol II pausing as a possible mechanism, and point toward Pol II pausing release.

Moreover, there is also evidence (from these authors) that Mxc phosphorylation is not always coupled to histone transcription in Drosophila ovaries. This work is also not discussed or cited:

Potter-Birriel et al. J Cell Sci 2021, "A region of SLBP outside the mRNA-processing domain is essential for deposition of histone mRNA into the Drosophila egg"

The current research may be of interest to the broad cell cycle field, but it may also be useful as a model for those conducting basic, foundational research who seek to describe how Pol II is released from pausing. The histone locus may be of interest as a novel, facile model for pausing.

Reviewer expertise: Drosophila, chromatin, gene expression

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

__Evidence, reproducibility and clarity __

This is a well-written manuscript that describes a thorough study of the functionality of individual residues of a central component of the ESX-3 type VII secretion system of Mycobacterium smegmatis, EccD3, in the essential role of this protein transport system in iron acquisition. Using the powerful and unbiased approach of deep mutational scanning (DMS), the authors assessed the impact of different mutations on a large number of residues of this component. This carefully executed research highlights the importance of hydrophobic residues at the center the ubiquitin-like domain, specific residues of the linker domain that connects this domain with the transmembrane domains and specific residues that connect EccD3 with the MycP3 component.

Major comments

Since the LOF effects in the iron-sufficient and iron-deficient condition differ less than expected, the differences of the DMS results between these two conditions should be better presented, explained and discussed: 1. The authors discuss: "Of the 270 LOF mutations seen in the iron-deficient condition, 37 (13.7%) were tolerant in the iron sufficient condition, and 39 (14.44%) had strong LOF effects but weak LOF effects in the iron sufficient condition." Do the authors mean that 39 (14.44%) had strong LOF effects in the iron-deficient condition, but weak LOF effects in the iron-sufficient condition. In turn, does this mean that the remaining mutants (71.9%) had similar LOF effects in the two conditions?

We thank this reviewer for their comment and for highlighting a lack of clarity. We have updated the main text to more effectively communicate our point - that 270 mutants had LOF effects in the iron-deficient media. 37 of these 270 mutants were tolerant in the iron-sufficient media. 39 of these 270 mutants had strong LOF effects in iron-deficient media, but were weak LOF in iron-sufficient media. The remaining 124/270 mutants had weak LOF effects in both conditions. The larger point is that removing iron leads to stronger selection - tolerant mutants become LOF, weak LOF become strong LOF. Removing iron pushes mutants at the bounds over the limit.

__ The diagonal shape of the scatter plot in Fig. 2C, which shows the correlation of the Enrich2 scores of all mutants in the two conditions, indicates that the growth of most mutants is affected similarly in these conditions, but in Fig. 2D lower graph, which shows only the Enrich2 scores of missense mutants, there are clear differences between the two conditions. How can this be explained?__

We apologize for any confusion created by this presentation of our data. We hoped to highlight that while effects are largely similar across conditions, there are some differences. As communicated in our first response, 270 out of our ~2700 missense mutations had LOF effects in the iron-deficient condition. 37 of these 270 mutants were tolerant in the iron-sufficient media. 39 of these 270 mutants had strong LOF effects in iron-deficient media, but were weak LOF in iron-sufficient media. The remaining 124 mutations had weak LOF effects in both conditions.

While Figure 2C shows this difference, it is hard to see by nature of using a scatter plot. We have added contours to highlight how our data is distributed. Our density plots in Figure 2D are meant to try to highlight these differences, where the top plot is showing the effects of all missense mutations. Negatively scored mutations represent LOF effects, mutations with scores around 0 are considered tolerant, and the extremely rare scores with positive scores have GOF effects. Our bottom plot specifically zooms into the negatively scored mutations, to show the 270 LOF mutants we discussed. Specifically, we were hoping to highlight the 39 mutations that have strong LOF effects in iron-deficient media (so the purple line scores are more negative), but weak LOF effects in iron-sufficient media (the green line scores are less negative).

__ Regarding the authors' explanation for the observed LOF effects in the permissive condition, "This speaks to the sensitivity of next-generation sequencing compared to the strong differences observed between conditions in phenotypic growth curves." But this sensitivity does not explain the observed large LOF effects but no growth difference in the permissive condition, unless the analysis is less quantitative than expected? Could it be that there is local iron depletion in this mixed culture, causing selection pressure even in the iron-sufficient condition? Moreover, the severity of the growth defect at the time of sampling, i.e., after 24 hours of growth, is unclear. Indeed, the growth curve in Fig. 1 shows that the growth of the double mutant in iron-deficient conditions is significantly impaired at that timepoint. In the growth curve in Fig. 2B (and also slightly in Fig. 2F), however, the growth defect is less pronounced: the double mutant has a similar OD600 as the WT strain, although the error bar is larger. Is this variability between replicates also seen in the DMS analysis? In general, no statistics are shown for the DMS analysis and there is no information on the significance of the observed LOF effects. In addition, the legend should explain how many replicates the DMS data are based on.__

We thank this reviewer for their comment and for highlighting a point of confusion. In addition to increased sensitivity in next generation sequencing compared to our growth curve experiments, our data analysis and variant scoring was performed by comparing growth rates of our mutant strains to our wild type strain. So, any effect on viability or growth rates seen by expression mutant variants will be more notable in our DMS scoring, as they are relative to wild type. In contrast, our growth curves are plotted as the raw OD600 values of each strain. We believe this difference underlies the difference seen in our heatmaps and growth rates.

It is also a relevant and important point that our libraries are grown as mixed cultures, where there is competition over the limited iron in their growth media, as we highlight in our discussion.

While the double mutant does show a stark growth defect at 24 hours in Figure 1 compared to the WT and complement, it grows just as well as those strains in Figure 2B. The growth defect becomes notable after 24 hours. Within this experiment, we observed variability in growth at the 24hr timepoint for the negative control strain, but also selection when compared to the positive control and library growth at later time points. We analyzed our DMS data in accordance with typical methods used in the field (see: https://doi.org/10.1186/s13059-017-1272-5). We include statistics for the DMS analysis as supplemental Figure 1. We apologize for any confusion regarding the figure caption, however in our manuscript we do point out that our library growth in Figure 2B was repeated in triplicate in the figure caption, and the samples collected during that experiment were the ones used to generate the DMS data.

Minor comments

1. Line and page numbering should be added to the manuscript to facilitate the reviewing process.

We have updated our manuscript to include line and page numbering.

__ "Knockout of the entire ESX-3 operon leads to inhibited M. smegmatis growth in a low-iron environment. When individual components of the ESX-3 system are deleted, growth is only available under impaired if the additional siderophore exochelin formyltransferase fxbA is also knocked out20." First, a reference should be added to the first sentence. Second, Siegrist et al. did not exactly show this. They showed that the fxbA/eccC3 double mutant grows slower that the fxbA single mutant. To my knowledge there is no publication showing that single esx-3 component mutants grow as WT in iron-deficient conditions. Do the authors have data demonstrating this? If true, it is surprising that mutating EccD3 has a milder phenotype compared the complete region deletion, as it is a crucial ESX-3 component.__

We apologize for any confusion. We had the relevant reference two lines prior, and have since added it to that sentence as well.

The reviewer is correct that Siegrest et al did not show the effects of just ESX-3 component single deletions. However, Siegrest et al. 2009 demonstrated that deleting the entire ESX-3 operon results in growth similar to the wild type strain in low-iron media. In contrast, the fxbA single knockout exhibits a notable growth defect, and the fxbA/ESX-3 double knockout has an even more severe growth defect. Following the logic that a double knockout is needed to observe a growth defect in low-iron media, Siegrest et al. 2014 demonstrated this also extends to single ESX-3 component knockouts, such as the fxbA/eccD3 double knockout strain. To ensure clarity and accuracy, I will edit the sentence to say "When individual components of the ESX-3 system are deleted, growth is significantly impaired when the additional siderophore exochelin formyltransferase fxbA is also knocked out."

__ Reference to Table 1, should be a reference to Table S1.__

We have updated our manuscript to correct this reference.

__ "Our heatmaps surprisingly reveal residues where substitutions are deleterious specifically in the iron-sufficient condition" Refer here to Fig. S2.__

We have updated our manuscript to include this reference.

__ "In the iron-deficient condition, 6/551 (1.08%) missense mutations have a weak LOF effect, and 0 have strong effects." More clearly explain this refers to the residues of the transmembrane region.__

We have updated our manuscript to provide more clarity.

__ "The MycP transmembrane helix has been hypothesized to be required for ESX complex specificity, targeting MycP to associate with the correct ESX homologue." I miss a reference here. And I thought that the transmembrane domain of MycP was required for complex stability not for specificity?__

We thank the reviewer for pointing out our missing citation, and asking us to clarify our point. I believe the literature suggests that both the protease and transmembrane domains of MycP are required for both complex stability and specificity. van Winden et al. 2016 https://doi.org/10.1128/mbio.01471-16 show that MycP5 needs to be present for secretion. The protease activity can be abolished and the ESX-5 complex can still secrete and be pulled down, as seen by BN-PAGE. van Winden et al. 2019 https://doi.org/10.1074/jbc.RA118.007090 show that truncated mutants missing either the protease domain or the transmembrane domain cannot rescue ESX-5 secretion or complex stability in a MycP knockout strain. More relevant, they attempted to rescue MycP1 and MycP5 mutants by creating chimeric proteins that either had the MycP1 protease domain and MycP5 transmembrane domain, or the MycP5 protease domain and MycP1 transmembrane domain. If the protease and transmembrane domains were required for complex stability and NOT specificity, we would see MycP5 rescue ESX-1 secretion in the MycP1 mutant strains and vice versa. We would also see the chimera proteins rescue both ESX-1 and ESX-5 secretion and complex stability. Instead, we see that neither chimera rescued ESX-1 nor ESX-5 secretion or complex stability, implying that both MycP domains are necessary.

We will amend our paper text to reference MycP's role in complex stability instead of specificity, and soften the language: "The MycP transmembrane helix has been shown to be required for ESX complex stability, as MycP knockouts and truncated mutants abolish ESX secretion and pulldowns of the entire complex."

__ "....role in ESX function relating to EccB3 and EccC3. In the transmembrane, ..... we" Insert "region" after "transmembrane"__

We have updated our manuscript to include this update.

Significance

The study provides insight into individual residues of a central component of the ESX-3 type VII secretion system for functionality, which is useful for those studying the functioning of mycobacterial type VII secretion systems. Moreover, because this system is essential for the growth of the important pathogen M. tuberculosis, this knowledge can be used to design new anti-tuberculosis compounds that block the ESX-3 system. Although the results mainly confirm previous observations (highlighting specific residues important for the stability of ubiquitin and residues of other parts of EccD important for protein-protein interactions within the ESX-3/ESX-5 membrane complex), to my knowledge this is the first time DMS has been applied to mycobacteria. This study is therefore of interest to mycobacteriologists.

---

Reviewer #2

__Evidence, reproducibility and clarity __

This work provides valuable insights into EccD3 function, a core component of the ESX-3 secretion system. The strength of this study lies in the development of a robust functional assay for the systematic mapping of functionally relevant amino acids in EccD3. The approach could potentially be expanded to analyze other ESX-3 components but remains limited to the ESX-3 secretion system. 1. The authors engineered an M. smegmatis knockout strain with deletions of fxbA and eccD3. Deletion of fxbA renders the exocholin iron uptake system non-functional, forcing the bacteria to rely on siderophore-mediated iron uptake under iron-limiting conditions. This process, in turn, depends on ESX-3 secretion activity, as PPE4, a known ESX-3 substrate, has been previously implicated in iron utilization in M. tuberculosis (Tufariello et al., 2016). This experimental setup links EccD3 function to a growth phenotype under iron-limiting conditions, as mutations impairing ESX-3 secretion disrupt iron utilization and mycobacterial growth. 2. By complementing the knockout strain with a library of EccD3 mutant variants, the authors systematically identify residues essential for protein-protein interactions within the ESX-3 core complex. Structural analysis corroborates the functional relevance of these residues, specifically those mediating interactions between EccD3 and other ESX-3 components, or those disrupting the hydrophobic core of the EccD3 ubiquitin-like (Ubl) domain. 3. Structural comparisons with the MycP5-bound ESX-5 complex allow the authors to predict residues within EccD3 that may interact with MycP3 during ESX-3 core complex assembly. Furthermore, comparisons with the ESX-5 hexamer suggest residues that may stabilize or drive oligomerization of the ESX-3 dimer into its putative hexameric state. These insights are significant and provide testable hypotheses for future studies. 4. The methodology is limited to ESX-3. The authors exploit the essentiality of ESX-3 for siderophore-dependent growth under iron-limiting conditions. However, this functional readout cannot be directly transferred to other ESX systems (ESX-1, ESX-2, ESX-4, ESX-5), which have distinct substrates, biological roles, and regulatory mechanisms.

Significance

This work provides valuable insights into EccD3 function, a core component of the ESX-3 secretion system. The strength of this study lies in the development of a robust functional assay for the systematic mapping of functionally relevant amino acids in EccD3. The approach could potentially be expanded to analyze other ESX-3 components but remains limited to the ESX-3 secretion system.

Thank you for your thoughtful and supportive feedback. We appreciate your time and effort in reviewing our study.

---

Reviewer #3

__Evidence, reproducibility and clarity __

The manuscript by Trinidad et al. provides a deep mutational scanning (DMS) analysis to investigate the functional roles of residues from the EccD3 subunit of the Type VII ESX-3 secretion apparatus from M. smegmatis. A previously published structure of ESX-3 from M. smegmatis by the Rosenberg group (Oren Rosenberg is also an author of this paper) is used as basis for structural interpretation of the DMS data presented in this contribution. A shortcoming of the previous structure, despite being very rich in terms of structural details, was in the lack of hexameric pore formation, which has been established more recently by structures of the related ESX-5 system.

Technically, DMS is state-of-the art and a powerful approach to systematically scan residues of potential functional interest. Therefore, the data presented here, provide a remarkable repository for further interpretation in this contribution and by other future investigations. The experimental data have been deposited in Github enabling access by others in the future.

Overall, the paper would benefit from an improved overall organisation. I found in part hard to extract some of the main points from the way the data are presented. In essence, two separate screens were performed, the first one focusing on the EccD3 Ubl domain and adjacent linker regions and a second one on the EccD3 TM region. I think the paper could be better structured accordingly. Tables of residues with strong effects in iron-deficient and iron-sufficient media, together with their structural annotation, would facilitate extracting main messages from this manuscript. Without going too much in detail, there is also scope for improvement of most of the structural figures. More consistency in terms of color coding with the previous paper by Powileit et al. (2019) would also help navigation.

A potential weakness of the paper is in the limited scope of interpretation of the data in the context of the dimeric ESX-3 assembly, which is actually acknowledged by the authors. Computational AI-based methods should allow generating a complete pore model of ESX-3, which would allow interpretation of some of the data in a more functional relevant context. This would enhance the validity of the current interpretations presented.

We acknowledge the lack of a hexameric ESX-3 structure, and would love to base our analysis on such a structure. Unfortunately, experimentally purifying and determining such a structure is beyond the scope of this manuscript. While AI-based methods are certainly exciting and helpful to make sense of mutational data, they are not able to computationally predict such large structures. The AlphaFold3 server website is commonly used for these purposes and allows predictions of up to 5000 tokens (or amino acids). An ESX-3 hexamer would be composed of 6x EccB proteins (519 AA each), 6x EccC proteins (1326 AA each), 12x EccD proteins (476 AA each), and 6x EccE proteins (310 AA each). Together, this complex would be made up of 18,642 amino acids.

We tried using alphafold to predict an ESX-5 dimer complex, as well as reproduce the ESX-3 dimer complex, and were unable to produce these structures. Each ESX protomer is assembled correctly, as each protein within the complex makes appropriate contacts with each other. We see the EccD-dimers still form the membrane vestibule within each ESX complex. The issue is the ESX dimer complex has not assembled correctly: the EccC transmembrane helix 1 of a protomer should interact with the EccB transmembrane helix of the neighboring protomer; and, the N-terminus of EccB in one protomer should interact with the loop between the EccD transmembrane helices 10 and 11 in the neighboring protomer. Instead, Alphafold creates contacts along the EccD proteins from both complexes. We have included a "top-down" view of the ESX-5 dimer, where the periplasmic domains of EccB have been cleaved off for clarity.

A side view:

Here we have the ESX-3 dimer structure published by Poweleit et al. side-by-side with the ESX-3 dimer predicted by alphafold, visualized in Pmyol. The alphafold structure largely has each proteins' domains and folds properly predicted, including even the EccD3 dimer found in each ESX protomer. However, the protomers are not assembled into a dimer properly as compared to the purified ESX-3 dimer from PDB: 6umm. We included a "front" and "side view", as well as a "top down" view where the cytoplasmic domains have been hidden for visual clarity.

The use of full names and acronyms needs to be more consistent. As an example, the terms "ubiquitin-like" and ubiquitin-like (Ubl) and UBl are used in parallel throughout the manuscript. The percentages given in various places of the paper could be reduced to integers, as they generally relate to relatively small data sets. Please express numbers with a precision, reasonable matching expected statistical significance.

We apologize for the lack of consistency in how we referred to the ubiquitin-like domain. I originally wrote "ubiquitin-like (Ubl)" once per section (intro, results, discussion). I have edited these all to just "Ubl" after the introduction, except for figure and section titles. We have also reduced our percentages to integers.

Some of the DMS experiments have been repeated three-fold, which should be a minimal number to allow extracting statistical significance, other experiments have only been repeated two-fold. Could this be clarified, please?

We apologize for this oversight, and thank the reviewer for pointing this out. All experiments were done in triplicate, the exception being the site-directed mutant growth curves, which were performed in duplicate. We have repeated this experiment in triplicate in response to this point. As we repeated this experiment, mutant R134A dropped out due to technical reasons, and so we did not include it in the updated growth curves.

Specific comments on text and figures:

Figure 1: The EM densities shown considerably deviate from those that were shown in the original publication by Poweleit et al (2019). If there is an aim is to reinterpret the data this needs to be described in sufficient technical detail. There may be a case for this, in light of recent advances in computational AI-vased structural biology.

We acknowledge this may be confusing and we apologize for that, as the EM density I have shown in this manuscript uses the same map we used to create the one seen in the original publication Poweleit et al 2019. There are existing crystal structures of EccB1 and the ATPase domains of EccC1 that we used to create homology models of EccB3 and EccC3 using the structure-prediction software RaptorX for the 2019 publication. These homology models were then combined with a low resolution EM density to create the model seen in the 2019 eLife paper. I did not include those homology models in this manuscript, as I did not believe those predictions were relevant to this study. I wanted to include the highest resolution and thus most accurate depiction of our ESX-3 structure.

Introduction, statement "We made comparisons to a prior DMS on ubiquitin to increase signal-to-noise in our interpretation of the Ubl domain mutagenesis data." Could this be further explained please? I could not find anything in addition in the Methods section and elsewhere.

__ __We apologize for the confusion!

EccD3 Ubl domain and ubiquitin DMS dataset comparisons

To compare the DMS data of EccD3 Ubl with that of ubiquitin, we first identified homologous residues in each structure. This was achieved by aligning the EccD3 Ubl domain with ubiquitin (PDB: 1ubq) using PyMOL and assessing the positional correspondence of side chains (e.g., ubiquitin residue I3 aligned with EccD3 residue V12). Next, we referenced missense mutation datasets to calculate the average DMS score for each residue position in both proteins. We then generated a scatter plot to compare the average missense scores for ubiquitin and EccD3 Ubl using ggplot2. Data points were color-coded according to the functional roles assigned to ubiquitin, with residues forming the hydrophobic patch and core highlighted, while all other residues were represented in grey.

Description of "vestibule" as a core feature of the ESX-3 structure. As mentioned above, this is very much a result of the presented dimeric arrangement. In the context of a complete pore model, these features may change or even disappear.

While we would certainly welcome an ESX-3 hexamer model to definitively determine whether this feature persists, such a model is not currently available. However, the highly homologous ESX-5 complex retains these EccD vestibules, and there is no reason to believe these features would change or disappear. Therefore, based on our interpretation of the ESX-3 dimer and ESX-5 hexamer we believe that the EccD membrane vestibule is not just an artifact of the ESX-3 dimer complex.

It is possible that the reviewer misunderstood what we were referring to as the vestibule. We updated the language in the text to improve clarity. However the vestibule is not a consequence of ESX-3 complex dimer formation. It is an inherent feature of the ESX monomer complexes, where two EccD proteins dimerize to form said vestibule. Furthermore, there is no evidence to suggest that this feature would be lost in a hexameric state.

Structurally, the ESX-3 dimer consists of two ESX-3 monomer complexes, each containing one EccB, one EccC, one EccE, and two EccD proteins. Therefore, each ESX-3 monomer inherently includes an EccD dimer. The presence of the EccD dimer is not exclusive to the ESX-3 dimer but is a fundamental component of each ESX-3 complex. Similarly, the ESX-5 hexamer retains the EccD dimer within each ESX-5 complex, further supporting the idea that this structural feature is conserved.

Figure 2, panel B: Isn't right that "positive" and "negative" need to exchanged? Perhaps, there is something I misunderstood.

We apologize for the confusion, and appreciate the reviewer pointing out this inconsistency. We have updated the manuscript to correct this.

Figure 2, panel F: it is hard to extract the assignments from the overlaid curves.

We apologize for a lack of clarity in how this growth curve was presented. We have included labels at the end point to show where each sample is.

Figure 3, caption "from low (red) to white (tolerant)": for the sake of consistency, please either put the color in parentheses, or functional description. Does this statement relate to panel A or B? "All other residues are colored white". I can't see this.

We apologize for the inconsistency, and have updated this label. We hope we have clarified the fact that the entire structure is white except for the residues we colored red.

Results text "In contrast to ubiquitin, all hydrophobic core residues in the EccD3 Ubl domain are equally intolerant to charged residue swaps. Unsurprisingly, residues important for ubiquitin's specific degradation interactions are not sensitive to substitutions in the EccD3 Ubl domain." Does this mean that proper folding of Ubl is less critical for ESX_3 function? Please elaborate on this further.

We apologize for any confusion. Our data shows that residues which side chains extend into the hydrophobic core of the Ubl domain are intolerant to swaps to charge residues. We hypothesize these missense mutations disrupt this hydrophobic core, and lead to destabilization of this domain. These intolerant missense mutations each have negative Enrich2 scores, implying a loss of ESX-3 function, and that proper folding of the Ubl is critical for ESX-3 function. We have updated our text to clarify this point:

Unsurprisingly, residues important for ubiquitin function's specific interactions are not sensitive to substitutions in the EccD3 Ubl domain. There is no simple discernable preference within the Ubl domain to any side that maintains protein-protein interactions, implying that the scores are dominated by stability effects and that the Ubl domain must maintain a stable β-grasp fold for ESX-3 function.

Figure 4, panel C: the surface does not provide residue-specific information, hence this panel is not very informative.

We agree with the reviewer that Figure 4 panel C was not very informative, and so we have removed it from Figure 4 for the sake of brevity.

Results text "T148 extends out from transmembrane helix 1 into a hydrophobic pocket between transmembrane helices 1, 2, and 3." Could this please be illustrated in one of the structural presentations?

We have updated figure 5 to include a snapshot of this residue and the hydrophobic pocket it extends into, as panel E.

Results text, last paragraph, Figure 5C-D: interpretation of the experimental ESX-3 data based on ESX-5 models is problematic, without showing proof of conservation of relevant sequence/structural features. As mentioned above, I would encourage the authors to establish a hexameric ESX-3 model and interpret the data starting from there. Extrapolation of the interpretation of data to other ESX systems, including ESX-5, would expand the scope by generalization, which however would open another chapter. The ESX-5 structure does not explain e.g. why W227 when mutated is less sensitive to iron depletion as opposed to iron being present.

We do not believe we can use AI to predict a hexameric ESX-3 model. We will update our supplement to include a figure showing proof of conservation between the EccD3 and EccD5 sequences. We can superpose the ESX-3 dimer structure onto the ESX-5 hexamer structure, and see that this dimeric complex overlays quite well on top of an ESX-5 subcomplex. We can imagine this hexamer as a trimer of dimers, where three copies of this dimeric complex interact to form the hexamer. The superposition is not perfect and there are slight rearrangements to different helices to allow for hexamer formation, but these do not imply we cannot compare these two homologous structures.

We have included a new structure snapshot in Figure 5, where panel D is the ESX-3 dimer (PDB: 6umm) shown as a side and top-down view. This allows for a comparison with panel C, the snapshot of the ESX-5 complex (PDB: 7np7) where in two protomers the EccB, EccC, and EccD proteins are colored the same way as ESX-3, and the other ESX-5 protomers are colored white. Note that in this hexamer, EccE is missing. We see the EccD membrane vestibule is conserved in both structures.

Significance

Strength and Limitations: already assessed under "Evidence, reproducibility and clarity".

There is scope for further interpretation using experimental structural and modeling data. There is also scope for applying complementary assays for selected mutants, most likely within a lower throughput format.

Advance: The contribution demonstrates well the power of DMS for systematic screening, in the context of Type VII secretion. The main advance is in the raw data generated and deposited.

Audience: microbiology with a specific interest in secretion, structural biology

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Trinidad et al. provides a deep mutational scanning (DMS) analysis to investigate the functional roles of residues from the EccD3 subunit of the Type VII ESX-3 secretion apparatus from M. smegmatis. A previously published structure of ESX-3 from M. smegmatis by the Rosenberg group (Oren Rosenberg is also an author of this paper) is used as basis for structural interpretation of the DMS data presented in this contribution. A shortcoming of the previous structure, despite being very rich in terms of structural details, was in the lack of hexameric pore formation, which has been established more recently by structures of the related ESX-5 system.

Technically, DMS is state-of-the art and a powerful approach to systematically scan residues of potential functional interest. Therefore, the data presented here, provide a remarkable repository for further interpretation in this contribution and by other future investigations. The experimental data have been deposited in Github enabling access by others in the future.

Overall, the paper would benefit from an improved overall organisation. I found in part hard to extract some of the main points from the way the data are presented. In essence, two separate screens were performed, the first one focusing on the EccD3 Ubl domain and adjacent linker regions and a second one on the EccD3 TM region. I think the paper could be better structured accordingly. Tables of residues with strong effects in iron-deficient and iron-sufficient media, together with their structural annotation, would facilitate extracting main messages from this manuscript. Without going too much in detail, there is also scope for improvement of most of the structural figures. More consistency in terms of color coding with the previous paper by Powileit et al. (2019) would also help navigation.

A potential weakness of the paper is in the limited scope of interpretation of the data in the context of the dimeric ESX-3 assembly, which is actually acknowledged by the authors. Computational AI-based methods should allow generating a complete pore model of ESX-3, which would allow interpretation of some of the data in a more functional relevant context. This would enhance the validity of the current interpretations presented.

The use of full names and acronyms needs to be more consistent. As an example, the terms "ubiquitin-like" and ubiquitin-like (Ubl) and UBl are used in parallel throughout the manuscript. The percentages given in various places of the paper could be reduced to integers, as they generally relate to relatively small data sets. Please express numbers with a precision, reasonable matching expected statistical significance.

Some of the DMS experiments have been repeated three-fold, which should be a minimal number to allow extracting statistical significance, other experiments have only been repeated two-fold. Could this be clarified, please?

Specific comments on text and figures:

Figure 1: The EM densities shown considerably deviate from those that were shown in the original publication by Poweleit et al (2019). If there is an aim is to reinterpret the data this needs to be described in sufficient technical detail. There may be a case for this, in light of recent advances in computational AI-vased structural biology.

Introduction, statement "We made comparisons to a prior DMS on ubiquitin to increase signal-to-noise in our interpretation of the Ubl domain mutagenesis data." Could this be further explained please? I could not find anything in addition in the Methods section and elsewhere.

Description of "vestibule" as a core feature of the ESX-3 structure. As mentioned above, this is very much a result of the presented dimeric arrangement. In the context of a complete pore model, these features may change or even disappear.

Figure 2, panel B: Isn't right that "positive" and "negative" need to exchanged? Perhaps, there is something I misunderstood.

Figure 2, panel F: it is hard to extract the assignments from the overlaid curves.

Figure 3, caption "from low (red) to white (tolerant)": for the sake of consistency, please either put the color in parentheses, or functional description. Does this statement relate to panel A or B? "All other residues are colored white". I can't see this.

Results text "In contrast to ubiquitin, all hydrophobic core residues in the EccD3 Ubl domain are equally intolerant to charged residue swaps. Unsurprisingly, residues important for ubiquitin's specific degradation interactions are not sensitive to substitutions in the EccD3 Ubl domain." Does this mean that proper folding of Ubl is less critical for ESX_3 function? Please elaborate on this further.

Figure 4, panel C: the surface does not provide residue-specific information, hence this panel is not very informative.

Results text "T148 extends out from transmembrane helix 1 into a hydrophobic pocket between transmembrane helices 1, 2, and 3." Could this please be illustrated in one of the structural presentations?

Results text, last paragraph, Figure 5C-D: interpretation of the experimental ESX-3 data based on ESX-5 models is problematic, without showing proof of conservation of relevant sequence/structural features. As mentioned above, I would encourage the authors to establish a hexameric ESX-3 model and interpret the data starting from there. Extrapolation of the interpretation of data to other ESX systems, including ESX-5, would expand the scope by generalization, which however would open another chapter. The ESX-5 structure does not explain e.g. why W227 when mutated is less sensitive to iron depletion as opposed to iron being present.

Referee cross-commenting

I especially second the comments of referee #1, major comments, point 3 (statistical significance of the data). Addressing this point is crucial for the paper. Referee #2, significance section "The approach could potentially be expanded to analyze other ESX-3 components but remains limited to the ESX-3 secretion system." I was considering making the same point but did not at the end. Of course, ultimately, it would be great if all components of ESX-3 could be analyzed they way it was done for the EccD3 component. However, I am afraid such exercise could become quite open ended. Already by now, there is some compromise on the depth of mechanistic interpretation in light of a large data set generated.

Significance

Strength and Limitations: already assessed under "Evidence, reproducibility and clarity".

There is scope for further interpretation using experimental structural and modeling data. There is also scope for applying complementary assays for selected mutants, most likely within a lower throughput format.

Advance: The contribution demonstrates well the power of DMS for systematic screening, in the context of Type VII secretion. The main advance is in the raw data generated and deposited.

Audience: microbiology with a specific interest in secretion, structural biology

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This work provides valuable insights into EccD3 function, a core component of the ESX-3 secretion system. The strength of this study lies in the development of a robust functional assay for the systematic mapping of functionally relevant amino acids in EccD3. The approach could potentially be expanded to analyze other ESX-3 components but remains limited to the ESX-3 secretion system.

1. The authors engineered an M. smegmatis knockout strain with deletions of fxbA and eccD3. Deletion of fxbA renders the exocholin iron uptake system non-functional, forcing the bacteria to rely on siderophore-mediated iron uptake under iron-limiting conditions. This process, in turn, depends on ESX-3 secretion activity, as PPE4, a known ESX-3 substrate, has been previously implicated in iron utilization in M. tuberculosis (Tufariello et al., 2016). This experimental setup links EccD3 function to a growth phenotype under iron-limiting conditions, as mutations impairing ESX-3 secretion disrupt iron utilization and mycobacterial growth.
2. By complementing the knockout strain with a library of EccD3 mutant variants, the authors systematically identify residues essential for protein-protein interactions within the ESX-3 core complex. Structural analysis corroborates the functional relevance of these residues, specifically those mediating interactions between EccD3 and other ESX-3 components, or those disrupting the hydrophobic core of the EccD3 ubiquitin-like (Ubl) domain.
3. Structural comparisons with the MycP5-bound ESX-5 complex allow the authors to predict residues within EccD3 that may interact with MycP3 during ESX-3 core complex assembly. Furthermore, comparisons with the ESX-5 hexamer suggest residues that may stabilize or drive oligomerization of the ESX-3 dimer into its putative hexameric state. These insights are significant and provide testable hypotheses for future studies.
4. The methodology is limited to ESX-3. The authors exploit the essentiality of ESX-3 for siderophore-dependent growth under iron-limiting conditions. However, this functional readout cannot be directly transferred to other ESX systems (ESX-1, ESX-2, ESX-4, ESX-5), which have distinct substrates, biological roles, and regulatory mechanisms.

Significance

This work provides valuable insights into EccD3 function, a core component of the ESX-3 secretion system. The strength of this study lies in the development of a robust functional assay for the systematic mapping of functionally relevant amino acids in EccD3. The approach could potentially be expanded to analyze other ESX-3 components but remains limited to the ESX-3 secretion system.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This is a well-written manuscript that describes a thorough study of the functionality of individual residues of a central component of the ESX-3 type VII secretion system of Mycobacterium smegmatis, EccD3, in the essential role of this protein transport system in iron acquisition. Using the powerful and unbiased approach of deep mutational scanning (DMS), the authors assessed the impact of different mutations on a large number of residues of this component. This carefully executed research highlights the importance of hydrophobic residues at the center the ubiquitin-like domain, specific residues of the linker domain that connects this domain with the transmembrane domains and specific residues that connect EccD3 with the MycP3 component.

Major comments

Since the LOF effects in the iron-sufficient and iron-deficient condition differ less than expected, the differences of the DMS results between these two conditions should be better presented, explained and discussed:

1. The authors discuss: "Of the 270 LOF mutations seen in the iron-deficient condition, 37 (13.7%) were tolerant in the iron sufficient condition, and 39 (14.44%) had strong LOF effects but weak LOF effects in the iron sufficient condition." Do the authors mean that 39 (14.44%) had strong LOF effects in the iron-deficient condition, but weak LOF effects in the iron-sufficient condition. In turn, does this mean that the remaining mutants (71.9%) had similar LOF effects in the two conditions?
2. The diagonal shape of the scatter plot in Fig. 2C, which shows the correlation of the Enrich2 scores of all mutants in the two conditions, indicates that the growth of most mutants is affected similarly in these conditions, but in Fig. 2D lower graph, which shows only the Enrich2 scores of missense mutants, there are clear differences between the two conditions. How can this be explained?
3. Regarding the authors' explanation for the observed LOF effects in the permissive condition, "This speaks to the sensitivity of next-generation sequencing compared to the strong differences observed between conditions in phenotypic growth curves." But this sensitivity does not explain the observed large LOF effects but no growth difference in the permissive condition, unless the analysis is less quantitative than expected? Could it be that there is local iron depletion in this mixed culture, causing selection pressure even in the iron-sufficient condition? Moreover, the severity of the growth defect at the time of sampling, i.e., after 24 hours of growth, is unclear. Indeed, the growth curve in Fig. 1 shows that the growth of the double mutant in iron-deficient conditions is significantly impaired at that timepoint. In the growth curve in Fig. 2B (and also slightly in Fig. 2F), however, the growth defect is less pronounced: the double mutant has a similar OD600 as the WT strain, although the error bar is larger. Is this variability between replicates also seen in the DMS analysis? In general, no statistics are shown for the DMS analysis and there is no information on the significance of the observed LOF effects. In addition, the legend should explain how many replicates the DMS data are based on.

Minor comments

1. Line and page numbering should be added to the manuscript to facilitate the reviewing process.
2. "Knockout of the entire ESX-3 operon leads to inhibited M. smegmatis growth in a low-iron environment. When individual components of the ESX-3 system are deleted, growth is only available under impaired if the additional siderophore exochelin formyltransferase fxbA is also knocked out20." First, a reference should be added to the first sentence. Second, Siegrist et al. did not exactly show this. They showed that the fxbA/eccC3 double mutant grows slower that the fxbA single mutant. To my knowledge there is no publication showing that single esx-3 component mutants grow as WT in iron-deficient conditions. Do the authors have data demonstrating this? If true, it is surprising that mutating EccD3 has a milder phenotype compared the complete region deletion, as it is a crucial ESX-3 component.
3. Reference to Table 1, should be a reference to Table S1.
4. "Our heatmaps surprisingly reveal residues where substitutions are deleterious specifically in the iron-sufficient condition" Refer here to Fig. S2.
5. "In the iron-deficient condition, 6/551 (1.08%) missense mutations have a weak LOF effect, and 0 have strong effects." More clearly explain this refers to the residues of the transmembrane region.
6. "The MycP transmembrane helix has been hypothesized to be required for ESX complex specificity, targeting MycP to associate with the correct ESX homologue." I miss a reference here. And I thought that the transmembrane domain of MycP was required for complex stability not for specificity?
7. "....role in ESX function relating to EccB3 and EccC3. In the transmembrane, ..... we" Insert "region" after "transmembrane"

Significance

The study provides insight into individual residues of a central component of the ESX-3 type VII secretion system for functionality, which is useful for those studying the functioning of mycobacterial type VII secretion systems. Moreover, because this system is essential for the growth of the important pathogen M. tuberculosis, this knowledge can be used to design new anti-tuberculosis compounds that block the ESX-3 system. Although the results mainly confirm previous observations (highlighting specific residues important for the stability of ubiquitin and residues of other parts of EccD important for protein-protein interactions within the ESX-3/ESX-5 membrane complex), to my knowledge this is the first time DMS has been applied to mycobacteria. This study is therefore of interest to mycobacteriologists.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02655

Corresponding author(s): Thierry SOLDATI

1. General Statements [optional]

The emergence of powerful model organisms for infection studies accelerates discoveries in innate immunity and conserved cell-autonomous defence mechanisms. Using the genetically tractable Dictyostelium discoideum/Mycobacterium marinum infection platform, we explored the critical interplay between pathogen-induced membrane damage and host repair pathways.

Recent findings highlight evolutionarily conserved membrane repair pathways as crucial for cellular integrity against both sterile and pathogenic insults. We previously demonstrated the involvement of ESCRT and autophagy machineries in repairing membrane damage and containing pathogenic mycobacteria within vacuoles. Crucially, we uncovered that TrafE, an evolutionarily conserved TRAF-like E3 ubiquitin ligase, coordinates these machineries to repair membrane damage, preventing cell death.

Here, we reveal that pathogenic mycobacteria manipulate host membrane microdomain scaffolding proteins and sterols to enhance toxin activity and facilitate bacterial escape. Genetic knockout of these microdomain organizers and sterol depletion significantly reduce membrane damage and bacterial escape, effectively containing mycobacteria and increasing host resistance. The conserved roles of flotillin and sterols are confirmed in murine microglial cells, underscoring evolutionary conservation.

These discoveries significantly advance understanding of intracellular host-pathogen interactions, offering broad implications for cellular microbiology and immunology and have already attracted wide interest at major international scientific meetings.

Thanks to the constructive criticisms and suggestions of the referees, we were able to significantly enhance the manuscript by integrating novel experimental strategies and improving presentation and discussion of previous results that together further strengthen our evidence.

2. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The proposed study aims to elucidate the role of membrane microdomains and associated proteins-Vacuolin A, B, and C-during the infection of Dictyostelium discoideum (Dd) amoebae by Mycobacterium marinum (Mm). The results demonstrate that Vacuolins are required for Mm virulence, and that the presence of membrane microdomains is essential for phagosome membrane damage and bacillary escape into the cytosol-key steps in establishing a successful infection and subsequent bacterial proliferation. The study is well-designed, employing methodologies with which the authors have demonstrated expertise. Overall, it is methodologically sound, and most conclusions are well-supported by the presented data. However, some points require clarification.

We thank the referee for their positive evaluation of the scope and strengths of our manuscript. The constructive criticisms of the referees were important to guide our revisions. We are convinced that the new data now integrated further strengthen our evidence.

Major Points:

The study aims to link the function of Dd Vacuolins to their potnetial facilitating role in phagosome escape and overall infection by Mm. To phenocopy the effect of Vac-KO, the authors used MβCD. Strikingly, this compound had a more significant impact on phagosome escape compared to Vac-KO, which either did not affect or only mildly affected this process. This likely reflects a difference in the underlying mechanisms being studied. Vac-KO cells may lack well-organized membrane domains but could retain a similar overall membrane composition. In contrast, MβCD disrupts these domains by chelating cholesterol, thus altering both the membrane composition and the domains themselves. This may explain why EsxA partitioning is more affected by MβCD than by triple KO. Consequently, this suggests that cholesterol, rather than the mere presence of membrane domains, plays a critical role in EsxA partitioning and activity in the phagosome. And even if LLOMe activity was lower in Vac-KO cells, this might be explained by the compartment targeted, the lysosomes which membrane composition may differ from the MCV. These points should be further discussed in the discussion section.

The referee is right on target, these are all excellent points, and we fully agree with the argumentation. If we compare EsxA to a cholesterol-dependent PFT such as SLO, the presence of sterol is an absolute requirement for pore formation, but the local concentration of sterols achieved via clustering and the organisation of lipids/sterols in microdomains "only" increases efficiency (see for example PMID: 39835825). Therefore, the respective impacts of vac-KO and CD treatment differ in "intensity", and are additive in most assays, but are not resulting from "different underlying mechanisms". The simplest and most plausible interpretation of the combined results is that EsxA requires a threshold of local concentration/clustering of sterols to act and vacuolins/flotillins is one of the means to achieve it. In other words, membrane composition inhomogeneities exist in physiological membranes, particularly sterol and sphingolipid clustering in rafts, and microdomain organisers probably regulate their size and dynamics. Without vacuolin/flotillin, these inhomogeneities persist. Only when sterol is depleted and/or redistributed, do they disappear. In brief, the local sterol concentration is the trigger for EsxA preferential partitioning and activity, and many factors besides microdomain organisers influence it.

The second interesting point is that LLOMe is a lysosomotropic membrane damaging agent, whereas EsxA targets the MCV membrane. We have already documented that the MCV has some endo-lysosomal properties and potentially resembles most the "post-lysosomal" compartment, characterized by a mildly acidic pH (pH ~6), the presence of Rab7 and zinc, ammonium and cupper transporters, for example. Our experiments also show that LLOMe is active in the whole endo-lysosomal pathway, including these post-lysosomes (PMID: 30596802, PMID: 37070811). The exact lipid composition of the MCV and post-lysosomes is not known, but both accumulate sterols in a similar manner. Both compartments are also akin to multivesicular bodies. These data are no direct proof but are compatible with our conclusions that both LLOMe and EsxA require similar threshold of local sterol concentration and that vacuolins are a mean to achieve this.

The presentation of these conclusions has been revised and enhanced in the discussion (for example lines 396-400 and 437-439).

Despite these similarities between LLOMe and EsxA activities, note that the mature MCV can be distinguished from all other endo-lysosomal compartments by the use of a Flipper probe that is sensitive to lipid composition and packing (Fig. 7C, and see below). In addition, RNAseq analyses of the impact of vac-KO and sterol depletion on infected and non-infected cells also highlight the interdependence between sterol concentration and vacuolin expression (Fig. 3G, 4G and H, Fig. EV5 and 6, and see below).

Based on this observation, in figure 2, does the D4H/filipin signal or association increase over time as the Vac signal "solidifies"? In Vac-KO cells, does the mScarlet-D4H signal change in intensity or pattern (building on fig. S1)? These insights could provide valuable information on cholesterol levels at the MCV in KO versus wild-type cells. If possible, the authors should quantify fluorescence or the frequency of signal association.

Qualitatively, sterols, as visualised by filipin and D4H, are present at all stages of the endo-lysosomal pathway and of MCV biogenesis. Now, there are many technical difficulties linked to a quantitative assessment, and therefore, please, let me present the framework. First, despite their wide use, the exact mechanism of binding of both reporters and which pool of sterol they visualise is still a mystery. This is often expressed as "they detect the accessible pool" of sterol, whatever it is. In addition, filipin detects sterols in both leaflets (and in intra-lumenal vesicles and other lipidic structures), while D4H detects sterols only in the cytosolic leaflet, and it is not known whether both leaflets have the same concentration of sterols. It is also known that filipin signal is only indirectly proportional to the sterol quantity in a cell, as measured by other quantitative methods. One of the best examples comes from studying the cellular phenotype of Niemann-Pick Type C disease, because many publications report a strong increase of filliping staining, whereas lipidomic analyses show at best a two-fold increase in cholesterol in NPC deficient cells. Moreover, technically speaking, D4H is a live probe, and fixation leads to some loss of localisation, probably because sterols are not fixable. On the other hand, filipin is mainly used after chemical fixation, but again sterols are not fixable, and the signal is very likely restricted to the membrane of origin, but not necessarily to the microdomains.

All this to admit that, despite numerous and rigorous tentatives, we have not been able to reliably obtain quantitative measurements of neither filipin nor D4H signals. Also, these features likely also explain why we were not able to document changes in "patterns" of signals during MCV maturation. We ask for the referee's indulgence about this. Vacuolins remain the best microdomain morphology reporters.

We nevertheless present additional qualitative D4H and VacC colocalization images in Fig. EV1C.

Additionally, since Vacuolins do not have a significant impact on phagosome damage or escape, the difference in intracellular growth may be indirect, as suggested in the team's previous study on Vacuolins (DOI: 10.1242/jcs.242974). The authors measured MCV pH in figure S6-could they repeat this experiment to test whether Vacuolins affect MCV maturation? This was investigated in a previous version of the pre-print (DOI: 10.1101/2021.11.16.468763), and if the results still hold, it would strengthen the hypothesis that Vacuolins promote escape by modulating membrane organization, rather than influencing phagosome maturation.

First, we respectfully disagree that vacuolins have no impact on membrane damage, we explained above why this impact is limited, but nevertheless additive with sterol depletion in most assays, during infection and sterile damage.

We thank the referee for their excellent knowledge of the literature. Indeed, we previously went to extreme experimental sophistication to interrogate the impact of vac-KO on endo-phagosomal maturation. We were able to demonstrate that the major impact is on the recycling of phagocytic receptors and therefore on the cytoskeleton- and motor-induced deformation of the membrane in a cup that is essential for efficient phagocytosis (but not macropinocytosis). We also demonstrated a minimal effect on maturation, on the kinetics of pH change and delivery/recycling of hydrolases, but these cell biological differences did not translate in an impact on bacteria killing and digestion. As mentioned above, the MCV shares characteristics with post-lysosomes but minimal alterations of endo-lysosomal maturation in vac-KO cells should not be responsible for the strong effect on Mm infection. In other words, we are convinced that these minimal (mainly loss-of-function) perturbations that do not impact killing of food bacteria do not lead to an increased phagosomal "ferocity" and restriction of tough mycobacteria.

Consequently, we decided not to repeat experiments to measure the pH around wt Mm in vac-KO cells, as it is anyway only slightly and transiently acidified in wt host cells, and previous work did not reveal major differences in endolysosomal compartment pH control (PMID: 32482795). But we agree with the referee that some of the MCV maturation data presented in the previous bioRxiv version are interesting for specialists, despite the indications of extremely small alterations between wt and vac-KO host cells. These data document that in absence of vacuolins, MCV characteristics are slightly altered, but we found no indication that they are more bactericidal in vac-KO cells (Fig. EV8F-H).

Finally, as a substantial part of this manuscript relies on microscopy and image analysis, the methods section should detail how these analyses were performed. Specifically, for figure 1f, it is unclear how the cells were segmented and fluorescence quantified-was total fluorescence per cell measured, or was an average value used? Figures 5c and 5h could be moved to the supplementary material, and the segmentation method should be explained in the methods section. Additionally, statistical analysis should be more clearly described, justifying the use of one-way or two-way ANOVA, and specifying the post-hoc tests used for group comparisons.

We fully agree with the referee and have therefore improved the detailed description of image analyses. For example, details for cell segmentation in images originating from infection and LLOMe experiments are succinctly described in the Materials & Methods section (lines 585-588, 594-597, and 639-640), but we now also refer to a methods chapter in press that describe in detail the whole segmentation pipeline (Perret et al. 2025).

Concerning specifically Fig. 1F, we distinguished infected or bystander cells by the presence of bacteria and quantitated the maximal fluorescence intensity for each cell. Then, we decided on an arbitrary threshold of intensity of 5,000, that corresponds to the maximal signal observed for cells in mock conditions. Then, we quantified the percentage of bystander and infected cells with a higher-than-threshold (>5,000) vacuolin signal intensity. This clarification is now added to the legend of Fig. 1F.

The statistical analyses applied are described in more detail in each figure legend.

Reviewer #1 (Significance (Required)):

This study provides the first direct evidence of the importance of membrane composition and organization in the virulence of Mycobacterium marinum, particularly in facilitating phagosome damage and bacillary escape. Using the well-established model of Dictyostelium discoideum infected with M. marinum, which has frequently been predictive of Mycobacterium tuberculosis behavior within phagosomes, the authors contribute critical insights into the mechanisms of mycobacterial phagosome escape-a key step in cellular invasion and dissemination. These findings have the potential to inform strategies aimed at blocking this escape mechanism, which, as demonstrated in this study, could prevent intracellular bacterial growth.

This work is significant for advancing our understanding of mycobacterial pathogenesis, particularly by linking membrane microdomain composition to bacterial virulence. It will be highly relevant to researchers studying mycobacteria, intracellular pathogens, and host-pathogen interactions. While the study's use of M. marinum provides valuable insights, a limitation is that these results may not fully translate to M. tuberculosis, and further testing with the latter pathogen will be essential.

We sincerely thank the referee for their very strong appraisal of our contributions, past and present, much appreciated. We agree that the translation of our findings to Mtb and macrophages is not guaranteed ... but has turned to be surprisingly and satisfyingly consistent in the past. To our delight, a recent article in Nature Communications reports about "Paired analysis of host and pathogen genomes identifies determinants of human tuberculosis" and clearly identified flotillin-1 as a susceptibility factor for tuberculosis (PMID: 39613754). We have introduced a sentence in the discussion that reads "Importantly and consistently with our findings, recent work has revealed flotillins as a major determinant of the fate of Mtb infection in patients, because overexpression of flotillin-1, resulting from particular allele variants, is a host susceptibility factor for Mtb infection (PMID: 39613754)." (Lines 477-480)

I am an expert in the infection of macrophages by Mycobacterium tuberculosis, the phagosome escape mechanism, and its associated effectors. I also have expertise in microscopy and image analysis. However, I do not specialize in Dictyostelium discoideum biology.

---

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

the authors of this manuscript reported that EsxA, a secreted virulent factor of Mtb or Mm, causes membrane lysis in sterol-rich micro domain. They used the Mm-infected amoeba as an infection model, and characterized the effects of microdomain in Mycobacterium-containing Vacuole (MCV) on EsxA-mediated membrane disruption. They found that disruption of the micro domain through knockout of vacuolins or sterol depletion diminished Mm-induced membrane damage and cytosolic escape. They also found that vacuolins and sterol are essential for EsxA inserting into the membranes in vitro, and flotillin knockdown and sterol depletion conferred the resistance of murine microglial cells to Mm infection. The experiments were well designed and controlled, and the data were convincing.

We thank the referee for this snappy summary of our main findings and for the positive comment on study design.

My major comment is that the authors need to justify the use of BV-2 cells that are murine microglial cells, instead of macrophage cell lines, which are more relevant to Mtb/Mm infection.

We understand the referee's concerns about the host used for Mm infection. First, we would like to argue that it is very true that the detailed biological processes accompanying the infection by Mtb, Mm or in fact any other pathogen depend on the origin and status of the host cell. In the TB field, a plethora of host macrophages, from murine and human origins, primary or immortalised, alveolar or interstitial, M1 or M2 have been used through the decades. Beside a robust agreement on many processes (phagosome maturation arrest, MCV membrane damage, role of xenophagy etc...), some of the crucial outcomes, for example the susceptibility or resistance to Mtb infection and the type of host cell death, have been hotly debated and depend on the host phagocyte identity and status.

Now, it is true that microglial cells have only rarely been used for Mtb (or Mm) research, but it does not mean that this is not relevant. First, we would like to remind the referee that TB is not only a pulmonary disease, and that among the most disastrous extra-pulmonary sites of infection is the brain, resulting in the devastating tuberculous meningitis. In fact, tuberculous meningitis is the most severe form of tuberculosis with a fatality rate of 20-50% in treated individuals (doi: https://doi.org/10.1101/2025.03.04.641394). A quick literature survey on the topic reveals over 9,000 publications, including very significant contributions, using both Mtb and Mm in animal and human models (PMID: 38745656, PMID: 38264653, PMID: 36862557, PMID: 32057291, PMID: 30645042, PMID: 29352446, PMID: 27935825, PMID: 26041993).

We have introduced a brief mention of these arguments in the discussion (Lines 456-459).

In addition, we have already shown that this BV-2 cell line is reliable, they are adherent, motile and constitutively phagocytic and thus do not need to be differentiated with mega-doses of PMA, or any other stimulus. They beautifully recapitulate our findings in the Dd-Mm model (PMID: 38270456, PMID: 25772333), including when used as a host phagocyte to validate anti-infective compounds that were primarily identified using the Dd-Mm platform (PMID: 29500372).

We have introduced a brief mention of these arguments in the results section (Lines 329-334).

We also introduced two novel experimental evidence to strengthen the link between the Dd and BV-2 model systems. First, we show using qRT-PCR that, like vacuolins, flotillin-1 is upregulated in BV-2 at 32hpi (Fig. EV9B). Excitingly, as mentioned as response to referee #1, a recent article in Nature Communications reports about "Paired analysis of host and pathogen genomes identifies determinants of human tuberculosis" and clearly identified flotillin-1 as a susceptibility factor for tuberculosis (PMID: 39613754). We have introduced a sentence in the discussion that reads "Importantly and consistently with our findings, recent work has revealed flotillins as a major determinant of the fate of Mtb infection in patients, because overexpression of flotillin-1, resulting from particular allele variants, is a host susceptibility factor for Mtb infection (PMID: 39613754)." (Lines 477-480)

Second, we used for the first time the LysoFlipper probe to monitor MCV lipid composition and packing during infection (Fig. 7C). These results indicate that in BV-2 cells, as in Dd, the membrane characteristics of the MCV are profoundly different from the standard endo-lysosomal compartments.

Reviewer #2 (Significance (Required)):

It is well known that EsxA is membrane-lytic protein playing a role in Mtb/Mm-mediated phagosomal escape. There are other studies that have indicated lipid raft or micro domains in the membrane may play a role in EsxA-mediated membrane damage. This study further confirmed that the sterol-rich micro domain on the membrane has significant influence on the EsxA-mediated membrane disruption both in vitro and in vivo. While this finding is expected, but confirmation with solid experimental evidence is welcomed. This study also identified the genes or proteins required for micro domain organization, vacuolins and flotillin, which could be a target of host-directed therapy. Overall, this study is performed well and the results are convincing.

We thank the referee for their expert views and comments on the function of EsxA and the lipidic environment in which it is supposed to act. We agree that EsxA has been the centre of attention for decades, but we respectfully disagree that its precise mode of action is known, neither in vitro nor in vivo. First, historically, it took the best of a decade for the field to accept that Mtb was not a strictly vacuolar pathogen. And even when the escape to the cytosol became a fact, the implication of EsxA remained extremely debated. For example, a "petition" was signed and published, arguing against its direct membrane damaging activity (PMID: 28119503). We agree that cumulated evidence now converges against a canonical "pore-forming" activity, but in favour of a "membrane-disrupting" activity. On the other hand, it is true that researchers have reached a form of consensus on the role of low pH to dissociate the EsxA-B dimer, and on the importance of the "physiological" composition of the acceptor membrane (PMID: 31430698, PMID: 35271388, PMID: 17557817). We are convinced that our evidence is not merely expected and confirmatory, but represents a novel, complete, solid, biochemical in vitro, molecular and genetics in vivo demonstration of the role of sterols clustering and microdomain organisers as susceptibility factors for Mm infection in evolutionary distant phagocytes.

---

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The manuscript by Bosmani, Perret et al examines the role of Dictyodistelium discoideum vacuolin proteins in the integrity of the Mycobacterium marinum vacuole membrane. The data demonstrates that loss of vacuolins, similar to sterol depletion, reduced vacuole membrane damage meaning less cytosolic escape of the pathogen and subsequently reduced bacterial replication. The authors demonstrate functional analogy in a mammalian model of infection - where flotillin plays a similar role to the vacuolins - and this is an important demonstration of the utility of the D. discoideum model. The data is well presented and clear.

We thank the referee for this positive summary of our main findings and of the clarity of results, interpretations and working model.

Major Comments:

There is no evidence presented in the manuscript of "microdomains" - while I believe this is likely a true description of what is happening on the vacuole membrane there is no visualisation of this. Both the GFP-Vac vacuole staining and the filipin staining show complete coverage of the vacuole. Perhaps at the 1 hour time points this is more convincing but I think it is worth looking at more of these earlier time points and quantifying these "microdomains" - i.e. proportion of vacuole membrane that is positive for the Vacs. Is it possible to look at the GFP-Vac signal and filipin staining at the same time? And other vacuole markers too?

We agree with the referee that microdomains are the central characters of our study. Now, we would like to argue with the referee that one has to distinguish between structural, morphological evidence for the existence of microdomains and the biochemical and genetic evidence of their functional implication.

On the one hand, microdomains are in fact nanometer-scale and are thus under the resolution limit of most optical microscopies. We and others already documented that during phagosome maturation, vacuolin distribution is patchy, reflecting the clustering of nanometer-scale inhomogeneities, and that the coating becomes more continuous with progressing maturation. The transition we observed here for vacuolins, as microdomain organisers, from a patchy to continuous coating reflects indirectly their macroscopic coalescence. As discussed above in response to the first referee, visualisation of the underlying lipidic clusters and microdomains is for technical reasons almost undoable. One cannot fix sterols. As replied to the first referee, we have not been able to improve much on the spatial resolution of lipidic microdomains, and, despite numerous and rigorous tentatives, we have not been able to reliably obtain quantitative measurements of neither filipin nor D4H signals, nor to document changes in "patterns" of signals during MCV maturation. We nevertheless present additional qualitative D4H and VacC colocalization images Fig. EV1C.

On the other hand, we respectfully disagree that our manuscript lacks in strong and direct evidence for the functionality of sterol-rich microdomains as susceptibility factors required for a full mycobacteria infection in evolutionary distant phagocytes.

In addition to the evidence presented previously, we have now added a large set of RNAseq analyses of the impact of vac-KO and sterol depletion on infected and non-infected cells, which also highlight the interdependence between sterol concentration and vacuolin expression (Fig. 3G, 4G and H, Fig. EV5 and 6). Moreover, we have now used a Flipper probe sensitive to lipid composition and packing to distinguish the mature MCV from all other endo-lysosomal compartments in microglial cells (Fig. 7C). Altogether, the simplest and most plausible interpretation of our cumulated evidence is that sterol-rich microdomains are necessary for EsxA-mediated MCV damage and escape to the cytosol.

I really like the data presented in Figure 1 that demonstrates the specific upregulation of Vacuolin C during M. marinum infection. This is an intriguing result that brings up a lot of new questions e.g. how is this regulated? In response to membrane damage? Sensed by what? Does this upregulation also hold true for flotillin in the mammalian model? (and more!) however none of these ideas are pursued in the manuscript and by the end I was wondering why this data was included in the manuscript because all of the phenotypic data uses either a VacBC or ABC mutant. The link between figure 1 and the rest of the manuscript would be aided by characterisation of a specific VacC mutant.

We share the referee's fascination with these data showing that VacC is a specific reporter of virulent mycobacteria infection. First, VacC expression at the transcriptional level, but also at the protein accumulation level both point toward a correlation with an infection with damage-causing mycobacteria. Specifically, one can distinguish two stages, one transient upregulation of all three isoforms that becomes sustained only for VacC and only when wt Mm causes damage (as opposed to the DRD1 mutant or M. smegmatis). This is clearly presented in multiple places in the manuscript (for example lines 377-380).

Now, how is MCV damage sensed is extremely interesting and is the focus of numerous past and on-going studies in our laboratory but is out of the scope of this article. Just to mention a few lines of research as food for thoughts, membrane damage (by EsxA and by LLOMe) triggers the recruitment of the E3 ubiquitin ligase TrafE (PMID: 37070811), and subsequently of the ESCRT and autophagy machineries (PMID: 37070811, PMID: 30596802). Upstream of TrafE, we know that decrease of membrane tension is one parameter, because transient hyperosmolar shock also recruits TrafE to endo-lysosomal compartments (PMID: 37070811). On-going experiments demonstrate that calcium leakage from endo-lysosomes and MCV is another major triggering factor.

As mentioned above, and in more direct response to the referee's questioning, we have now included RNAseq experiments that unequivocally indicate the link between vac-KO and sterol depletion and the direct effect on reducing membrane damage, because the two conditions lead to a down-regulation of the damage-dependent transcriptomic signatures of the ESCRT and autophagy related genes (Fig. 4G-H and Fig. EV5). Moreover, it clearly establishes that sterol depletion, which decreases sterile and EsxA-mediated damage, decreases vacuolin expression in infected cells (Fig 3G). Finaly, qRT-PCR on infected BV-2 microglial cells indeed documents an up-regulation of flotillin-1, reminiscent of vacC regulation in Dd (Fig. EV9B).

All in all, we would like to respectfully ask the editor and referee to acknowledge that the signalling pathway between damage sensing and the vacuolin responses will be the focus of future studies.

We understand that investigating the phenotypic consequences of only a single vacC-KO might be interesting, but we would like to argue that it is superfluous. First, for intricate biological reasons, KO of single and combinations of vacuolin genes result in very qualitatively and quantitatively similar phenotypes associated to motility, phagocytosis, endosome maturation etc... (PMID: 32482795). The present study extends this remarkable phenomenon by interrogating multiple parameters, reporters and phenotypes linked to infection, some shown and some unpublished (for example Fig. EV3B and Fig. 4D-E).

Are the MMVs positive for all three vacuolins? It would be great if you could quantify which are present together or whether there are more distinct populations that are positive for just one or all three for example.

The referee points to an interesting mechanistic aspect. We have therefore directly assessed the colocalization of pairs of vacuolin isoforms (Fig. EV1B), which clearly indicate that every MCV is coated with two vacuolins, which therefore arithmetically implies that all three isoforms are present together and that there is no isoform-specific MCV (Fig 2B). This is potentially also corroborated by earlier studies that showed vacuolin hetero-oligomerization (PMID: 16750281), a characteristic shared by flotillins (PMID: 38985763).

Minor Comments:

Fig 1F - this graph is quite striking but I think the individual data points should be presented as it is unclear whether this intensity threshold is an arbitrary value or genuinely represents two different populations. Perhaps better represented as a scatter plot?

We fuly agree with the referee and have accordingly replotted all the graphs where this improved the visualisation and contributed to the interpretation of the data. We did not change the representation in Fig. 7E and G, Fig. EV3C, because the error bar already represents the deviation of the Area Under the Curve (AUC) that was calculated for the average curves resulting from a biological triplicate of experiments.

The bar graphs early in the manuscript should shoe the individual data points from replicates. While the presentation is clear and differences are striking I think this article explains why showing the replicate data is important: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002128

We fully agree with the referee and have accordingly replotted all the graphs where this improved the visualisation and contributed to the interpretation of the data.

In Figure 2: F and G should include quantification, in G the arrow on the24 hpi filipin panel is not in the right location

As mentioned in response to referee #1 and #2, qualitatively, sterols, as visualised by filipin and D4H, are present at all stages of the endo-lysosomal pathway and of MCV biogenesis. Now, there are many technical difficulties linked to a quantitative assessment, and therefore, please, let me present the framework. First, despite their wide use, the exact mechanism of binding of both reporters and which pool of sterol they visualise is still a mystery. This is often expressed as "they detect the accessible pool" of sterol, whatever it is. In addition, filipin detects sterols in both leaflets (and in intra-lumenal vesicles and other lipidic structures), while D4H detects sterols only in the cytosolic leaflet, and it is not known whether both leaflets have the same concentration of sterols. It is also known that filipin signal is only indirectly proportional to the sterol quantity in a cell, as measured by other quantitative methods. One of the best examples comes from studying the cellular phenotype of Niemann-Pick Type C disease, because many publications report a strong increase of filliping staining, whereas lipidomic analyses show at best a two-fold increase in cholesterol in NPC deficient cells. Moreover, technically speaking, D4H is a live probe, and fixation leads to some loss of localisation, probably because sterols are not fixable. On the other hand, filipin is mainly used after chemical fixation, but again sterols are not fixable, and the signal is very likely restricted to the membrane of origin, but not necessarily to the microdomains.

We corrected the arrow localisation.

Reviewer #3 (Significance (Required)):

The key strength of this manuscript is the use of the Dictyostelium model to dissect host-pathogen interactions. This provides an interesting evolutionary lens to the research findings presented here and is further strengthened by the data demonstrating that these findings are relevant in a mammalian model as well. The weaknesses are articulated in my "major comments" section. The phenotypic data presented here is strong - it is clear that these vacuolin proteins are important for the intracellular success of M. marinum however the data demonstrating the mechanism for this is less clear.

We thank the referee for this overall positive summary of our main findings and of the clarity of results, interpretations and working model. As detailed above, we respectfully disagree with the final conclusion and are pleased to note that the other two referees are more satisfied with the level of mechanistic evidence.

I am an academic researcher who is interested in the molecular host-pathogen interactions mediated by intracellular microbial pathogens. Scientists in my research field will be a key audience for this research. Predominantly this is basic researchers but the interest will be broader than host-pathogen interactions as researchers in the membrane integrity and membrane dynamics field will be interested here.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Bosmani, Perret et al examines the role of Dictyodistelium discoideum vacuolin proteins in the integrity of the Mycobacterium marinum vacuole membrane. The data demonstrates that loss of vacuolins, similar to sterol depletion, reduced vacuole membrane damage meaning less cytosolic escape of the pathogen and subsequently reduced bacterial replication. The authors demonstrate functional analogy in a mammalian model of infection - where flotillin plays a similar role to the vacuolins - and this is an important demonstration of the utility of the D. discoideum model. The data is well presented and clear.

Major Comments:

1. There is no evidence presented in the manuscript of "microdomains" - while I believe this is likely a true description of what is happening on the vacuole membrane there is no visualisation of this. Both the GFP-Vac vacuole staining and the filipin staining show complete coverage of the vacuole. Perhaps at the 1 hour time points this is more convincing but I think it is worth looking at more of these earlier time points and quantifying these "microdomains" - i.e. proportion of vacuole membrane that is positive for the Vacs. Is it possible to look at the GFP-Vac signal and filipin staining at the same time? And other vacuole markers too?
2. I really like the data presented in Figure 1 that demonstrates the specific upregulation of Vacuolin C during M. marinum infection. This is an intriguing result that brings up a lot of new questions e.g. how is this regulated? In response to membrane damage? Sensed by what? Does this upregulation also hold true for flotillin in the mammalian model? (and more!) however none of these ideas are pursued in the manuscript and by the end I was wondering why this data was included in the manuscript because all of the phenotypic data uses either a VacBC or ABC mutant. The link between figure 1 and the rest of the manuscript would be aided by characterisation of a specific VacC mutant.
3. Are the MMVs positive for all three vacuolins? It would be great if you could quantify which are present together or whether there are more distinct populations that are positive for just one or all three for example.

Minor Comments:

1. Fig 1F - this graph is quite striking but I think the individual data points should be presented as it is unclear whether this intensity threshold is an arbitrary value or genuinely represents two different populations. Perhaps better represented as a scatter plot?
2. The bar graphs early in the manuscript should shoe the individual data points from replicates. While the presentation is clear and differences are striking I think this article explains why showing the replicate data is important: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002128
3. In Figure 2: F and G should include quantification, in G the arrow on the24 hpi filipin panel is not in the right location

Significance

The key strength of this manuscript is the use of the Dictyostelium model to dissect host-pathogen interactions. This provides an interesting evolutionary lens to the research findings presented here and is further strengthened by the data demonstrating that these findings are relevant in a mammalian model as well. The weaknesses are articulated in my "major comments" section. The phenotypic data presented here is strong - it is clear that these vacuolin proteins are important for the intracellular success of M. marinum however the data demonstrating the mechanism for this is less clear.

I am an academic researcher who is interested in the molecular host-pathogen interactions mediated by intracellular microbial pathogens. Scientists in my research field will be a key audience for this research. Predominantly this is basic researchers but the interest will be broader than host-pathogen interactions as researchers in the membrane integrity and membrane dynamics field will be interested here.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

the authors of this manuscript reported that EsxA, a secreted virulent factor of Mtb or Mm, causes membrane lysis in sterol-rich micro domain. They used the Mm-infected amoeba as an infection model, and characterized the effects of microdomain in Mycobacterium-containing Vacuole (MCV) on EsxA-mediated membrane disruption. They found that disruption of the micro domain through knockout of vacuolins or sterol depletion diminished Mm-induced membrane damage and cytosolic escape. They also found that vacuolins and sterol are essential for EsxA inserting into the membranes in vitro, and flotillin knockdown and sterol depletion conferred the resistance of murine microglial cells to Mm infection. The experiments were well designed and controlled, and the data were convincing.

My major comment is that the authors need to justify the use of BV-2 cells that are murine microglial cells, instead of macrophage cell lines, which are more relevant to Mtb/Mm infection.

Significance

It is well known that EsxA is membrane-lytic protein playing a role in Mtb/Mm-mediated phagosomal escape. There are other studies that have indicated lipid raft or micro domains in the membrane may play a role in EsxA-mediated membrane damage. This study further confirmed that the sterol-rich micro domain on the membrane has significant influence on the EsxA-mediated membrane disruption both in vitro and in vivo. While this finding is expected, but confirmation with solid experimental evidence is welcomed. This study also identified the genes or proteins required for micro domain organization, vacuolins and flotillin, which could be a target of host-directed therapy. Overall, this study is performed well and the results are convincing.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The proposed study aims to elucidate the role of membrane microdomains and associated proteins-Vacuolin A, B, and C-during the infection of Dictyostelium discoideum (Dd) amoebae by Mycobacterium marinum (Mm). The results demonstrate that Vacuolins are required for Mm virulence, and that the presence of membrane microdomains is essential for phagosome membrane damage and bacillary escape into the cytosol-key steps in establishing a successful infection and subsequent bacterial proliferation. The study is well-designed, employing methodologies with which the authors have demonstrated expertise. Overall, it is methodologically sound, and most conclusions are well-supported by the presented data. However, some points require clarification. Major Points: The study aims to link the function of Dd Vacuolins to their potnetial facilitating role in phagosome escape and overall infection by Mm. To phenocopy the effect of Vac-KO, the authors used MβCD. Strikingly, this compound had a more significant impact on phagosome escape compared to Vac-KO, which either did not affect or only mildly affected this process. This likely reflects a difference in the underlying mechanisms being studied. Vac-KO cells may lack well-organized membrane domains but could retain a similar overall membrane composition. In contrast, MβCD disrupts these domains by chelating cholesterol, thus altering both the membrane composition and the domains themselves. This may explain why EsxA partitioning is more affected by MβCD than by triple KO. Consequently, this suggests that cholesterol, rather than the mere presence of membrane domains, plays a critical role in EsxA partitioning and activity in the phagosome. And even if LLOMe activity was lower in Vac-KO cells, this might be explained by the compartment targeted, the lysosomes which membrane composition may differ from the MCV. These points should be further discussed in the discussion section.

Based on this observation, in figure 2, does the D4H/filipin signal or association increase over time as the Vac signal "solidifies"? In Vac-KO cells, does the mScarlet-D4H signal change in intensity or pattern (building on fig. S1)? These insights could provide valuable information on cholesterol levels at the MCV in KO versus wild-type cells. If possible, the authors should quantify fluorescence or the frequency of signal association. Additionally, since Vacuolins do not have a significant impact on phagosome damage or escape, the difference in intracellular growth may be indirect, as suggested in the team's previous study on Vacuolins (DOI: 10.1242/jcs.242974). The authors measured MCV pH in figure S6-could they repeat this experiment to test whether Vacuolins affect MCV maturation? This was investigated in a previous version of the pre-print (DOI: 10.1101/2021.11.16.468763), and if the results still hold, it would strengthen the hypothesis that Vacuolins promote escape by modulating membrane organization, rather than influencing phagosome maturation. Finally, as a substantial part of this manuscript relies on microscopy and image analysis, the methods section should detail how these analyses were performed. Specifically, for figure 1f, it is unclear how the cells were segmented and fluorescence quantified-was total fluorescence per cell measured, or was an average value used? Figures 5c and 5h could be moved to the supplementary material, and the segmentation method should be explained in the methods section. Additionally, statistical analysis should be more clearly described, justifying the use of one-way or two-way ANOVA, and specifying the post-hoc tests used for group comparisons.

Significance

This study provides the first direct evidence of the importance of membrane composition and organization in the virulence of Mycobacterium marinum, particularly in facilitating phagosome damage and bacillary escape. Using the well-established model of Dictyostelium discoideum infected with M. marinum, which has frequently been predictive of Mycobacterium tuberculosis behavior within phagosomes, the authors contribute critical insights into the mechanisms of mycobacterial phagosome escape-a key step in cellular invasion and dissemination. These findings have the potential to inform strategies aimed at blocking this escape mechanism, which, as demonstrated in this study, could prevent intracellular bacterial growth.

This work is significant for advancing our understanding of mycobacterial pathogenesis, particularly by linking membrane microdomain composition to bacterial virulence. It will be highly relevant to researchers studying mycobacteria, intracellular pathogens, and host-pathogen interactions. While the study's use of M. marinum provides valuable insights, a limitation is that these results may not fully translate to M. tuberculosis, and further testing with the latter pathogen will be essential.

I am an expert in the infection of macrophages by Mycobacterium tuberculosis, the phagosome escape mechanism, and its associated effectors. I also have expertise in microscopy and image analysis. However, I do not specialize in Dictyostelium discoideum biology.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility, and clarity)

The manuscript by Song et al presents evidence to show that the predicted cysteine protease type 6 secretion system (T6SS) effector Cpe1 inhibits target cell growth by cleaving type II DNA Topoisomerases GyrB and ParE. The authors determined the structure of the protein complex formed by Cpe1 and its immunity protein Cpi1, which allowed them to reveal the mechanism of inhibition. Moreover, the authors identified type II DNA topoisomerases GyrB and ParE as the targets of Cpe1. Overall, the major conclusions were well supported by experimental data of high quality. The findings have expanded our appreciation of the mechanism utilized by T6SS effectors to inhibit target cell growth.

We thank the reviewer for their positive remarks and valuable suggestions to improve this manuscript.

---

Major comments

To better establish that GyrB and ParE are the sole targets of Cpe1, the authors should express the GG mutant in target cells and determine whether these cells become resistant to Cpe1-mediated killing (inhibition). They can also determine whether co-expression of the cleavage resistant mutants suppresses the toxicity of Cpe1.

We appreciate the reviewer’s suggestion to investigate additional substrates of Cpe1 beyond GyrB and ParE, which may not have been fully captured in our crosslinking-mass spectrometry experiments due to technical limitations or low protein abundance. To address this topic, we generated target cells heterologously expressing cleavage-resistant GyrB and ParE variants (GyrBΔG102 and ParEΔG98) that are not susceptible to Cpe1, as described in our original manuscript (Figures 3h, i). We performed both Cpe1 expression assay and competition assay to assess if expression of the cleavage-resistant variants suppresses Cpe1 toxicity (Author Response Figures 1a, b). However, we did not observe a substantial protective effect. While this outcome could suggest that GyrB and ParE are not the sole targets of Cpe1, alternative explanations are also plausible. In the Cpe1 expression assay, high levels of Cpe1 could still act on endogenous wild-type GyrB and ParE, and although we attempted to increase variant expression, precise quantification remains challenging. In the competition assay, highly active Cpe1 may have continued to target wild-type substrates throughout the experiment, potentially masking any protective effect. Additionally, reduced activity of the mutant proteins could contribute to the observed results. Finally, deletion of the global repressor H-NS in the Cpe1-producing E. coli strain may have induced other interbacterial competition mechanisms1, leading to growth inhibition independently of Cpe1. Addressing these questions comprehensively would require a more systematic investigation under a wider range of conditions. We consider this an important avenue for future studies.

Results in Figure 7 clearly show that Cpi1 is capable of displacing ParE from Cpe1 due to higher affinity. Yet, the "competitive inhibition model" described in the last result section does not completely match what is really happening in Cpe1-mediated interbacterial competition. If Cpi1 is in the target cell, it would more likely engage the incoming Cpe1 before it can interact with ParE or GyrB, so competition does not occur in this scenario. Similarly, in the predatory cells expressing Cpe1 and Cpi1, these two proteins will form a stably protein complex, and no competition with the target will occur. The authors should reconsider their model.

We thank the reviewer for their comments and appreciate the opportunity to clarify this point. First, we believe the reviewer is referring to Figure 5 rather than Figure 7. In our model, the primary role of immunity proteins in interbacterial competition is to neutralize cognate toxins and prevent self- or kin-intoxication. These immunity proteins exhibit high specificity and strong binding affinity toward their associated toxins, ensuring effective protection2. In predatory cells, immunity proteins are typically co-expressed with their corresponding toxins, likely enabling immediate suppression upon translation. During kin competition, immunity proteins can protect cells even after foreign toxins engage their substrates.

Our results demonstrate that Cpi1 binds Cpe1 with higher affinity than its substrates and can displace them from pre-formed Cpe1-substrate complexes (Figures 5b-f). This aligns with the established function of immunity proteins in interbacterial competition and provides a mechanistic basis for how they confer protection, even when toxins have initially engaged their targets2. We acknowledge the reviewer’s point that in both scenarios—whether in the recipient cell or the toxin-producing cell—Cpe1 may first encounter Cpi1. However, our model underscores that Cpi1 not only binds at the substrate site but also exhibits superior affinity for Cpe1, ensuring robust protection against Cpe1-mediated toxicity.

Minor comments

"Intoxication" was used throughout the text numerous times to describe the activity of Cpe1. Looking in the Marriam-Webster dictionary, "Intoxication" means "a condition of being drunk". This word should be replaced with "toxicity" or some other terms in this line.

We thank the reviewer for this comment. We acknowledge that the term "intoxication" is commonly associated with alcohol consumption, yet the Merriam-Webster dictionary also defines it as "an abnormal state that is essentially a poisoning" (https://www.merriam-webster.com/dictionary/intoxication). This definition aligns with its well-established usage in the field of interbacterial competition to describe the effects of interbacterial toxins during antagonism3-5, which we have adopted in our manuscript. However, we appreciate the reviewer’s concern and remain open to revising the terminology if deemed necessary for clarity.

Lines 46-48, references on contact-dependent killings by these systems mentioned should cited. Ref. 9 cited does NOT cover the information at all.

We thank the reviewer for this comment. We have revised the citation and now reference studies that specifically describe contact-dependent killing systems in the relevant sentences (Lines 45–____50)

"characterizations" should be "characterization".

We have now modified the sentence as requested (Line 69)

Line 229 "Cpe1-Bpa monomers" should be " apo Cpe1-Bpa". The results cannot distinguish whether these bands are monomers or multimers.

We appreciate the reviewer’s careful assessment of our manuscript. The results in Line 233 (Figure 3c) show the enrichment of His-tagged proteins, including crosslinked complexes and overproduced Cpe1-Bpa. Based on the molecular weight marker, the Cpe1-Bpa bands appear between 10–15 kDa, consistent with the molecular weight of Cpe1 monomers (Figure 3a). Therefore, we have labeled this band as “Cpe1-Bpa monomers” and maintained this terminology throughout the text. This designation aligns with previous studies utilizing site-specific crosslinking via Bpa incorporation6,7

Line 283, was the mutation deletion? Substitution was used I think.

We thank the reviewer for highlighting this point. The GyrB and ParE mutants used to confirm the cleavage sites were deletion mutants, with a single glycine removed from the predicted double-glycine motifs. We have now revised the text for clarity (Lines 285–290)

Lines 439-444 the discussion should be extended to include other bacterial toxins that target type II DNA topoisomerases (e.g. PMID: 26299961 and PMID: 26814232).

We appreciate the reviewer’s suggestion. The studies referenced (PMID: 26299961 and PMID: 26814232) describe FicT toxin with adenylyl transferase activity that target and post-translationally modify GyrB and ParE at their ATPase domains, highlighting a potential hotspot for topoisomerase inhibition. We have now incorporated an additional paragraph in the Discussion section to describe these findings (Lines 424–439).

Reviewer #1 (Significance)

The authors determined the structure of the protein complex formed by Cpe1 and its immunity protein Cpi1, which allowed them to reveal the mechanism of inhibition. Moreover, the authors identified type II DNA topoisomerases GyrB and ParE as the targets of Cpe1. Overall, the major conclusions were well supported by experimental data of high quality. The findings have expanded our appreciation of the mechanism utilized by T6SS effectors to inhibit target cell growth.

We sincerely thank the reviewer for their positive comments and for the suggestions to improve our manuscript.

Reviewer #2 (Evidence, reproducibility, and clarity)

The manuscript, titled "An Interbacterial Cysteine Protease Toxin Inhibits Cell Growth by Targeting Type II DNA Topoisomerases GyrB and ParE", describes how an effector family was identified and characterized as a papain-like cysteine protease (PLCP) that negatively impacts bacterial growth in the absence of its co-encoded immunity protein. This thorough report includes (1) bioinformatic analysis of prevalence, finding this PLCP effector encoded in many gram-negative bacteria, (2) confirming conservation of catalytic active site via structural (crystallographic) analysis, as well as visualizing contacts with the immunity protein, (3) validation of results using growth studies combined with mutagenesis, (4) using a cell-based cross-linking method to pull out potential targets, which were subsequently identified via mass spectrometry, (5) validation of these results using in vitro protease assays with purified (potential) substrates, including verification of the motif recognized on the substrate(s), and cell-based phenotype analyses, and finally, (6) demonstrating competition between immunity protein and ParE substrate using an in vitro pull-down approach. Overall, this is a strong body of work with compelling conclusions that are well supported by multiple experimental approaches.

We appreciate the reviewer for their positive comments regarding our original submission.

Major comments

The claims made based on the presented results are well supported, including that this PLCP effector toxin is widespread, is neutralized in a competitive mechanism by its immunity partner, and that it effectively cleaves both GyrB and ParE (subunits of bacterial type II topoisomerases) at a conserved motif, resulting in suppression of bacterial cell growth via mis-regulating chromosome segregation. No additional experiments are needed to further validate these results, and the authors are commended on the cell-based and in vitro studies to deduce very specific mechanisms and structural details.

We appreciate the reviewer’s positive feedback.

Minor comments

While the writing and data presentation are extremely clear, in general I recommend the authors indicate the level(s) of replication for experiments. Figure legends generally note that mean values with standard deviations are shown, but I did not find where the number of replicates (and independent versus technical) were listed.

We appreciate the reviewer’s suggestion. We have now revised the manuscript to specify the levels of replication (independent vs. technical) for each experiment in the figure legends, particularly in Figures 2 and 3.

The figures are very clear, but in many instances the addition of PLCP toxin is indicated as "before" and "after"; while a modest change, I recommend altering this to some type of "-" and "+" type nomenclature rather than a time-based notation (especially as presumably both samples were treated identically, just with or without protease).

We thank the reviewer for this helpful comment. In Figures 3 and Supplementary Figures 5, 9, we used "before" and "after" to indicate the time points for in vitro cleavage assays verifying Cpe1 cleavage. To minimize variations between reactions, the catalytic mutant Cpe1tox (Cpe1toxC362A) was used as a comparison rather than a reaction without Cpe1tox. In these assays, duplicate reaction mixtures were prepared: one was denatured immediately after preparation ("before" reaction) to serve as a baseline, while the other was incubated to allow enzymatic activity ("after" reaction). This labeling clarifies the comparison between initial and processed samples. We believe this approach clearly distinguishes the effects of Cpe1 activity and provides a reliable basis for assessing proteolysis in our assays.

I also suggest quantifying the intensities of the gel images presented in Figure 5c, d (for example, Cpe1 intensity as a ratio to that of the ParE ATPase domain), to make the interpretation even more evident.

We thank the reviewer for the valuable suggestion to quantify the signal intensities of the gel images presented in Figures 5c, d. We have now included the quantification results in Supplementary Figures 9e, f and have updated the respective text in the manuscript (Lines 826-828 and 1066-1087).

Crystallographic structure: the PDB report notes some higher-than-expected RZR (RSRZ) scores; I interpret this to mean that there was strain around the catalytic site of one of the two toxins in the asymmetric unit, or that this copy was less well ordered. The RZR outliers likely arise from non-optimal weighting for geometric restraints. While no figures of electron density are presented, these modest outliers are not expected to alter the conclusions reached in the current work. One point of interest that is not addressed, however, is if any variance between the two complexes in the asymmetric unit are noted? A passage compares the current toxins to others in the larger subfamily and notes a rotation of a side chain is needed to superpose (Line 159). Can the authors please clarify around which bond this rotation is needed, and if both copies in the asymmetric unit are in the same orientation at this site?

We appreciate the reviewer’s insightful comments.

1. We have provided the electron density map for the RSR-Z outlier residues along with the model (Author response Figure 2a). These outlier residues are located at the loop regions of a molecule within the asymmetric unit in the crystal (Chain B). As a result, the electron density for their side chains appears to be noisier compared to residues in the well-folded regions, leading to higher RSR-Z scores. Notably, when we superimposed the models of two complexes within the asymmetric unit, the calculated RMSD value was 0.402 Å (Author response Figure 2b), indicating that the two models are structurally very similar and that these residues are properly assigned. Therefore, the RSR-Z outliers do not significantly impact the overall structure.
2. Here, we provide a zoomed-in view of Figure 2d, highlighting the superimposed crystal structures of Cpe1 and the closely related PLCPs, ComA and LahT (Author response Figure 2c). As shown, the side chain of the catalytic cysteine residue in ComA adopts a different orientation, positioning it slightly farther from the homologous residues in Cpe1 and LahT. However, since the backbone and catalytic pockets remain structurally intact, we believe that this deviation arises due to results from crystal packing effects rather than an inherent functional distinction. We have now modified the main text (Lines 159-166) to clarify this and prevent any potential misinterpretation.

Reviewer #2 (Significance)

Bacteria encode numerous effectors to successfully compete in natural environments or to mediate virulence; these effectors are typically associated with type VI secretion system machinery or referred to as contact dependent inhibition systems. The current work has identified a sub-family of papain-like cysteine protease effectors that are unique by targeting type II topoisomerases. Among the actionable findings is the identification of both the specific site of interaction with the topo substrates, as well as the specific motif recognized for cleavage. This should enable the field to move forward probing for this activity with other toxins and substrates. The insights provided by the competitive neutralization mechanism also stand out as an important contribution that can be more broadly applied. Within the literature, few effector targets are identified, making the current study stand out as impactful by the well-executed experiments that directly support the conclusions.

While the current study has strong elements of novelty and is complete, it also nicely sets up future studies for remaining open questions. For example, does the nucleotide-bound status of the ATPase domain, or other catalytic intermediate, impact the susceptibility of topoisomerases to cleavage? Is this identified motif found in other ATPase domains? Is the negative supercoiling activity unique to gyrase also impacted, or is the phenotypic mechanism of cell toxicity reliant only on chromosome segregation? What types of kinetic parameters do this class of toxins demonstrate, and does sequence variability alter this? These ideas are a testament to the intriguing study as presented, capturing the readers' curiosity for additional details that are clearly beyond the scope of the current work.

I anticipate this work will be of interest to the broad field of microbiologists that study interbacterial communication as well as pathogenic mechanisms. While the research is largely fundamental in nature, it is wide in scope with applications to many gram-negative bacteria that inhabit a myriad of niches. The work will also be of interest to specialists in topoisomerases, as the list of toxins that target these essential enzymes is growing and the therapeutic utility of topoisomerase inhibition remains vital. My interest lies in the latter, in toxin-mediated inhibition of topoisomerase enzymes as a means to alter bacterial cell growth. While I have strong expertise in structural biology, I am lacking in expertise for mass spectrometry. I note this because this method was used for the identification of the target substrate.

We appreciate the reviewer’s insightful discussion and interest in our study. We agree that further investigations are crucial to address the open questions posed, and we have initiated work on some of these avenues.

For example, considering Cpe1's specificity for the ATPase domain of GyrB and ParE, we have begun examining whether Cpe1 targets other ATPase domains by searching for the consensus sequence or double glycine motifs in the sequences of ATPase domains beyond GyrB and ParE. Among the 42 E. coli ATPase domains identified by the PEC database8, we found several with double glycine residues. However, none contained the exact LHAGGKF consensus sequence identified in GyrB and ParE, which are targeted by Cpe1 (Author Response Figure 3). These findings suggest that Cpe1 is less likely to target other ATPase domains. Nonetheless, due to Cpe1’s potential tolerance of certain variations within the consensus sequence, we cannot draw a definitive conclusion without further investigation into the cleavage sites.

Another critical open question is the impact of Cpe1-mediated cleavage on the function of GyrB and ParE. To address this topic, we have begun investigating if Cpe1 cleavage affects the ATPase activity of these proteins. As expected, our biochemical analysis has demonstrated a significant decrease in ATP hydrolysis in the presence of active Cpe1tox, but not in the presence of the catalytic mutant Cpe1toxC362A (Author response Figures 4a, b). These results confirm that the ATP-dependent activities of both GyrB and ParE are disrupted following Cpe1 cleavage9. Previous work on FicT toxin that inhibits GyrB and ParE ATPase activity through post-translational modification found that ATP-dependent activities such as DNA supercoiling, relaxation, and decatenation were inhibited10,11. Interestingly, GyrB’s relaxation of negative supercoiled DNA, which does not require ATP, was also affected to some extent. This outcome raises the question as to whether Cpe1-cleaved GyrB results in similar downstream defects. Investigating this possibility would provide valuable insights into Cpe1’s mode of action, although we feel doing so is beyond the scope of the current study. Consequently, we view this as an important area for future research.

Finally, regarding the potential applications of Cpe1, we are interested in further investigating its enzymatic specificity and properties. In this study, we analyzed the binding kinetics between Cpe1 and its substrate (Figure 5f) and currently we are endeavoring to characterize the kinetics of Cpe1-mediated proteolysis. To better probe hydrolytic dynamics, we plan to utilize a substrate with a reporting group (such as a chromogenic or fluorogenic leaving group) to monitor cleavage over time. We could achieve this by designing a recombinant substrate based on our knowledge of Cpe1’s native substrates (GyrB and ParE) and the target sequence (“LHAGGKF”). Alternatively, a secondary reaction leading to colorimetric changes could be employed for detection. We consider this an exciting research direction and an important next step for this study.

Overall, we are grateful for the reviewer’s recognition of the novelty and importance of our work in advancing the understanding of interbacterial toxins and their inhibitory effects on topoisomerases. We plan to further investigate the consequences of Cpe1 cleavage on GyrB and ParE and to explore Cpe1 kinetics and its mechanistic actions in more detail. This will not only deepen our understanding of bacterial toxin-mediated inhibition but may also provide critical insights into strategies for targeting type II DNA topoisomerases. The reviewer’s insightful feedback has proven invaluable in shaping our ongoing and future research directions.

Reviewer #3 (Evidence, reproducibility, and clarity)

Bacterial warfare in microbial communities has become illuminated by recent discoveries on molecular weapons that allow contact-dependent injection of bacterial toxins between competitors. Among the best characterized systems are the type VI secretion system (T6SS) or the contact-dependent inhibition (CDI) system (i.e. some of the T5SSs). These systems are delivering a plethora of toxins with various biochemical activities and a broad range of targets. In recent years many such toxins have been characterized and their relevance in pointing at appropriate drug targets is increasing.

In this study the authors built on a previously published association of a family of proteins, papain-like cysteine proteases (PLCPs), with their delivery by T6SS or CDI into target bacterial cells. Whereas this observation is not particularly novel, the findings that this set of proteins, that the authors called now Cpe1, can specifically target bacterial proteins such as ParE and GyrB, so that it affects chromosome partitioning and cell division, is groundbreaking. The authors are clearly demonstrating that Cpe1 cleaves their target proteins at double glycine recognition site which is in line with previous characterization of such proteases when fused to a particular category of ABC transporters. Even more remarkably they can show using biochemical approaches that Cpi1 is a cognate immunity for CpeI, preventing its activity, not by interfering with the catalytic site, but instead with the substrate binding site. The mechanism of competitive inhibition between immunity and substrate is also substantiated by biochemical data.

We sincerely appreciate the reviewer’s interest in and support of our study.

Major comments

• This is a very well conducted study which combines bacterial genetics and phenotypes with excellent biochemical evidence.

We thank the reviewer for their positive comments.

• There are 8 targets identified for Cpe1 and yet only two are cleaved by the enzyme. It is intriguing that FtsZ is one identified target by the pull down but not confirmed for cleavage. The authors rules this as false positive but the cell division defect associated with Cpe1 activity would be consistent here. Are there any double glycine in FtsZ that could be identified as cleavage site? Is it possible that slightly different incubation conditions may promote degradation of FtsZ?

We appreciate the reviewer’s thoughtful comment regarding FtsZ as a potential substrate of Cpe1. This was indeed an intriguing possibility, especially given the cell division defects observed following Cpe1 intoxication. Early on in the project, we also identified FtsZ as a Cpe1 interactor in our proteomic crosslinking assays, which further fueled the hypothesis that FtsZ might be a target.

To explore this possibility, first we examined the FtsZ protein sequence for potential Cpe1 cleavage sites and identified several double glycine motifs (Author response Figure 5a). However, none of these motifs matched the consensus sequence identified in GyrB and ParE, which is LHAGGKF, a sequence that we have shown to be critical for Cpe1 cleavage activity. In an effort to better understand if FtsZ could still be cleaved by Cpe1, we conducted additional cleavage assays under various conditions (Author response Figure 5b). We tested different incubation temperatures, including increasing the temperature to 37 °C, and extended the reaction time to overnight. However, we did not observe any cleavage of FtsZ under these conditions. Given that FtsZ undergoes significant conformational changes upon binding to GTP12, we also considered the possibility that the GTP-bound form of FtsZ might be cleaved by Cpe1. However, even under those conditions, no significant cleavage of FtsZ was detected (Author response Figure 5b). Based on these results, we do not have any evidence to support that FtsZ is a target of Cpe1. The observed cell division defects are more likely a secondary effect resulting from the cleavage of GyrB and ParE, direct targets of Cpe1 that are crucial for chromosome segregation.

• Could it be structurally predicted whether the GG of ParE or GyrB is fitted into the catalytic site of Cpe1.

We appreciate the reviewer’s insightful question regarding the structural prediction of the GG motif of ParE and GyrB fitting into the catalytic site of Cpe1. To address this possibility, we used Alphafold 3 to predict the interaction structure between Cpe1 and its substrates13. The resulting model of Cpe1 interacting with the ATPase domain of GyrB (GyrBATPase) is shown in Supplementary Figure 9c. As illustrated, the loop of the GyrB ATPase domain containing the consensus targeting sequence (“LHAGGKF”) fits into the catalytic site of Cpe1, with the GG motif positioned closest to the catalytic cysteine residue, which likely facilitates hydrolysis. We also attempted to model the interaction between Cpe1 and the ATPase domain of ParE. However, confidence for this model was lower (ipTM = 0.74, pTM = 0.71), possibly due to Alphafold’s preference for certain protein configurations. To gain a more accurate understanding of how Cpe1 binds and recognizes its substrates, we are currently working on co-crystallizing Cpe1tox with GyrB and ParE. This long-term project aims to provide precise structural insights into the Cpe1-substrate interaction and further elucidate the mechanism of cleavage.

Minor comments

• The authors described a family of proteases, PLPCs, and characterized one here called Cpe1. Not clear whether this is a generic name or one specific protein from one particular bacterial species. Indeed, it is unclear from which bacterial strain the Cpe1 protein studied here originates.

We thank the reviewer for this comment and apologize for the lack of clarity. To provide better context, we have now revised the manuscript (Lines 136-137 and 141-145) to clearly state that the Cpe1 protein characterized in this study originates from E. coli strain ATCC 11775.

• It may be worth to emphasize that the Cpe1 domain is found in all possible configurations as T6SS cargo and that is to be linked to VgrG, PAAR or Rhs.

Thank you for this suggestion. We have revised the manuscript accordingly to emphasize this point (Lines 106-109).

• Line 49 the authors could indicate that the Esx system is also known as type VII secretion system (T7SS).

Thank you for this suggestion. We have revised the manuscript accordingly (Line 48-50).

• Line 113 it may be better to use Proteobacteria instead of Pseudomonadota

We have revised the manuscript (Lines 114-115) as suggested by the reviewer. It is important to note that following the recent decision by the International Committee on Systematics of Prokaryotes (ICSP) to amend the International Code of Nomenclature of Prokaryotes (ICNP) and formally recognize "phylum" under official nomenclature rules14,15, the taxonomy database used in our analysis has adopted the updated nomenclature. To ensure consistency, we followed this updated nomenclature throughout the original manuscript.

Reviewer #3 (Significance)

This is an excellent piece of work. The characterization of Cpe1 might look poorly novel at the start when compared to previous studies. Yet the findings go crescendo by characterizing original mechanisms of action of the cognate immunity, and by identifying the molecular target of Cpe1. This is providing real conceptual advance in the T6SS field and not just reporting yet another T6SS toxin.

As a T6SS expert I genuinely feel that these findings are groundbreaking and could be targeted to broad audience since the possible implications of these observations for future antimicrobial drugs discovery or therapeutic approaches is highly relevant.

We sincerely appreciate the reviewer’s positive remarks and support of our study.

References

1. Ishihama, A., and Shimada, T. (2021). Hierarchy of transcription factor network in Escherichia coli K-12: H-NS-mediated silencing and Anti-silencing by global regulators. FEMS Microbiol Rev 45. 10.1093/femsre/fuab032.
2. Hersch, S.J., Manera, K., and Dong, T.G. (2020). Defending against the Type Six Secretion System: beyond Immunity Genes. Cell Rep 33, 108259. 10.1016/j.celrep.2020.108259.
3. Russell, A.B., Singh, P., Brittnacher, M., Bui, N.K., Hood, R.D., Carl, M.A., Agnello, D.M., Schwarz, S., Goodlett, D.R., Vollmer, W., and Mougous, J.D. (2012). A widespread bacterial type VI secretion effector superfamily identified using a heuristic approach. Cell Host Microbe 11, 538-549. 10.1016/j.chom.2012.04.007.
4. Jana, B., Fridman, C.M., Bosis, E., and Salomon, D. (2019). A modular effector with a DNase domain and a marker for T6SS substrates. Nat Commun 10, 3595. 10.1038/s41467-019-11546-6.
5. Halvorsen, T.M., Schroeder, K.A., Jones, A.M., Hammarlof, D., Low, D.A., Koskiniemi, S., and Hayes, C.S. (2024). Contact-dependent growth inhibition (CDI) systems deploy a large family of polymorphic ionophoric toxins for inter-bacterial competition. PLoS Genet 20, e1011494. 10.1371/journal.pgen.1011494.
6. Nguyen, T.T., Sabat, G., and Sussman, M.R. (2018). In vivo cross-linking supports a head-to-tail mechanism for regulation of the plant plasma membrane P-type H(+)-ATPase. J Biol Chem 293, 17095-17106. 10.1074/jbc.RA118.003528.
7. Liu, Y., Yu, J., Wang, M., Zeng, Q., Fu, X., and Chang, Z. (2021). A high-throughput genetically directed protein crosslinking analysis reveals the physiological relevance of the ATP synthase 'inserted' state. FEBS J 288, 2989-3009. 10.1111/febs.15616.
8. Yamazaki, Y., Niki, H., and Kato, J. (2008). Profiling of Escherichia coli Chromosome database. Methods Mol Biol 416, 385-389. 10.1007/978-1-59745-321-9_26.
9. Reece, R.J., and Maxwell, A. (1991). DNA gyrase: structure and function. Crit Rev Biochem Mol Biol 26, 335-375. 10.3109/10409239109114072.
10. Harms, A., Stanger, F.V., Scheu, P.D., de Jong, I.G., Goepfert, A., Glatter, T., Gerdes, K., Schirmer, T., and Dehio, C. (2015). Adenylylation of Gyrase and Topo IV by FicT Toxins Disrupts Bacterial DNA Topology. Cell Rep 12, 1497-1507. 10.1016/j.celrep.2015.07.056.
11. Lu, C., Nakayasu, E.S., Zhang, L.Q., and Luo, Z.Q. (2016). Identification of Fic-1 as an enzyme that inhibits bacterial DNA replication by AMPylating GyrB, promoting filament formation. Sci Signal 9, ra11. 10.1126/scisignal.aad0446.
12. Matsui, T., Han, X., Yu, J., Yao, M., and Tanaka, I. (2014). Structural change in FtsZ Induced by intermolecular interactions between bound GTP and the T7 loop. J Biol Chem 289, 3501-3509. 10.1074/jbc.M113.514901.
13. Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A.J., Bambrick, J., et al. (2024). Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493-500. 10.1038/s41586-024-07487-w.
14. Oren, A., Arahal, D.R., Rossello-Mora, R., Sutcliffe, I.C., and Moore, E.R.B. (2021). Emendation of Rules 5b, 8, 15 and 22 of the International Code of Nomenclature of Prokaryotes to include the rank of phylum. Int J Syst Evol Microbiol 71. 10.1099/ijsem.0.004851.
15. Oren, A., and Garrity, G.M. (2021). Valid publication of the names of forty-two phyla of prokaryotes. Int J Syst Evol Microbiol 71. 10.1099/ijsem.0.005056.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary

Bacterial warfare in microbial communities has become illuminated by recent discoveries on molecular weapons that allow contact-dependent injection of bacterial toxins between competitors. Among the best characterized systems are the type VI secretion system (T6SS) or the contact-dependent inhibition (CDI) system (i.e. some of the T5SSs). These systems are delivering a plethora of toxins with various biochemical activities and a broad range of targets. In recent years many such toxins have been characterized and their relevance in pointing at appropriate drug targets is increasing. In this study the authors built on a previously published association of a family of proteins, papain-like cysteine proteases (PLCPs), with their delivery by T6SS or CDI into target bacterial cells. Whereas this observation is not particularly novel, the findings that this set of proteins, that the authors called now Cpe1, can specifically target bacterial proteins such as ParE and GyrB, so that it affects chromosome partitioning and cell division, is groundbreaking. The authors are clearly demonstrating that Cpe1 cleaves their target proteins at double glycine recognition site which is in line with previous characterization of such proteases when fused to a particular category of ABC transporters. Even more remarkably they can show using biochemical approaches that Cpi1 is a cognate immunity for CpeI, preventing its activity, not by interfering with the catalytic site, but instead with the substrate binding site. The mechanism of competitive inhibition between immunity and substrate is also substantiated by biochemical data.

Major comments

• This is a very well conducted study which combines bacterial genetics and phenotypes with excellent biochemical evidence.
• There are 8 targets identified for Cpe1 and yet only two are cleaved by the enzyme. It is intriguing that FtsZ is one identified target by the pull down but not confirmed for cleavage. The authors rules this as false positive but the cell division defect associated with Cpe1 activity would be consistent here. Are there any double glycine in FtsZ that could be identified as cleavage site? Is it possible that slightly different incubation conditions may promote degradation of FtsZ?
• Could it be structurally predicted whether the GG of ParE or GyrB is fitted into the catalytic site of Cpe1.

Minor comments

• The authors described a family of proteases, PLPCs, and characterized one here called Cpe1. Not clear whether this is a generic name or one specific protein from one particular bacterial species. Indeed, it is unclear from which bacterial strain the Cpe1 protein studied here originates.
• It may be worth to emphasize that the Cpe1 domain is found in all possible configurations as T6SS cargo and that is to be linked to VgrG, PAAR or Rhs.
• Line 49 the authors could indicate that the Esx system is also known as type VII secretion system (T7SS).
• Line 113 it may be better to use Proteobacteria instead of Pseudomonadota

Significance

This is an excellent piece of work. The characterization of Cpe1 might look poorly novel at the start when compared to previous studies. Yet the findings go crescendo by characterizing original mechanisms of action of the cognate immunity, and by identifying the molecular target of Cpe1. This is providing real conceptual advance in the T6SS field and not just reporting yet another T6SS toxin. As a T6SS expert I genuinely feel that these findings are groundbreaking and could be targeted to broad audience since the possible implications of these observations for future antimicrobial drugs discovery or therapeutic approaches is highly relevant.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The manuscript, titled "An Interbacterial Cysteine Protease Toxin Inhibits Cell Growth by Targeting Type II DNA Topoisomerases GyrB and ParE", describes how an effector family was identified and characterized as a papain-like cysteine protease (PLCP) that negatively impacts bacterial growth in the absence of its co-encoded immunity protein. This thorough report includes (1) bioinformatic analysis of prevalence, finding this PLCP effector encoded in many gram-negative bacteria, (2) confirming conservation of catalytic active site via structural (crystallographic) analysis, as well as visualizing contacts with the immunity protein, (3) validation of results using growth studies combined with mutagenesis, (4) using a cell-based cross-linking method to pull out potential targets, which were subsequently identified via mass spectrometry, (5) validation of these results using in vitro protease assays with purified (potential) substrates, including verification of the motif recognized on the substrate(s), and cell-based phenotype analyses, and finally, (6) demonstrating competition between immunity protein and ParE substrate using an in vitro pull-down approach. Overall, this is a strong body of work with compelling conclusions that are well supported by multiple experimental approaches.

Major comments:

The claims made based on the presented results are well supported, including that this PLCP effector toxin is widespread, is neutralized in a competitive mechanism by its immunity partner, and that it effectively cleaves both GyrB and ParE (subunits of bacterial type II topoisomerases) at a conserved motif, resulting in suppression of bacterial cell growth via mis-regulating chromosome segregation. No additional experiments are needed to further validate these results, and the authors are commended on the cell-based and in vitro studies to deduce very specific mechanisms and structural details.

Minor comments:

While the writing and data presentation are extremely clear, in general I recommend the authors indicate the level(s) of replication for experiments. Figure legends generally note that mean values with standard deviations are shown, but I did not find where the number of replicates (and independent versus technical) were listed.

The figures are very clear, but in many instances the addition of PLCP toxin is indicated as "before" and "after"; while a modest change, I recommend altering this to some type of "-" and "+" type nomenclature rather than a time-based notation (especially as presumably both samples were treated identically, just with or without protease). I also suggest quantifying the intensities of the gel images presented in Figure 5c, d (for example, Cpe1 intensity as a ratio to that of the ParE ATPase domain), to make the interpretation even more evident.

Crystallographic structure: the PDB report notes some higher-than-expected RZR scores; I interpret this to mean that there was strain around the catalytic site of one of the two toxins in the asymmetric unit, or that this copy was less well ordered. The RZR outliers likely arise from non-optimal weighting for geometric restraints. While no figures of electron density are presented, these modest outliers are not expected to alter the conclusions reached in the current work. One point of interest that is not addressed, however, is if any variance between the two complexes in the asymmetric unit are noted? A passage compares the current toxins to others in the larger subfamily and notes a rotation of a side chain is needed to superpose (Line 159). Can the authors please clarify around which bond this rotation is needed, and if both copies in the asymmetric unit are in the same orientation at this site?

Significance

Bacteria encode numerous effectors to successfully compete in natural environments or to mediate virulence; these effectors are typically associated with type VI secretion system machinery or referred to as contact dependent inhibition systems. The current work has identified a sub-family of papain-like cysteine protease effectors that are unique by targeting type II topoisomerases. Among the actionable findings is the identification of both the specific site of interaction with the topo substrates, as well as the specific motif recognized for cleavage. This should enable the field to move forward probing for this activity with other toxins and substrates. The insights provided by the competitive neutralization mechanism also stand out as an important contribution that can be more broadly applied. Within the literature, few effector targets are identified, making the current study stand out as impactful by the well-executed experiments that directly support the conclusions.

While the current study has strong elements of novelty and is complete, it also nicely sets up future studies for remaining open questions. For example, does the nucleotide-bound status of the ATPase domain, or other catalytic intermediate, impact the susceptibility of topoisomerases to cleavage? Is this identified motif found in other ATPase domains? Is the negative supercoiling activity unique to gyrase also impacted, or is the phenotypic mechanism of cell toxicity reliant only on chromosome segregation? What types of kinetic parameters do this class of toxins demonstrate, and does sequence variability alter this? These ideas are a testament to the intriguing study as presented, capturing the readers' curiosity for additional details that are clearly beyond the scope of the current work.

I anticipate this work will be of interest to the broad field of microbiologists that study interbacterial communication as well as pathogenic mechanisms. While the research is largely fundamental in nature, it is wide in scope with applications to many gram-negative bacteria that inhabit a myriad of niches. The work will also be of interest to specialists in topoisomerases, as the list of toxins that target these essential enzymes is growing and the therapeutic utility of topoisomerase inhibition remains vital. My interest lies in the latter, in toxin-mediated inhibition of topoisomerase enzymes as a means to alter bacterial cell growth. While I have strong expertise in structural biology, I am lacking in expertise for mass spectrometry. I note this because this method was used for the identification of the target substrate.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Song et al presents evidence to show that the predicted cysteine protease type 6 secretion system (T6SS) effector Cpe1 inhibits target cell growth by cleaving type II DNA Topoisomerases GyrB and ParE. The authors determined the structure of the protein complex formed by Cpe1 and its immunity protein Cpi1, which allowed them to reveal the mechanism of inhibition. Moreover, the authors identified type II DNA topoisomerases GyrB and ParE as the targets of Cpe1. Overall, the major conclusions were well supported by experimental data of high quality. The findings have expanded our appreciation of the mechanism utilized by T6SS effectors to inhibit target cell growth.

Specific comments:

Main points:

1. To better establish that GyrB and ParE are the sole targets of Cpe1, the authors should express the GG mutant in target cells and determine whether these cells become resistant to Cpe1-mediated killing (inhibition). They can also determine whether co-expression of the cleavage resistant mutants suppresses the toxicity of Cpe1.
2. Results in Figure 7 clearly show that Cpi1 is capable of displacing ParE from Cpe1 due to higher affinity. Yet, the "competitive inhibition model" described in the last result section does not completely match what is really happening in Cpe1-mediated interbacterial competition. If Cpi1 is in the target cell, it would more likely engage the incoming Cpe1 before it can interact with ParE or GyrB, so competition does not occur in this scenario. Similarly, in the predatory cells expressing Cpe1 and Cpi1, these two proteins will form a stably protein complex, and no competition with the target will occur. The authors should reconsider their model.

Minor points:

1. "Intoxication" was used throughout the text numerous times to describe the activity of Cpe1. Looking in the Marriam-Webster dictionary, "Intoxication" means "a condition of being drunk". This word should be replaced with "toxicity" or some other terms in this line.
2. Lines 46-48, references on contact-dependent killings by these systems mentioned should cited. Ref. 9 cited does NOT cover the informatin at all.
3. "characterizations" should be "characterization".
4. Line 229 "Cpe1-Bpa monomers" should be " apo Cpe1-Bpa". The results cannot distinguish whether these bands are monomers or multimers.
5. Line 283, was the mutation deletion? Substitution was used I think.
6. Lines 439-444 the discussion should be extended to include other bacterial toxins that target type II DNA topoisomerases (e.g. PMID: 26299961 and PMID: 26814232).

Significance

The authors determined the structure of the protein complex formed by Cpe1 and its immunity protein Cpi1, which allowed them to reveal the mechanism of inhibition. Moreover, the authors identified type II DNA topoisomerases GyrB and ParE as the targets of Cpe1. Overall, the major conclusions were well supported by experimental data of high quality. The findings have expanded our appreciation of the mechanism utilized by T6SS effectors to inhibit target cell growth.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reply to the Reviewers

We thank the reviewers for their evaluation of our previous submission and have responded to each point in detail below. Overall, we have revised the manuscript with the addition of several new data and corresponding figure panels that strengthen our previous conclusions and add new insights allowing us to extend the conclusions of the study. Important additions include new data showing the impact of loss of CLU on adapting to additional stressors during metabolic transitions that supports a mechanistic understanding of our omics results; by poly(dT) FISH we show that fly Clu granules indeed contain mRNAs; FRAP microscopy analysis supports that Clu1 granules have dynamic content similar to other LLPS membraneless organelles; and we have re-analysed our data to demonstrate more clearly the impact of Clu1 on translation efficiency and also the relative binding of mRNAs during translation. In addition, we provide some extra control analyses for completeness.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary:

In this manuscript the authors study the Clustered mitochondrial proteins Clu of Drosophila melanogaster and Clu1 of Saccharomyces cerevisiae, two homologues of the mammalian protein CLUH. They show in compelling microscopy analysis that both proteins form granules. This was the case for flies fed on yeast paste after starvation and in yeast in post-diauxic phase, in respiratory media or during mitochondrial stress. They show that these granules are found in proximity to mitochondria and that they behave like liquid-liquid-phase separated condensates. They show by co-staining for P-bodies and stress granules that Clu1-granules are distinct from these RNA granules. Furthermore, they found that the formation required active translation. In the second part, they show that Clu1 interacts with ribosomal and mitochondrial proteins by BioID. The deletion of Clu1 leads to slightly impaired growth on media containing Ethanol as a carbon source. They find that nascent polypeptides of some mitochondrial precursor proteins are decreased in the deletion of Clu1 and conclude that Clu1 regulates translation of these proteins. Using RNA immunoprecipitation of Clu1-GFP in presence of cycloheximid, EDTA and puromycin. The mRNAs of nuclear-encoded mitochondrial proteins found to be interacting with Clu1 were purified in conditions when the ribosomes are intact and the RNAs showed no interaction when ribosomes were disassembled. They show in sucrose gradients that Clu1 co-migrates with polysomes independent of its distribution state or carbon source. However, when cells are grown in conditions of granule formation, then polysomes and Clu1 run less deeply into the gradient. Form these data, the authors conclude that Clu/Clu1 regulates the translation of nuclear-encoded mitochondrial proteins.

Major comments:

-The authors state that Clu1 is regulating translation during metabolic shifts. However, it is not clear what the real impact on mitochondrial function is. They show that there is a minor growth defect on ethanol media when CLU1 is deleted. However, if Clu1 is necessary mainly for adaptation, the phenotype will be strongest observed in conditions where cells switch carbon sources. Growth curves would be suitable in which the lag-phase of yeast cells precultured either in glucose or glycerol switched to media of different carbon sources (glucose to glycerol or glycerol to glucose) are measured. One would expect that the deletion mutant shows a longer lag-phase compared to the wild type when shifted from glucose to glycerol media.

We agree that this is an important question, and, duly, we previously attempted to address this exactly as the reviewer described. Surprisingly, we were not able to observe any substantial differences in the duration of the lag phase between the wild-type and CLU1 knockout strains under these conditions. However, we did note that CLU1 knockout cells consistently reached stationary phase with a lower optical density when switched to ethanol media, consistent with these cells having a different metabolic efficiency during growth on ethanol media.

To further explore the role of Clu1, we noted that several of the Clu1 mRNA interactors were mitochondrial heat shock proteins (HSPs), which are crucial for mitochondrial protein folding and import during the transition from fermentation to respiration. Hence, we hypothesised that the absence of Clu1 might lead to increased sensitivity to heat shock during the metabolic shift.

To test this, we subjected both wild-type and CLU1 knockout cells to heat shock under three different conditions: (1) during growth on glucose-containing media (fermentation), (2) after shifting cells to media containing ethanol during the lag phase, when cells are adapting to respiration, and (3) after cells had fully adapted to ethanol and resumed growth. Interestingly, CLU1 knockout cells were more sensitive to heat shock selectively during the adaptation to respiration, which involves the translation of an extensive number of mitochondrial proteins. We think that the small difference in translation of mitochondrial HSPs becomes evident only upon additional heat shock, likely due to a deficient mitochondrial protein folding and import. These findings support our hypothesis that Clu1 is essential for optimal mitochondrial function during metabolic shifts.

These results have been added to the manuscript and shown in Fig. S6 and described on page 9.

-In line with this, how different is the mitochondrial proteome of the WT and the mutant? Do hits of the BioID, RIP and Punch-P experiments change at steady state or during metabolic shifts? Either proteomics of isolated mitochondria or western blots of whole cells or isolated mitochondria of WT and the deletion mutant grown in conditions of Clu1-granule formation or no granules for the hits could answer this question.

We also considered this question during the course of the work. However, in exploratory analyses we saw no obvious differences in overall mitochondrial proteomics at steady-state which is what prompted us to look at more subtle effects on translation. Considering this further, changes in steady-state levels can be complex to interpret as they represent the combined effects of protein production and degradation. Small changes arising from altered production could be masked by compensatory changes in turnover rate. In light of this, we believe that the translational regulation differences identified in our study remain central to understanding the role of Clu1, and any downstream proteomic changes would not alter our primary conclusions.

-The authors analyze RNAs bound in polysomes to assess translation efficiency. Translation efficiency is usually calculated by the fraction of RNA bound by ribosomes to the total RNA amount of an RNA species. Thus, doing RT-qPCR from whole cells would be necessary to assess if the occupancy of ribosomes on the transcripts is due to changes in RNA abundance or other regulatory pathways and would help to further assess what causes the observed changes.

Thanks for this recommendation. To address this and expand our analysis to other proteins differentially translated in clu1Δ cells, we measured the mRNA steady-state levels by performing RNAseq on WT and clu1Δ strains grown under the same conditions as used for Punch-P. We then calculated the translation efficiency by dividing the nascent protein levels (Punch-P) by steady-state mRNA levels (RNAseq), as previously described for Punch-P data (PMID: 26824027). The translation efficiency for the majority of proteins with reduced translation in the clu1Δ cells by Punch-P analysis was lower. Similarly, the majority of proteins with increased translation had higher translation efficiency.

The mRNA quantification in polysomes we originally presented in the manuscript, further showed that the decrease in translation efficiency is not caused by a simple decrease of mRNA engaged in translation and that Clu1 is regulating protein translation at the ribosome level. In contrast, for higher translated proteins, we detected an increase in mRNAs engaged in polysomes, likely underlying the increased translation. These results further support our conclusions regarding the regulatory effects of Clu1 on translation.

These results have been added to the manuscript and shown in Fig. 7E and described on page 9.

OPTIONAL:

-The authors show a co-localization of Clu/Clu1 with mitochondrial fission factors and conclude that the granules appear likely near fission sites. Indeed, CLUH has been implied in the past to play a role in mitochondrial fission (Yang, H., Sibilla, C., Liu, R. et al. Clueless/CLUH regulates mitochondrial fission by promoting recruitment of Drp1 to mitochondria. Nat Commun 13, 1582 (2022). https://doi.org/10.1038/s41467-022-29071-4). Thus, are fission sites required for Clu-granule localizations? What is the role of the mitochondrial network integrity for the granule distribution? Expressing Clu-GFP/Clu1-GFP in cells depleted for the fission factors would provide information on that.

Thanks for this suggestion. We agree that it would be interesting to know whether Clu1 granules still appear when mitochondrial fission is blocked. We tried to address this question but encountered some technical limitations. First, overexpression of Clu1-GFP via a plasmid did not replicate the endogenous Clu1 behaviour, making it necessary to delete the fission factors in the Clu1-GFP background. While crossing the Clu1-GFP strain with already available knockout strains would be straightforward, we would need access to a tetrad dissecting microscope, which unfortunately was not available to us. We also attempted PCR-based gene deletion but the sequence homology between the GFP-tagging cassette and the deletion cassettes made this very challenging. Given these limitations, and as the lab's yeast expert had already left, we were not able to pursue this experiment further and have removed these observations from our manuscript. We hope that future studies will explore this question in more detail.

-The author assess convincingly that Clu1 interacts with ribosomes and runs with polysomal fractions. However, how it actually regulates translation is not clear. To answer this question, selective ribosomal profiling would be necessary. The authors have established conditions which would be suitable for the experiment. They could use crosslinking and sucrose cushions to IP ribosomes with Clu1-GFP bound to be used for ribosomal profiling. However, this experiment is quite time-intensive (3-4 months) and expensive, thus, an optional suggestion.

We thank the reviewer for this suggestion. We agree that ribosome profiling could provide novel insights into the function of Clu1/Clu. While we recognise the potential of this approach, as the reviewer points out, this experiment would indeed be time- and resource-intensive. Based on our initial tests, where we included cross-linked samples (UV and formaldehyde) we anticipate that it could even take longer than the estimated 3-4 months, as the IP using cross-linked lysates was not as successful as the IP using non-cross-linked samples: we were not able to immunoprepitate Clu1 so efficiently likely to the epitope being poorly exposed to the antibody. Although we have optimised working conditions for co-immunoprecipitating Clu1 with ribosomes, performing ribosome profiling using our setup within the timeframe and resources of this study is unfortunately not currently feasible.

Minor comments:

Fig1: B, C, please add scale bars into the zoom ins.

These have been added.

Fig 2 would profit from inlets of zoom ins to visualize the distribution better.

These have been added.

Fig.3: Panel C does not really add much information. I would rather remove it or put it into supplements and therefore show a zoom of Panel E with a line plot showing the rings. It is not clear from the represented images where the rings are formed.

We think some confusion has arisen from the text description. It seems that the reviewer was under the impression that Fig. 3C and 3E were intended to be showing the Clu1 rings around the mitochondria, but this was shown only in Fig. S3A. We have re-written these sentences for better clarity. To be clear, Fig. 3C is a 3D rendering of the left-hand cell in 3B (3D is a line plot of part of the right-hand cell) and 3E is a different experiment showing the formation of Clu1 granules under a different respiratory stress (galactose plus CCCP). We have also added a line plot showing Clu1-GFP and mito-mCherry fluorescence intensity to highlight the Clu1 rings around the mitochondria in Fig. S3A.

Fig.3 panel F: Max projections are not appropriate to show colocalization as they can lead to false-positive overlaps. Just remove the max projections.

We tried a number of different approaches to improve this analysis but, ultimately, we were not able to generate sufficiently robust data to be convincing so we decided to remove this from the manuscript. The coincidence of Clu1 granules with mitochondrial fission factors was an adjunct observation and not a major part of the story and has been discussed by others relating to fly Clu (PMID: 35332133), so removal from the current manuscript does not impact the key conclusions of the study.

References 21 and 22 are the same.

Thanks. This has been fixed.

Reviewer #1 (Significance (Required)):

This manuscript shows in a convincing way that Clu and Clu1 form RNA granules and that Clu1 interacts with ribosomes. It is written in a clear way and the figures support the conclusions drawn in the text. The finding that Clu/Clu1 is important for metabolic adaptation has not been shown in fly or yeast to my knowledge. It is in line with findings for the mammalian homologue CLUH. Thus, the findings are supported by earlier work. This study is of value for a broader audience of the basic research field, especially of the mitochondrial and RNA granule field, as it supports the idea of post-transcriptional regulation of nuclear-encoded mitochondrial protein gene expression for dynamic adaptation of mitochondrial function. The conditions when Clu granules form is studied in detail, followed up by identification of target RNAs and interaction partners. Though the interaction of Clu1 with ribosomes is shown in a compelling way, a detailed mechanism of the function of Clu/Clu1 is missing and would require more experiments. Thus, even though a detailed mechanism is missing, the study does expand on our understanding of Clu/Clu1 in regulating mitochondrial biogenesis and is therefore of high interest of the mitochondrial field.

Expertise: mitochondria, yeast, RNA granules, mitochondrial biogenesis, next-generation sequencing, fluorescence microscopy

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary:

In this manuscript the authors use D. melanogaster and S. cerevisiae to study the role of CLUH in the translation of nuclear-encoded mitochondrial proteins. During conditions requiring aerobic respiration, CLUH forms RNA-dependent granules that localise in the proximity to mitochondria. Furthermore, the authors demonstrate that CLUH interacts with translating ribosomes to facilitate the translation of specific target mRNAs. For this, the authors use a combination of GFP-tagged CLUH models. BioID, polysome translating proteomics, RNA-IP. The authors' main conclusions are that (i) CLUH forms dynamic, membrane-less, RNA-dependent granules under conditions that demand aerobic respiration, (ii) CLUH interacts with specific mRNAs encoding metabolic factors, and (iii) CLUH interacts with the translating ribosome. The manuscript is well written and the conclusions stand in proportion to the experimental output and the results. The main concern is with regards to lack of advancement in relationship to published data.

We appreciate the reviewer's feedback and specific comments which we respond to individually below. However, we would like to first address the point regarding "lack of advancement" and the use of the "CLUH" terminology which the reviewer uses throughout their critique. We would like to reiterate, as the reviewer states, our work focussed exclusively on yeast Clu1 and Drosophila Clu. None of our data relates to mammalian CLUH. While these proteins share substantial sequence homology, it is imprudent and scientifically unsound to assume cross-species equivalence without directly testing. Indeed, one of the central aims of our study was to characterise the molecular function of yeast Clu1, which remains almost entirely unstudied.

We acknowledge that some of the observations contained within our study have been described by others and we have appropriately noted and cited these in context. Nevertheless, (a) independent replication is always valuable but easily criticised as lacking novelty, and (b) the majority of the work was analysing the molecular dynamics and function of yeast Clu1 which is almost completely unstudied and may help provide hypotheses for others to test for conservation in mammalian CLUH. Hence, we consider that summarising the work as 'lacking advancement' is misplaced.

Comments:

To this reviewer it is not clear how CLUH can regulate the translation of specific mRNAs while being bound to ribosomes, regardless of being in a diffuse or granular state. The authors suggest that under metabolically active conditions, CLUH might aggregate translating ribosomes, forming the granular structures. How CLUH though can both be bound to translating ribosomes and recruit specific mRNAs at the same time is not explained.

It was indeed surprising to us that the data indicate that Clu1 can bind both mRNAs and ribosomes to affect translation, and we share the reviewer's curiosity about the precise mechanism of how this occurs. While we have provided novel insights into this situation, dissecting the precise molecular mechanisms is beyond the scope of the current study.

The authors might want to discuss how changes in metabolic demands signal the aggregation of CLUH, and how CLUH can recognise its target mRNAs.

We appreciate the reviewer's point here but as this would be pure speculation we have made only brief comments on this at the end of the Discussion.

What was the rationale to perform the RIP or the PUNCH-P experiments only under non-challenged conditions, but not under conditions demanding aerobic respiration?

We appreciate the reviewer's question. In fact, the Punch-P analysis was carried out on cells that had been transferred to ethanol to induce respiration. This was stated in the Methods, but we appreciate that this may have been missed so we have now clarified this in the main text (p9).

Regarding the RIP, our initial tests showed that mRNAs encoding proteins found to interact with Clu1 by BioID were interacting with Clu1 in both fermenting and respiring conditions. Due to this consistency, it did not seem necessary to perform the RIP experiments under both metabolic conditions, so we chose to conduct the experiment under the simpler growth condition.

If CLUH is ubiquitously bound to ribosomes, has CLUH been seen in any structural representation of the cytosolic ribosome?

This is a good question, and we wondered the same. To our knowledge, Clu1/Clu/CLUH has not been observed in any structural studies of the ribosome, and no formal structure of any Clu family proteins has been resolved.

Nevertheless, we would like to clarify that we do not think, or suggest in the manuscript, that Clu/Clu1 is ubiquitously bound to ribosomes. First, current evidence supports that Clu/Clu1 only regulates a specific subset of mRNAs. Second, our work, particularly the sucrose gradient experiments, shows that Clu1 interacts transiently with ribosomes, as cross-linking was required to capture the full extent of this interaction. This transient and selective interaction of Clu/Clu1 with the ribosome, together with the fact that transient interactors are often lost during ribosome purification, makes Clu/Clu1 detection in structural studies unlikely. Due to the transient interaction and dynamic localisation of Clu/Clu1, capturing Clu/Clu1 in ribosomal structures will require significant work in the future.

Reviewer #2 (Significance (Required)):

CLUH has been studied in various publications, showing data very similar to that presented in this manuscirpt. However, the authors provide a comprehensive analysis on both yeast and fly CLUH. The strength of the manuscript is the combination of several elegant methods and genetically modified model systems in two species to elucidate the role of CLUH during the translation of specific mRNA. In my view through, the advancement of understanding the function of CLUH is limited.

Although the authors work in yeast and DM, the results seem applicable to other species, including humans, and thus, the presented results will be of interest in a range of researchers working in the field of metabolic regulation and gene expression.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary: This study from Miller-Fleming et al. employs yeast and Drosophila as model systems to explore the function of the RNA-binding protein Clu1, which is involved in mitochondrial biogenesis. The first part of the manuscript characterizes so called "Clu1 granules", and their dependance from metabolic transitions. In particular, using yeast, they find a relocalisation of Clu1 upon starvation and several mitochondrial stress conditions. These granules are not stress granules, and are dissolved by RNAse and puromycin treatment. The second part of the study aims to understand the molecular function of the protein and its link to translation. The results confirm an evolutionary conserved role of Clu1 in binding mRNAs for mitochondrial proteins and in interacting with mitochondrial proteins, ribosomal components and polysomes. In addition, the authors claim that binding of Clu1 to RNA is enhanced when mRNAs are trapped in polysomes by treatment with cycloheximide (CHX), leading to the proposal that Clu1 binds mRNAs during active translation.

Major comments:

-The claim of Clu1 granule localization next to mitochondria (Figure 3) would be more convincing if any of the experiment would be quantified. Especially in the case of panel 3G in Drosophila egg chambers where there are a lot of mitochondria, one wonders whether the closeness to mitochondria is just random. Furthermore, mdv1-signal does not look very convincing, being blurry and not dotty as expected. Thus, the conclusion that Clu1 granules partially colocalization with site of fission appears premature.

The claim that Clu/Clu1 granules are often found in close proximity to mitochondria was inferred from observations from multiple analyses from yeast (we looked at hundreds of cells in several different conditions) and flies, where it had already been demonstrated (Cox and Spradling, 2009). We agree that observations of the fly egg chambers are challenging due to the very high density of mitochondria (and other cellular components - see the new analysis of poly(A) mRNAs) in these highly active cells. These considerations motivated us to take the CLEM approach (in addition to investigating the membraneless nature), to gain a much higher resolution view of the localisation of the granules. This analysis unequivocally showed that the Clu granules were exactly juxtaposed to several mitochondria. It is noteworthy that even in the TEM images shown, there is ample cytoplasm in which the Clu granule could be located if the association with mitochondria was coincidental and all granules had mitochondria in close proximity.

Regarding the possible coincidence of Clu1 with mitochondrial fission factors, as mentioned above for Reviewer 1, we tried a number of different approaches to improve this analysis but, ultimately, we were not able to generate sufficiently robust data to be convincing so have decided to remove this from the manuscript. Since this was an adjunct observation and not a major part of the story and has been discussed by others relating to fly Clu (PMID: 35332133), removal from the current manuscript does not impact the key conclusions of the study.

Based on the ability of 1,6-hexanediol to dissolve the granules (Figure 4), the authors conclude that: "Clu1 foci have membraneless nature". As they correctly state in the discussion, treatment with 1,6-hexanediol can have other effects. I suggest to be more cautious with the conclusions or add additional experiments. Are the granules dynamics if using FRAP? Do they fuse?

The inference that the Clu1 granules are membraneless organelles was not solely based on the observation that they disassemble upon 1,6-hexanediol treatment but was made in conjunction with the CLEM analysis that showed unambiguously that Clu granules are not associated with any detectable membrane, which is strong evidence that these granules are membraneless in nature. Indeed, as the reviewer mentioned, we are cautious in concluding they have been formed by liquid-liquid phase separation (LLPS) and we do acknowledge that 1,6-hexanediol can have other effects in cells. Nevertheless, following the reviewer's suggestion we have analysed Clu1 granule dynamics using FRAP, even though we are aware that FRAP is also not a definitive proof that a structure is formed by LLPS. The FRAP analysis, shown in new Figure 4C, D, revealed approximately 50% recovery over 10 min imaging timeframe. As discussed on page 13, this indicates a dynamic nature of these granules, but this dynamism can vary widely between different types of granules and even different proteins within the same granule. Further work is warranted to fully investigate the dynamic nature of Clu/Clu1 granule components.

The experiment in which the granules are dissolved by treatment with RNAse is very interesting. However, per se this does not directly demonstrate that the granules contain mRNA. To state this the author should perform FISH experiments for example using a probe to detect poly-A.

We thank the reviewer for this suggestion. We have performed poly(dT) FISH in egg chambers. Initial analysis showed that the fluorescence was diffuse and widely distributed, as expected for these highly active cells, but with no specific accumulation in Clu granules. Interestingly, we observed that treatment with RNase A, which we initially used to demonstrate probe specificity, revealed an enrichment of poly(A) RNAs in Clu granules. So, while treating the live egg chambers with RNase revealed that granules depend on RNA for their stability, treating fixed egg chambers revealed more directly the presence of RNAs in granules.

These results have been added to the manuscript and shown in Fig. 5 and described on page 7.

The authors show that puromycin prevents the granule formation before insulin addition in the fly. Are these results (upon RNAse treatment and puromycin treatment) recapitulated in the yeast system? The authors conclude that Clu1 formation depends on mRNAs being engaged in translation, but never show that the granules are site of active translation. More experiments in this direction (for example using puro-PLA of specific mRNAs) are missing and would clearly improve the manuscript.

Thanks for this very interesting consideration. We agree that we have not formally shown that the Clu1 granules are sites of active translation. A major limitation to addressing this is that puromycin is not able to penetrate the yeast cell wall, so cannot be used for analysis of intact cells as would be needed in this case. We agree that this would be a welcome addition but is beyond the scope of the current study.

The interactome of Clu1-neighbouring proteins (Figure 6) is interesting and a valuable addition to data in other organisms. I am wondering why the authors have not used as a control a cytosolic BirA-GFP, which would have been the right control for this experiment, especially since GFP tends to form aggregates.

We thank the reviewer for this comment. With hindsight, we agree that a cytosolic BirA-GFP would have been a better control. However, we are confident in our results for the following reasons:

1. The levels of GFP obtained from Clu1-GFP expression are low, and under these conditions, we observed no evidence of GFP aggregation. Even in experiments where GFP is overexpressed from a high-copy 2µ plasmid under a strong promoter, we do not detect aggregation. Aggregation is not a concern in our experimental setup.
2. Our conclusions are not solely based on the interactome analysis (BioID) but are supported by complementary findings. Specifically, several proteins identified in the proximity to Clu1 in the BioID analysis showed reduced translation in Clu1 knockout cells, and their corresponding mRNAs were found to interact with Clu1 during translation. These complementary results from independent techniques provide strong evidence for Clu1's role and validate the findings of the interactome analysis. Given this robust and complementary dataset, having BirA as a control strain was sufficient to validate our conclusions.

Figure 7B: The log 2 FC for the changed proteins are in many cases small, implying that the difference in translation for these proteins is not so large. For this reason, it is relevant to know how was the statistical significance calculated for these MS measurements. In the supplementary Tables and in Fig 7B, a p value is indicated and it is not clear if this is a simple p value or an adjusted p value (FDR or q value). If not shown, I recommend showing the adjusted p value, so that one can have an idea of the solidity of the data and the claim. Again, this is an important piece of evidence, since the authors base on this experiment the conclusion that Clu1 controls translation of these mRNAs.

Thanks for this comment. We have now included the q-value in the supplementary table.

Minor comments:

-Figure 1: The change in Clu1 localisation in post-diauxic phase or upon changing of the medium is evident from the images shown. However, it seems that the experiment has been performed only once (the same for Figure 2). Is this the case? An important information would be to show the expression levels of Clu1-GFP in the different conditions. Does recruitment of CLU1 to granules associate to increased expression levels?

The experiments shown in figures 1 and 2 were performed independently at least three times, as stated in the figure legends. The numbers shown are indicative values from one of the replicate experiments. This has now been added to the figure legends.

We agree that providing the information regarding the expression levels of Clu1-GFP is important to address whether the recruitment of Clu1 to granules is associated with changes in its abundance. To this end, we have performed an additional experiment to quantify Clu1-GFP levels under the conditions where Clu1 is diffuse (log growth phase in glucose-containing media) and when Clu1 is in granules (sodium azide treatment).

These results have been added to the manuscript and shown in Fig. S2 and described on page 4.

Figure 2 A-B. The authors claim that the only stressor capable of inducing Clu1 granules formation alone is inhibition of complex IV activity via sodium azide treatment. Other mitochondrial stresses like CCCP treatment or OA treatment are efficient only when combined to starvation. It should be mentioned that sodium azide treatment is not only capable of inhibiting complex IV but has also uncoupling function.

Thanks for this comment. We have now mentioned this (p4).

Figure 2 D-E: investigation of colocalization with Bre5 would help to understand how similar the yeast Clu1 granules are compared to the mammalian CLUH granules (Pla-Martin et al., 2020).

This is an interesting suggestion and one that we also considered, but with limited time and resources we were not able to pursue this line of inquiry as well.

Figure 8. This figure summarizes one of the most novel pieces of data about Clu1, the interaction with mRNAs via the ribosome. The way how panel A-C are represented is however a bit misleading. The Y axis in Figure B and C has the same amplitude as the one in A. Therefore, potential differences in Clu1-RNA pull-down in presence of EDTA or puromycin cannot be assessed. It is true that in presence of CHX there is much more pulled down RNA, but one cannot judge from these panels if there is any difference between Clu1 targets and controls also in the other conditions. The graphs should be modified and statistics added.

We appreciate the reviewer's feedback regarding the presentation of the RIP-qPCR data in Fig. 8. Based on the comments, we have revised how the results are represented, improved the normalisation of the data and added statistical analysis.

First, it is worth clarifying that the presentation of the original charts was done specifically to highlight the huge differences between RNA-pulldown in CHX versus disrupted ribosomes. It is also important to note that these RIP experiments were performed simultaneously under identical experimental conditions, so any differences lie in the treatments applied. To improve cross-comparison between treatments we have now incorporated an additional normalisation step. We normalised the enrichment levels of each mRNA tested against the non-specific binding observed with the negative control housekeeping genes (UBC6 and TAF10). This ensures that differences in bead loss or other technical variations are accounted for.

We now show the comparison of the six positive hits and two negative controls normalised as described above, on the same scale (Fig. 8A). We now also present the relative effects of the three conditions (CHX, EDTA, and puromycin) within the same graph for each mRNA tested (Fig. 8B). This format enables direct comparison of Clu1 target mRNA enrichment and two negative controls across treatments, which is the relevant comparison for testing the hypothesis of ribosome-dependent interactions. We have adjusted the Y-axis scaling for each mRNA, as requested by the reviewer, and added statistical comparisons. For clarity, the data shown in Fig. 8A are also represented in the panels of Fig. 8B (CHX). We have amended the text appropriately and hope that these changes improve the comparisons between treatments and more readily demonstrate that Clu1 target enrichment is lost upon ribosome disassembly, either by EDTA or by puromycin.

In addition, RNAse treatment in panel L does not seem to have really worked.

These samples were cross-linked prior to treatment to preserve the transient interaction of Clu1 with the ribosome, hence, the normal dramatic effect of RNase to collapse the polysomes is much less pronounced. Nevertheless, the purpose of this experiment was to monitor whether Clu1 co-migrated with ribosomes, which it does.

The authors should cite Vornlocher et al. (PMID: 10358023), who were the first to implicate Clu1 (Tif31) with translation.

Thank you for this prompt. We have now added a comment on this in the Discussion (page 13).

References 21 and 22 are the same.

Thanks. This has been fixed.

Reviewer #3 (Significance (Required)):

The data reported in this manuscript are valuable, because they confirm an evolutionary conserved role of Clu1 in binding mRNAs for mitochondrial proteins and regulating their translation. It is also interesting that in yeast, similar to Drosophila and mammalian cells, Clu1 can form granular structures upon metabolic rewiring. A limitation of the study is that direct experiments to support the claim that Clu1 concentrates ribosomes engaged in translation are not provided. Furthermore, it is not clear what is the functional role of the Clu1 granules, since the proximity interactome and the binding of Clu1 to the polysomes is not affected by treatments that dissolve or stimulate granule formation.

The study is of interest to a general cell biology audience.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This study from Miller-Fleming et al. employs yeast and Drosophila as model systems to explore the function of the RNA-binding protein Clu1, which is involved in mitochondrial biogenesis. The first part of the manuscript characterizes so called "Clu1 granules", and their dependance from metabolic transitions. In particular, using yeast, they find a relocalisation of Clu1 upon starvation and several mitochondrial stress conditions. These granules are not stress granules, and are dissolved by RNAse and puromycin treatment. The second part of the study aims to understand the molecular function of the protein and its link to translation. The results confirm an evolutionary conserved role of Clu1 in binding mRNAs for mitochondrial proteins and in interacting with mitochondrial proteins, ribosomal components and polysomes. In addition, the authors claim that binding of Clu1 to RNA is enhanced when mRNAs are trapped in polysomes by treatment with cycloheximide (CHX), leading to the proposal that Clu1 binds mRNAs during active translation.

Major comments:

• The claim of Clu1 granule localization next to mitochondria (Figure 3) would be more convincing if any of the experiment would be quantified. Especially in the case of panel 3G in Drosophila egg chambers where there are a lot of mitochondria, one wonders whether the closeness to mitochondria is just random. Furthermore, mdv1-signal does not look very convincing, being blurry and not dotty as expected. Thus, the conclusion that Clu1 granules partially colocalization with site of fission appears premature.
• Based on the ability of 1,6-hexanediol to dissolve the granules (Figure 4), the authors conclude that: "Clu1 foci have membraneless nature". As they correctly state in the discussion, treatment with 1,6-hexanediol can have other effects. I suggest to be more cautious with the conclusions or add additional experiments. Are the granules dynamics if using FRAP? Do they fuse?
• The experiment in which the granules are dissolved by treatment with RNAse is very interesting. However, per se this does not directly demonstrate that the granules contain mRNA. To state this the author should perform FISH experiments for example using a probe to detect poly-A.
• The authors show that puromycin prevents the granule formation before insulin addition in the fly. Are these results (upon RNAse treatment and puromycin treatment) recapitulated in the yeast system? The authors conclude that Clu1 formation depends on mRNAs being engaged in translation, but never show that the granules are site of active translation. More experiments in this direction (for example using puro-PLA of specific mRNAs) are missing and would clearly improve the manuscript.
• The interactome of Clu1-neighbouring proteins (Figure 6) is interesting and a valuable addition to data in other organisms. I am wondering why the authors have not used as a control a cytosolic BirA-GFP, which would have been the right control for this experiment, especially since GFP tends to form aggregates.
• Figure 7B: The log 2 FC for the changed proteins are in many cases small, implying that the difference in translation for these proteins is not so large. For this reason, it is relevant to know how was the statistical significance calculated for these MS measurements. In the supplementary Tables and in Fig 7B, a p value is indicated and it is not clear if this is a simple p value or an adjusted p value (FDR or q value). If not shown, I recommend showing the adjusted p value, so that one can have an idea of the solidity of the data and the claim. Again, this is an important piece of evidence, since the authors base on this experiment the conclusion that Clu1 controls translation of these mRNAs.

Minor comments:

• Figure 1: The change in Clu1 localisation in post-diauxic phase or upon changing of the medium is evident from the images shown. However, it seems that the experiment has been performed only once (the same for Figure 2). Is this the case? An important information would be to show the expression levels of Clu1-GFP in the different conditions. Does recruitment of CLU1 to granules associate to increased expression levels?
• Figure 2 A-B. The authors claim that the only stressor capable of inducing Clu1 granules formation alone is inhibition of complex IV activity via sodium azide treatment. Other mitochondrial stresses like CCCP treatment or OA treatment are efficient only when combined to starvation. It should be mentioned that sodium azide treatment is not only capable of inhibiting complex IV but has also uncoupling function.
• Figure 2 D-E: investigation of colocalization with Bre5 would help to understand how similar the yeast Clu1 granules are compared to the mammalian CLUH granules (Pla-Martin et al., 2020).
• Figure 8. This figure summarizes one of the most novel pieces of data about Clu1, the interaction with mRNAs via the ribosome. The way how panel A-C are represented is however a bit misleading. The Y axis in Figure B and C has the same amplitude as the one in A. Therefore, potential differences in Clu1-RNA pull-down in presence of EDTA or puromycin cannot be assessed. It is true that in presence of CHX there is much more pulled down RNA, but one cannot judge from these panels if there is any difference between Clu1 targets and controls also in the other conditions. The graphs should be modified and statistics added. In addition, RNAse treatment in panel L does not seem to have really worked.
• The authors should cite Vornlocher et al.. ( PMID: 10358023), who were the first to implicate Clu1 (Tif31) with translation.
• References 21 and 22 are the same.

Significance

The data reported in this manuscript are valuable, because they confirm an evolutionary conserved role of Clu1 in binding mRNAs for mitochondrial proteins and regulating their translation. It is also interesting that in yeast, similar to Drosophila and mammalian cells, Clu1 can form granular structures upon metabolic rewiring. A limitation of the study is that direct experiments to support the claim that Clu1 concentrates ribosomes engaged in translation are not provided. Furthermore, it is not clear what is the functional role of the Clu1 granules, since the proximity interactome and the binding of Clu1 to the polysomes is not affected by treatments that dissolve or stimulate granule formation. The study is of interest to a general cell biology audience.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this manuscript the authors use D. melanogaster and S. cerevisiae to study the role of CLUH in the translation of nuclear-encoded mitochondrial proteins. During conditions requiring aerobic respiration, CLUH forms RNA-dependent granules that localise in the proximity to mitochondria. Furthermore, the authors demonstrate that CLUH interacts with translating ribosomes to facilitate the translation of specific target mRNAs. For this, the authors use a combination of GFP-tagged CLUH models. BioID, polysome translating proteomics, RNA-IP. The authors' main conclusions are that (i) CLUH forms dynamic, membrane-less, RNA-dependent granules under conditions that demand aerobic respiration, (ii) CLUH interacts with specific mRNAs encoding metabolic factors, and (iii) CLUH interacts with the translating ribosome. The manuscript is well written and the conclusions stand in proportion to the experimental output and the results. The main concern is with regards to lack of advancement in relationship to published data.

Comments:

• To this reviewer it is not clear how CLUH can regulate the translation of specific mRNAs while being bound to ribosomes, regardless of being in a diffuse or granular state. The authors suggest that under metabolically active conditions, CLUH might aggregate translating ribosomes, forming the granular structures. How CLUH though can both be bound to translating ribosomes and recruit specific mRNAs at the same time is not explained.
• The authors might want to discuss how changes in metabolic demands signal the aggregation of CLUH, and how CLUH can recognise its target mRNAs.
• What was the rationale to perform the RIP or the PUNCH-P experiments only under non-challenged conditions, but not under conditions demanding aerobic respiration?
• If CLUH is ubiquitously bound to ribosomes, has CLUH been seen in any structural representation of the cytosolic ribosome?

Significance

CLUH has been studied in various publications, showing data very similar to that presented in this manuscirpt. However, the authors provide a comprehensive analysis on both yeast and fly CLUH. The strength of the manuscript is the combination of several elegant methods and genetically modified model systems in two species to elucidate the role of CLUH during the translation of specific mRNA. In my view through, the advancement of understanding the function of CLUH is limited.

Although the authors work in yeast and DM, the results seem applicable to other species, including humans, and thus, the presented results will be of interest in a range of researchers working in the field of metabolic regulation and gene expression.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

In this manuscript the authors study the Clustered mitochondrial proteins Clu of Drosophila melanogaster and Clu1 of Saccharomyces cerevisiae, two homologues of the mammalian protein CLUH. They show in compelling microscopy analysis that both proteins form granules. This was the case for flies fed on yeast paste after starvation and in yeast in post-diauxic phase, in respiratory media or during mitochondrial stress. They show that these granules are found in proximity to mitochondria and that they behave like liquid-liquid-phase separated condensates. They show by co-staining for P-bodies and stress granules that Clu1-granules are distinct from these RNA granules. Furthermore, they found that the formation required active translation. In the second part, they show that Clu1 interacts with ribosomal and mitochondrial proteins by BioID. The deletion of Clu1 leads to slightly impaired growth on media containing Ethanol as a carbon source. They find that nascent polypeptides of some mitochondrial precursor proteins are decreased in the deletion of Clu1 and conclude that Clu1 regulates translation of these proteins. Using RNA immunoprecipitation of Clu1-GFP in presence of cycloheximid, EDTA and puromycin. The mRNAs of nuclear-encoded mitochondrial proteins found to be interacting with Clu1 were purified in conditions when the ribosomes are intact and the RNAs showed no interaction when ribosomes were disassembled. They show in sucrose gradients that Clu1 co-migrates with polysomes independent of its distribution state or carbon source. However, when cells are grown in conditions of granule formation, then polysomes and Clu1 run less deeply into the gradient. Form these data, the authors conclude that Clu/Clu1 regulates the translation of nuclear-encoded mitochondrial proteins.

Major comments:

• The authors state that Clu1 is regulating translation during metabolic shifts. However, it is not clear what the real impact on mitochondrial function is. They show that there is a minor growth defect on ethanol media when CLU1 is deleted. However, if Clu1 is necessary mainly for adaptation, the phenotype will be strongest observed in conditions where cells switch carbon sources. Growth curves would be suitable in which the lag-phase of yeast cells precultured either in glucose or glycerol switched to media of different carbon sources (glucose to glycerol or glycerol to glucose) are measured. One would expect that the deletion mutant shows a longer lag-phase compared to the wild type when shifted from glucose to glycerol media. In line with this, how different is the mitochondrial proteome of the WT and the mutant? Do hits of the BioID, RIP and Punch-P experiments change at steady state or during metabolic shifts? Either proteomics of isolated mitochondria or western blots of whole cells or isolated mitochondria of WT and the deletion mutant grown in conditions of Clu1-granule formation or no granules for the hits could answer this question.
• The authors analyze RNAs bound in polysomes to assess translation efficiency. Translation efficiency is usually calculated by the fraction of RNA bound by ribosomes to the total RNA amount of an RNA species. Thus, doing RT-qPCR from whole cells would be necessary to assess if the occupancy of ribosomes on the transcripts is due to changes in RNA abundance or other regulatory pathways and would help to further assess what causes the observed changes.

Optional:

• The authors show a co-localization of Clu/Clu1 with mitochondrial fission factors and conclude that the granules appear likely near fission sites. Indeed, CLUH has been implied in the past to play a role in mitochondrial fission (Yang, H., Sibilla, C., Liu, R. et al. Clueless/CLUH regulates mitochondrial fission by promoting recruitment of Drp1 to mitochondria. Nat Commun 13, 1582 (2022). https://doi.org/10.1038/s41467-022-29071-4). Thus, are fission sites required for Clu-granule localizations? What is the role of the mitochondrial network integrity for the granule distribution? Expressing Clu-GFP/Clu1-GFP in cells depleted for the fission factors would provide information on that.
• The author assess convincingly that Clu1 interacts with ribosomes and runs with polysomal fractions. However, how it actually regulates translation is not clear. To answer this question, selective ribosomal profiling would be necessary. The authors have established conditions which would be suitable for the experiment. They could use crosslinking and sucrose cushions to IP ribosomes with Clu1-GFP bound to be used for ribosomal profiling. However, this experiment is quite time-intensive (3-4 months) and expensive, thus, an optional suggestion.

Minor comments:

Fig1: B, C, please add scale bars into the zoom ins.

Fig 2 would profit from inlets of zoom ins to visualize the distribution better.

Fig.3: Panel C does not really add much information. I would rather remove it or put it into supplements and therefore show a zoom of Panel E with a line plot showing the rings. It is not clear from the represented images where the rings are formed.

Fig.3 panel F: Max projections are not appropriate to show colocalization as they can lead to false-positive overlaps. Just remove the max projections.

References 21 and 22 are the same.

Significance

This manuscript shows in a convincing way that Clu and Clu1 form RNA granules and that Clu1 interacts with ribosomes. It is written in a clear way and the figures support the conclusions drawn in the text. The finding that Clu/Clu1 is important for metabolic adaptation has not been shown in fly or yeast to my knowledge. It is in line with findings for the mammalian homologue CLUH. Thus, the findings are supported by earlier work. This study is of value for a broader audience of the basic research field, especially of the mitochondrial and RNA granule field, as it supports the idea of post-transcriptional regulation of nuclear-encoded mitochondrial protein gene expression for dynamic adaptation of mitochondrial function. The conditions when Clu granules form is studied in detail, followed up by identification of target RNAs and interaction partners. Though the interaction of Clu1 with ribosomes is shown in a compelling way, a detailed mechanism of the function of Clu/Clu1 is missing and would require more experiments. Thus, even though a detailed mechanism is missing, the study does expand on our understanding of Clu/Clu1 in regulating mitochondrial biogenesis and is therefore of high interest of the mitochondrial field.

Expertise: mitochondria, yeast, RNA granules, mitochondrial biogenesis, next-generation sequencing, fluorescence microscopy

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer 1

1. The structures of the lamina propria of murine colon mucosa are nicely described. However, in the introduction of the manuscript the structures of fibroblasts, myofibroblasts and ECM are not described. The structures of the lamina propria of murine colon mucosa should be well described in the induction and discussed in the discussion.

We will revise the Introduction to include a more detailed description of fibroblasts, myofibroblasts, and the ECM within the lamina propria of the murine colon mucosa. We will also expand the Discussion section to address these structures in the context of our findings.

2. The UMAP plot suggests potential heterogeneity within Cluster 1, raising questions about whether the chosen clustering resolution (e.g., parameter settings in Seurat's "FindClusters") optimally captures subpopulations.

We appreciate this insightful observation. We agree that the UMAP plot suggests potential heterogeneity within Cluster 1 and that the current clustering resolution may not fully capture underlying subpopulations. We could revisit the clustering parameters and explore reclustering at a lower resolution. However, we note that lowering the resolution often increases the total number of clusters, which may introduce noise and complicate biological interpretation. To more precisely dissect the heterogeneity within Cluster 1 while minimizing artificial subdivisions, we propose to perform subclustering specifically within Cluster 1.

3. Some subpopulations express marker genes characteristic of pericytes and smooth muscle cells (e.g., Desmin). How did the authors ensure proper discrimination between fibroblasts and these other cell types?

We thank the reviewer for this important comment. We acknowledge the challenge in distinguishing between fibroblasts, pericytes, and smooth muscle cells (SMCs) based solely on single-cell RNA sequencing data, particularly given the overlapping expression of markers such as Desmin.

Pericytes vs. Myofibroblast/SM-Pericyte-Like Fibroblasts: Due to the highly similar transcriptional profiles of pericytes and pericyte-like fibroblasts, scRNA-seq alone does not allow for unambiguous discrimination between these populations. However, we were able to distinguish them based on morphology and spatial localization observed in high-resolution imaging. Notably, we identified a population of large (50–150 µm), elongated myofibroblast/SM-pericyte-like fibroblasts that, unlike typical pericytes, are not positioned directly on blood vessels but are distributed around the crypts. Some of these cells also appear to contact both blood vessels and the muscle layer, raising the possibility that they represent a specialized pericyte-like population. While their precise function remains uncertain, we agree that further characterization is warranted. To address this, we propose additional staining for canonical pericyte markers to help clarify their identity and spatial relationship to the vasculature.

Smooth Muscle Cells vs. Myofibroblast/SM-Pericyte-Like Fibroblasts: We are confident that the analyzed fibroblast populations do not include smooth muscle cells. The mucosa was carefully dissected and separated from the underlying smooth muscle layer prior to RNA sequencing, which was performed exclusively on the mucosal compartment. Therefore, contamination by SMCs is unlikely.

4. The manuscript also did not show the distribution and structures of ECM. It is better to show the relationships of fibroblasts and myofibroblasts with in the lamina propria of murine colon mucosa.

In the supplementary material we show distribution of main ECM proteins such as Laminin, Collagen I, Collagen IV, and Fibronectin1.

5. The integration with previously published datasets lacks clear connection to the authors' own findings. A more detailed comparison and discussion of how these integrated analyses relate to the newly generated data would improve the manuscript's coherence.

We thank the reviewer for this helpful comment. Our RNA-seq dataset shows strong consistency with previously published datasets, supporting the robustness of our fibroblast isolation and transcriptional profiling strategy. We agree that a more explicit integration and comparison will improve the manuscript. We have now revised the Discussion to better highlight the spatial localization and organization of the different fibroblast populations identified in our study, with an emphasis on the duality of their functions. In particular, we discuss how our findings extend existing datasets by providing spatial context and functional insights that were not previously resolved. These comparisons underscore the novelty and value of our integrated approach.

6. While the authors focus on colonic mucosa, the integrated public datasets include data from both colon and small intestine. Were these distinct tissue sources accounted for in the analysis? Clarification on this point is necessary to ensure the validity of comparisons.

We thank the reviewer for raising this important point. Among the integrated datasets, only one—McCarthy et al. (GEO GSE130681)—originates from the small intestine; all others, including our own, were derived from the colon. Specifically, we used the following datasets:

• GEO GSE113043 (Degirmenci et al., PMID: 29875413) – Colon (1 sample)
• GEO GSE114374 (Kinchen et al., PMID: 30270042) – Colon (3 samples)
• GEO GSE130681 (McCarthy et al., PMID: 32884148) – Small intestine (2 samples)
• GEO GSE142431 (Roulis et al., PMID: 32322056) – Colon (5 samples) We selected these datasets based on their relevance to fibroblast biology, particularly those that specifically focused on mural fibroblasts. The inclusion of the McCarthy dataset was guided by its high-quality profiling of fibroblast populations and its utility in expanding our comparative framework.

Importantly, review by McCarthy et al. (https://doi.org/10.1038/s41556-020-0567-z) reported minimal differences in fibroblast clustering between the small intestine and colon. Our integrated analysis supports this conclusion: fibroblasts from both regions consistently co-cluster, indicating a high degree of transcriptional similarity. This suggests that inclusion of the small intestine dataset did not bias or compromise the integrity of our colon-focused findings.

Nevertheless, our primary emphasis remains on the colon, particularly due to the relative scarcity of studies addressing fibroblast localization and morphology in this tissue compared to the small intestine. Additionally, at the time of analysis, the datasets we used represented the most comprehensive publicly available single-cell profiles of intestinal mural fibroblasts.

7. Many aspects of the described fibroblast subpopulations, including their single-cell expression profiles and physiological functions, appear to have been previously reported. The authors should more explicitly highlight the novel contributions of their work to advance our understanding of intestinal fibroblast biology.

We thank the reviewer for this important observation. While it is true that aspects of fibroblast heterogeneity have been previously reported, our study provides several novel contributions that advance the current understanding of intestinal fibroblast biology. We will revise the manuscript to more explicitly highlight the following key findings:

1. Functional distinction between ECM production and contractility: Our integrative analysis reveals a clearer separation between fibroblast subpopulations based on their functional specializations—specifically, ECM production versus contractile properties. This distinction has not been well delineated in prior studies and is particularly relevant in the context of inflammatory bowel disease, where fibrosis remains a major complication. Our findings may help identify specific fibroblast subtypes that contribute to pathological remodeling.
2. Detailed characterization of fibroblast localization and morphology: We provide new spatial insights by demonstrating the lack of overlap between GFP⁺ and CD34⁺ basket cell populations in vivo. Additionally, we highlight the presence of large, elongated myofibroblasts and pericyte/smooth muscle-like fibroblasts that span from the vasculature to the underlying muscle layer—morphologies and arrangements that have not been thoroughly described before. These observations offer a more refined anatomical and functional framework for understanding fibroblast roles within the colonic mucosa. We will revise both the Results and Discussion sections to more explicitly emphasize these novel contributions.

Reviewer 2:

Major points:

1. The order of the present manuscript should be reconstructed. The main message is in the discussion part. It is worth bringing it to the front.

We appreciate this thoughtful suggestion. We agree that the main message of the manuscript is currently more prominent in the Discussion section, and bringing it forward would improve the overall clarity and impact of the work. We will restructure the manuscript accordingly to ensure that the key findings and their significance are introduced earlier and more clearly communicated throughout the text.

2. Figure 1A, the authors employed the "vimentin+" filter to distinguish between fibroblasts and other cell types in the single-cell RNA sequencing (scRNA-seq) data. However, they did not provide a rationale for this choice in the manuscript. It would be worthwhile to consider the incorporation of an "Epcam-" or "E-cadherin-" filter as well, given the potential impact on the subsequent analysis's significance. Notably, the original UMAP plot generated before the application of the "vimentin+, Krt8-" filter, is absent from both the main figures and the supplementary data. The availability of this data is crucial for the identification of specific fibroblast populations among the sorted cells.

The rationale for using the “vimentin⁺” filter is based on its long-standing use as a canonical marker for fibroblasts and mesenchymal cells in both developmental and adult tissues, including the intestinal lamina propria. Vimentin has consistently been used to distinguish fibroblasts from epithelial and immune cell populations in scRNA-seq studies.

Regarding the exclusion of epithelial cells, we chose to apply a “Krt8⁻” filter instead of “Epcam⁻” or “E-cadherin⁻”, as Krt8 is a highly specific marker for colonocytes in the intestinal epithelium. We found this to be a reliable criterion for excluding epithelial cells in our dataset. We will revise the Methods section to clearly explain this rationale and selection.

Additionally, we agree that the original UMAP plot—prior to the application of the “vimentin⁺, Krt8⁻” filter—would provide valuable context. We will include this plot in the supplementary figures to allow better visualization of the initial clustering and to support the identification of fibroblast populations among the sorted cells.

3. Page4 line 12, the authors claim that they did not find specific markers for the cluster 1, despite the fact that cluster 1 is distinctly separated from clusters 0, 5, 4 and 3 in figure 1B. Furthermore, the cells in the cluster 1 do not cluster together based on the resolution applied in the present manuscript. The authors claim that cells in cluster 1 are in a transition state, and therefore, they did not include them in the functional analysis. However, later they claim that the cluster 1 are multipotent progenitors, which is not clear.

We appreciate the reviewer’s careful reading and valuable critique. We acknowledge the confusion regarding the identity and interpretation of Cluster 1 and would like to clarify our reasoning and planned revisions.

When identifying marker genes using Seurat’s FindMarkers() or FindAllMarkers() functions, the output highlights genes that are significantly enriched in a given cluster relative to others—but these genes are not necessarily uniquely or exclusively expressed in that cluster. This is the case with Cluster 1: although it is spatially distinct in the UMAP (Figure 1B), many of the top-ranked marker genes are also expressed in other clusters, albeit at lower levels. As a result, defining Cluster 1 based solely on unique gene expression signatures is challenging, and we initially interpreted this cluster as a “transitional population” due to its ambiguous marker profile.

However, we acknowledge the apparent inconsistency in referring to Cluster 1 as both "in transition" and "multipotent progenitors." We will clarify our interpretation and terminology in the revised manuscript. Specifically, we will refer to Cluster 1 as a __ transitory population__, and provide a more nuanced discussion of its potential roles.

As mentioned in our response to Reviewer 1 (Comment 2), we will also perform reclustering within Cluster 1 to better explore its internal heterogeneity. Additionally, we will now include Cluster 1 in the functional enrichment analysis to further assess its biological relevance and contribution to fibroblast diversity.

4. Figure 1E and F, authors only use gene ontology to define the functions of different clusters of fibroblasts which constrain the present manuscript at the hypothesis stage. To substantiate the claims, it is imperative to conduct more precise experiments. At the very least, co-staining with cluster marker genes and candidate genes identified in GO analysis is necessary. In the event that antibodies are not available, RNA scope can serve as a viable alternative. Further functional experiments will be required to prove their unique function. For instance, the identification of specific cell surface markers to isolate different clusters of fibroblasts for coculture with intestinal organoids in vitro can be facilitated by scRNA-seq data.

We appreciate the reviewer’s insightful suggestions regarding the functional validation of GO-based predictions.

While we recognize that RNAscope is a valuable alternative when antibodies are unavailable, its use requires much thinner tissue sections than those employed in our current imaging approach. Our analysis is based on thicker sections, which preserve the 3D architecture and spatial relationships of fibroblasts within the colonic mucosa—an essential aspect of our study. Transitioning to thinner sections would compromise our ability to visualize these cells in their full anatomical context.

To suppor the GO analysis with experimental validation, we will include __co-staining for cluster __marker genes along with representative candidate genes____ identified through GO analysis to better substantiate the predicted functions of different fibroblast clusters.

We acknowledge the importance of functional studies such as co-culture assays with intestinal organoids, and indeed, several such experiments have been reported by other groups. Additionally, isolating specific fibroblast populations via FACS sorting for in vitro studies presents practical challenges, including low cell survival rates, which limit the feasibility of downstream functional assays. Thus, we believe that these types of experiments are beyond the scope of the current manuscript. We hope that our integrative approach and spatial validation will serve as a valuable foundation for future functional investigations into fibroblast biology.

5. DAPI staining is absent in the majority of the images, which complicates the task of distinguishing cells from different clusters. Multiplex staining is necessary to show all specific markers: EGFP, SMA, CD34, Desmin, Pdgfra, Pil6, and Clu, regarding six clusters in one section or image.

We appreciate the reviewer’s comment and the emphasis on the importance of cellular context in multiplex imaging.

We acknowledge that DAPI staining is absent in some of the presented images, which may limit nuclear visualization and make it more challenging to distinguish cell boundaries. However, to achieve high-content multiplexing, we employed protocols allowing up to 5–6 fluorophores per section, as previously demonstrated by Chikina et al. (Cell, 2020). Due to spectral limitations and the risk of fluorophore overlap and signal bleed-through, we occasionally excluded DAPI to allocate the 405 nm channel for markers of greater relevance to our study. In these cases, Tomato or EGFP signals served as effective surrogates for cellular localization, as they label cell membranes, providing sufficient morphological context.

Regarding multiplex staining for Pi16 and Clu, we tested several commercially available antibodies, but unfortunately, none yielded specific or reproducible signals in our hands. As a result, we were unable to include these markers reliably in our multiplex panels.

6. Figure 4, the authors utilize supervised methods to execute trajectory analysis, defining cluster 1 as the initial point based on its hybrid expression state of genes. This assertion, however, lacks sufficient substantiation, as cluster 1 could also function as a transition point, not necessarily an initial point.The data presented in the current manuscript is inadequate to support the conclusion of multipotency in cluster 1.To substantiate these claims, the authors should employ additional evidence, such as SENIC analysis, to demonstrate the expression of specific transcription factors for each lineage along the trajectory. In order to substantiate the assertion that cluster 1 is a multipotent progenitor capable of differentiating into other specific populations, such as fibroblasts, further functional experiments are required. These experiments could include isolating the population in question and conducting a differentiation test in vitro or tracking the population's response to wound healing.The absence of immunofluorescence images or gene signatures for this cluster in the study is a cause of confusion for the reader.

We thank the reviewer for this thoughtful and constructive comment. We agree that Cluster 1 could plausibly represent either an initial or transitional state. In trajectory analysis, the starting point must be defined, and we selected Cluster 1 due to its hybrid gene expression profile—exhibiting low-level expression of marker genes associated with multiple other clusters—suggesting a less differentiated or “primed” state. However, we fully acknowledge that this assignment does not preclude its interpretation as a transitional population, and we will revise the manuscript text to reflect both possibilities more clearly and cautiously.

We appreciate the suggestion to perform __SENIC __analysis (https://www.nature.com/articles/nmeth.4463). This algorithm aims to identify gene regulatory networks and their associated transcription factors for each cell cluster. While interpreting such analysis can be quite challenging, it could provide interesting insights and thus we propose to apply it.

Regarding functional validation, we agree that experiments such as isolation and in vitro differentiation assays, or in vivo lineage tracing during injury models, would offer more definitive insights. However, as we noted, the lack of specific surface markers currently makes it challenging to isolate Cluster 1 by FACS for such assays.

We also acknowledge the reviewer’s concern about the __lack of immunofluorescence images or distinct gene signatures__for Cluster 1. We will revise the text to clearly communicate that this limitation.

7. Figure 5B, the data set of Kinchen et al is from human samples. Is it relevant and significant to merge murine data and the human data together?

We appreciate the reviewer’s attention to detail. To clarify, the dataset from Kinchen et al__.__ used in Figure 5B refers exclusively to their murine samples, not the human data. Only murine datasets were included in our analysis to ensure consistency and biological relevance. Therefore, merging the Kinchen murine data with other murine datasets in Figure 5B is both appropriate and justified.

We will revise the figure legend and Methods section to clearly state that only mouse data were used throughout the analysis.

Minor points:

8. The chapter entitled "Subepithelial Fibroblast Do Not Proliferate" is not necessary to be an independent chapter. It can be considered a fusion chapter, as it is combined with Chapter 2. Further experiments such as Brdu or Edu are needed to strengthen the current hypothesis.

We agree with the reviewer that this section does not require a standalone chapter and would be better integrated into Chapter 2. We will revise the manuscript accordingly.

In addition, to further support our observations regarding fibroblast proliferation, we will perform a 2-hour EdU pulse-chase experiment and include the results in the revised manuscript. We believe this will strengthen our conclusions and provide more direct evidence regarding the proliferative status of subepithelial fibroblasts.

1. DAPI staining is absent in majority of the images.

indeed this is a limitation of the unmixing technique we use.

10. Put the number of each cluster next to the arrow in all the IF images.

We will do this.

11. Immunofluorescent staining of cluster markers identified in the previous studies should be included in the present study such as: CD81, FoxL1, Myh11, Pdgfrb and Gli1.

We will include those markers.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The summitted article entitle "Intestinal fibroblast heterogeneity: unifying RNA-seq studies and introducing consensus-driven nomenclature" by Glisovic et al., identify six distinct populations of fibroblast with unique molecular signatures, spatial localization and specific function in mouse colon using scRNA-seq. Moreover, with different bioinformatic methods, they show the potential differentiation trajectories of fibroblast in mouse colon mucosa. Finally, they propose a standardized nomenclature for colonic fibroblast by integrating the data of this manuscript and the four published scRNA-seq data of mouse and human intestinal colonic fibroblast. Several similar studies cited by the authors in the present manuscript have been done and the different populations of colonic fibroblasts have been well characterized in these previous studies. Here the authors utilized another mouse model, the "aSMAcreERT2" to target the murine colonic fibroblast population which is novel compared to previous published data. Although the authors have provided multiple bioinformatic analyses and immunofluorescent staining of certain markers to support their conclusions, many points are overclaimed or not clear based on the data of the present manuscript, especially for the differentiation trajectories and unique function of different clusters of subepithelial colonic fibroblast. Functional experiment data are absent from the present manuscript.

Major comments

• The order of the present manuscript should be reconstructed. The main message is in the discussion part. It is worth bringing it to the front.
• Figure 1A, the authors employed the "vimentin+" filter to distinguish between fibroblasts and other cell types in the single-cell RNA sequencing (scRNA-seq) data. However, they did not provide a rationale for this choice in the manuscript. It would be worthwhile to consider the incorporation of an "Epcam-" or "E-cadherin-" filter as well, given the potential impact on the subsequent analysis's significance. Notably, the original UMAP plot generated before the application of the "vimentin+, Krt8-" filter, is absent from both the main figures and the supplementary data. The availability of this data is crucial for the identification of specific fibroblast populations among the sorted cells.
• Page4 line 12, the authors claim that they did not find specific markers for the cluster 1, despite the fact that cluster 1 is distinctly separated from clusters 0, 5, 4 and 3 in figure 1B. Furthermore, the cells in the cluster 1 do not cluster together based on the resolution applied in the present manuscript. The authors claim that cells in cluster 1 are in a transition state, and therefore, they did not include them in the functional analysis. However, later they claim that the cluster 1 are multipotent progenitors, which is not clear.
• Figure 1E and F, authors only use gene ontology to define the functions of different clusters of fibroblasts which constrain the present manuscript at the hypothesis stage. To substantiate the claims, it is imperative to conduct more precise experiments. At the very least, co-staining with cluster marker genes and candidate genes identified in GO analysis is necessary. In the event that antibodies are not available, RNA scope can serve as a viable alternative. Further functional experiments will be required to prove their unique function. For instance, the identification of specific cell surface markers to isolate different clusters of fibroblasts for coculture with intestinal organoids in vitro can be facilitated by scRNA-seq data.
• DAPI staining is absent in the majority of the images, which complicates the task of distinguishing cells from different clusters. Multiplex staining is necessary to show all specific markers: EGFP, SMA, CD34, Desmin, Pdgfra, Pil6, and Clu, regarding six clusters in one section or image.
• Figure 4, the authors utilize supervised methods to execute trajectory analysis, defining cluster 1 as the initial point based on its hybrid expression state of genes. This assertion, however, lacks sufficient substantiation, as cluster 1 could also function as a transition point, not necessarily an initial point. The data presented in the current manuscript is inadequate to support the conclusion of multipotency in cluster 1.To substantiate these claims, the authors should employ additional evidence, such as SENIC analysis, to demonstrate the expression of specific transcription factors for each lineage along the trajectory. In order to substantiate the assertion that cluster 1 is a multipotent progenitor capable of differentiating into other specific populations, such as fibroblasts, further functional experiments are required. These experiments could include isolating the population in question and conducting a differentiation test in vitro or tracking the population's response to wound healing. The absence of immunofluorescence images or gene signatures for this cluster in the study is a cause of confusion for the reader.
• Figure 5B, the data set of Kinchen et al is from human samples. Is it relevant and significant to merge murine data and the human data together?

Minor comments

• The chapter entitled "Subepithelial Fibroblast Do Not Proliferate" is not necessary to be an independent chapter. It can be considered a fusion chapter, as it is combined with Chapter 2. Further experiments such as Brdu or Edu are needed to strengthen the current hypothesis.
• DAPI staining is absent in majority of the images.
• Put the number of each cluster next to the arrow in all the IF images.
• Immunofluorescent staining of cluster markers identified in the previous studies should be included in the present study such as: CD81, FoxL1, Myh11, Pdgfrb and Gli1.

Significance

In this study, the researchers employed an alternative mouse model, the "aSMAcreERT2," to target the murine colonic fibroblast population. This approach represents a novel contribution to the field, offering a fresh perspective on previous findings. While the authors have presented several bioinformatic analyses and immunofluorescent staining of specific markers to support their conclusions, certain aspects of their argument require further elaboration or clarification, particularly regarding the differentiation trajectories and unique functions of the various clusters of subepithelial colonic fibroblasts. The present manuscript is constrained at the descriptive level due to an absence of functional experiment data.

Strengths: The authors utilize "aSMAcreETR2" as a research model to target murine colonic fibroblasts, a novel approach that complements previously published data. By comparing and combining four published single-cell RNA sequencing (scRNA-seq) of colonic fibroblasts, they proposed a novel classification with five distinct subpopulations: telocytes, trophocytes/extracellular matrix (ECM) fibroblast, fibroblast, myofibroblast, and smooth muscle/pericyte-like fibroblast.This new classification, together with their unique molecular signature, can be useful for people in the colon and intestine research field. However, the manuscript is not without its limitations. First, the novel classification and unique molecular signature are not substantiated by functional experimentation, which is essential for validating the fibroblast subcluster's functionality. Additionally, the characterization of cluster 1 is lacking, particularly concerning its ability to differentiate into the five distinct subcultures, which is crucial for confirming its status as a multipotent progenitor. Despite the proposal of a novel classification and detailed molecular signature of the colonic fibroblasts, no isolation strategy is proposed in the present manuscript to allow further characterization. If the authors can address these points, the manuscript can make a significant contribution to the field. This study might interest people who perform basic research in the intestine and colon.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study utilizes scRNA-seq to delineate six fibroblast subpopulations in mouse colonic mucosa, revealing their molecular heterogeneity, functional specialization, and spatial distribution. The high-quality confocal microscopy images effectively illustrate the spatial distribution of cells within the colon mucosa. However, several concerns should be addressed:

1. The structures of the lamina propria of murine colon mucosa are nicely described. However, in the introduction of the manuscript the structures of fibroblasts, myofibroblasts and ECM are not described. The structures of the lamina propria of murine colon mucosa should be well described in the induction and discussed in the discussion.
2. The UMAP plot suggests potential heterogeneity within Cluster 1, raising questions about whether the chosen clustering resolution (e.g., parameter settings in Seurat's "FindClusters") optimally captures subpopulations.
3. Some subpopulations express marker genes characteristic of pericytes and smooth muscle cells (e.g., Desmin). How did the authors ensure proper discrimination between fibroblasts and these other cell types?
4. The manuscript also did not show the distribution and structures of ECM. It is better to show the relationships of fibroblasts and myofibroblasts with in the lamina propria of murine colon mucosa.
5. The integration with previously published datasets lacks clear connection to the authors' own findings. A more detailed comparison and discussion of how these integrated analyses relate to the newly generated data would improve the manuscript's coherence.
6. While the authors focus on colonic mucosa, the integrated public datasets include data from both colon and small intestine. Were these distinct tissue sources accounted for in the analysis? Clarification on this point is necessary to ensure the validity of comparisons.
7. Many aspects of the described fibroblast subpopulations, including their single-cell expression profiles and physiological functions, appear to have been previously reported. The authors should more explicitly highlight the novel contributions of their work to advance our understanding of intestinal fibroblast biology

Significance

The proposed standardized nomenclature for intestinal fibroblasts represents a valuable contribution toward unifying classification in the field.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements

We thank the editor for handling our manuscript and the reviewers for their constructive critiques. We are deeply convinced that the reviewers’ suggestions have substantially raised the quality and possible impact of our manuscript. We also like to thank the reviewers for their judgements that the subject of our manuscript is biologically and clinically significant and of high importance, and that our manuscript might help to increase focus and visibility for affected individuals.

New text passages in the manuscript are colored in red. Below is a point-by-point response to the reviewers’ comments.

2. Point-by-point description of the revisions

Response to reviewer 1 comments

Major comments

Point 1-1

The authors performed qRT-PCR validation for markers of differentiation and hypoxia, with a major absence of VEGF and HIF1a. The paper would be strengthened by mention of these factors, especially by qRT-PCR or Western blot.

We thank the reviewer for the suggestion to include the bona fide hypoxia markers Vegfa and Hif1-alpha. We followed the suggestion and performed qRT-PCR on Vegfa transcripts at each tested condition (Figs. 1A,2A,3A,4A,5A,5D,5I,5N). As Hif1α is rather regulated on protein than on transcript level, we followed the advice to perform Western blots. We analyzed Hif1α protein levels on proliferating cells and quantified by normalization to actin (Figs. 1B,C and 5 B,C).

Point 1-2

Please provide justification of selection 0.5% as their hypoxic condition or perhaps repeat experiments in a less extreme environment to see if their conclusions still hold true.

We admit that our approach to use 0.5% hypoxia was a drastic challenge for the cells. It should be noted, however, that physiologic oxygen levels during pregnancy at times drop to lower than 1% (Hansen et al, 2020; Ng et al, 2017). In the first place, we had used oxygen levels lower than this, because we had wanted to ensure that we can detect responses by bulk RNA-seq with a limited number of samples. As we had many conditions to compare, we did not want to use more than 3-4 samples per condition. The fact that the cells showed normal proliferation underscores the fact that 0.5% O2 per se was not so low that it would be overly stressful to the cells.

Nevertheless, we are very grateful to the reviewer for the suggestion to include a milder hypoxic condition. We chose 2% O2, because this equals the physiological oxygen concentration shortly before the onset of cranial neural crest cell (CNCC) differentiation. We could recapitulate the phenomenon of impaired differentiation to chondrocytes, osteoblasts and smooth muscle cells at these mild hypoxic conditions, as shown by qRT-PCR and immunofluorescence of typical markers (Figs. 5D-R). Moreover, the differentiation-specific induction of the two central hypoxia-attenuated risk genes associated with orofacial clefts that we had identified by our bioinformatic analyses at 0.5% O2 (Boc and Cdo1), was still observable at 2% O2 (Figs. S6C,D). Interestingly, in some rare cases, the attenuation of induction was lost or not as drastic as in 0.5% O2.

We are convinced that the experiments at 2% O2 strongly increased the relevance of our manuscript, because we thus detected that oxygen levels prevailing shortly before the onset of CNCC differentiation still can influence their differentiation. This leads to the conclusion that only slight decreases of intra-uterine oxygen levels indeed might interfere with correct differentiation of CNCC.

Point 1-3

Standard immunohistochemistry or histology of differentiated cells would strengthen the authors' claims of reduced differentiation under hypoxic conditions, e.g., Alcian blue, alk-phos or Alizarin red, and smooth muscle actin or other indicator.

We are grateful to the reviewer for the suggestion to include stainings of cells, as these stainings visualized the drastic effects of hypoxia on the cells. We performed immunofluorescent stainings against at least one marker protein for each differentiation paradigm. At 0.5% O2, each protein signals were nearly completely absent and cell morphology was disrupted (Figs. 2E,F, 3E, 4E). At 2% O2, we detected some more protein deposition than at 0.5%. Importantly, cells had retained their normal shape at mild hypoxia (Figs. 5H,M,R, S5A).

Point 1-4

The authors identify a few genes that appear down-regulated in all three differentiation conditions. If it is within the scope of the study, it would strengthen the claim of these genes' function to show the effect of knock-down or knock-out for validation.

We thank the reviewer for the suggestion of gene knock-down or knock-out in order to prove functional relevance of our findings. As this would have been too much effort and beyond the scope of our study, we rather followed the suggestion of reviewer 2 (cf. points 2-6, and 2-8) that headed to the same direction: we mined publicly available sequence data on orofacial development for gene expression or marks of active enhancers. We found robust expression of the two central hypoxia-attenuated OFC risk genes Boc and Cdo1 during human craniofacial development (Fig. 7A) and we identified enhancers that are active in embryonic craniofacial mouse tissue (Fig. 7B). Moreover, we detected expression of both genes during murine craniofacial development in undifferentiated mesenchymal cells, osteoblasts, chondrocytes and smooth muscle cells with the help of a single cell RNA-seq dataset (Figs. 7C-E, S6B).

Thus, we found evidence for the in vivo relevance of Boc and Cdo1 and could rule out a possible important role of Actg2, the third gene we had identified. We therefore are grateful for the suggestion to circumvent gene knockouts by reviewer 2, as we think these data strongly emphasized the importance of our findings.

Point 1-5

Another major critique lies in the initial claim that proliferation of O9-1 cells is not significantly impacted by hypoxia. In figures 1E-H, photograms of the cells cultured 24 -72 hours and quantifications of live vs dead cells are shown as evidence for this argument. However, the increased density of cells in normoxic conditions may be a confounding variable in this assay. It would be interesting for the researchers to assess the percent of dead vs alive cells between normoxic and hypoxic conditions when the plates reach equivalent densities.

We apologize for the use of image sections from photographs with different cell densities. Of course, as demonstrated by our quantification, cell densities between 0.5% and 21% O2 in total were equal (cf. Figs. 1D,E). We therefore replaced the formerly used sections with new image sections with equal cell numbers.

We thank the reviewer for the suggestion to examine if cell numbers influence cell death rates. We followed this advice by several approaches: first, we seeded cells at different densities, incubated them for 72 h (the same time span where a minimal difference had been detected) and performed live/dead stainings (Fig. S1B). The seeding density did not affect percentages of dead cells and the values were in the same range as in our initial experiment (Fig. 1J). Moreover, we performed TUNEL stainings of apoptotic cells at different time points to have an additional readout of cell death (Figs. 1K,L). As expected, the percentages of TUNEL-positive cells were identical between hypoxic and normoxic cells at all analyzed time points.

We therefore concluded that hypoxia does not influence the rate of cell death of proliferating CNCC and accordingly specified our wording in the results section.

Point 1-6

At end of Fig 1 section authors attempt to tie phenotypes observed in a cell line in vitro to the complex biological processes. They are not comparable and in vivo models would be better suited for these types of comparisons.

We apologize for the overconfident wording in our manuscript. Of course, our in vitro experiments cannot fully simulate the complex developmental processes taking place in vivo. We therefore changed the text to a more careful formulation. Moreover, we kept the wording in the discussion section that we cannot exclude that in the in vivo situation proliferation of CNCC is also affected by low oxygen levels because nutrients might not be available in such excess as they are in cell culture.

Point 1-7

Fig 2: if qRT-PCR did not show statistically different results between experimental and control groups why move on to bulk RNA seq?

We apologize that the sentence about statistical significance was misleading. What we wanted to express is that there was only a little difference (if any at all) between differentiated cells at 0.5% O2 and proliferating cells at 0.5% O2 or 21% O2. For the sake of clarity and readability, we deleted this misleading sentence.

Point 1-8

Fig 5: hypoxia this intense is going to affect broad range of biological processes and genes. Finding a few genes that are affected in extreme hypoxia that are also risk genes is highly unlikely. How can the authors be assured that these overlaps are actually significant and not just by chance?

We thank the reviewer for the suggestion to test for statistical significance. We tested significance of the overlap of respective gene sets (nsOFC vs. hyp-a; OFC vs. hyp-a) by Fisher’s exact test. We included Venn diagrams depicting the overlap and present the exact p-values (Figs. S5C,D). In each case where overlap of genes occurred, p-values indicated significance.

Point 1-9

Would appreciate discussion on how examination of neural crest is relevant for OFC, as most animal models of OFC demonstrate the pathogenesis in embryonic epithelium or periderm, not in the neural crest. Defects in neural crest are associated with other congenital craniofacial anomalies such as craniosynostosis or complex (Tessier) clefts, not the typical orofacial cleft. Please revise rationale of study, interpretation of data and Discussion to specifically state how neural crest cells are involved in the pathogenesis of orofacial cleft.

We apologize for not pointing out enough the role of epithelial cells in the emergence of orofacial clefts. We revised our introduction, results and discussion sections in this regard and emphasized the role of epithelial cells. Importantly, we addressed the possible influence of the results gained in CNCC on epithelial cells by analyzing scRNA-seq data with the algorithm CellChat, as suggested by reviewer 2 (cf. point 2-8). We detected several cell communication pathways from CNCC to epithelial cells which contain components that are misexpressed upon hypoxia in our dataset (Figs. 7F-I). Therefore, during hypoxia, these pathways might influence epithelial cells and therefore indirectly cause orofacial clefts. We outlined this possible interplay in the discussion and briefly mentioned it in the abstract.

We have not discussed more strongly the role of CNCC in the emergence of OFC in the revised manuscript, because we did not want to put even more emphasis on this matter. Numerous studies have proven the contribution of cranial neural crest tissue to the emergence of orofacial clefts. This fact is also pointed out in several review articles about orofacial clefts. In most cases, this knowledge was achieved by mouse models, because tissue-specific conditional knockouts are feasible (in contrast to genetic studies on patients), usually via deletion with the Wnt1-Cre driver. Funato et al. give an excellent (but quite old) overview of mouse models in which the neural crest-specific knockout of a gene leads to emergence of OFC and lists 17 genes for which this is the case (Funato et al, 2015). Moreover, several recent studies also report on the emergence of orofacial clefts upon neural crest-specific deletion (Forman et al, 2024; Li et al, 2025). These include genes responsible for DNA methylation (Ulschmid et al, 2024), and a study on subunits of chromatin remodeling complexes that are necessary for correct transcription of their target genes, which was conducted by our group (Gehlen-Breitbach et al, 2023).

Minor comments

__Point 1-10 __

The author should replace "Final proof" in the introduction with "further evidence supporting."

We apologize for the incorrect wording. Of course, it is highly questionable if there is such a thing as final proof in life sciences. We re-phrased the text according to the reviewer’s suggestion.

Point 1-11

Authors are inconsistent when referring to Figures- sometimes they capitalize (i.e. 1J) and other times they leave lower case (i.e. 1i). Needs to be consistent throughout. Figures are not numbered.

We apologize for the inconsistency. We corrected the references to figures. Moreover, we apologize for the missing figure numbers. We also corrected this and included figure numbers.

Point 1-12

In figures authors would sometimes list 21% O2 first then 0.5% O2 or vice versa. (i.e. Fig on page 21 panels I, J, K). Needs to be consistent.

We again apologize for being inconsistent. We corrected the inconsistency in Fig. 1D. Now, 21% O2 is presented before/above 0.5% O2.

Point 1-13

Figures on pages 28, 29, 30 panel J and page 31 panel F: there is no legend on what the scale/measurement is for the difference in expression level other than it ranges from -1 to +3.

We thank the reviewer for the hint. We are aware that from the heatmaps we used one cannot infer relative expression rates of different genes or similar. If we would have considered expression strength of single genes, many of the gene-specific differing expression rates under the different conditions would have been hard to detect, as presentation would have been dominated by the differences in expression rates between genes. We therefore plotted gene-wise scaled expression.

We included an explanation of the procedure in the materials and methods section.

Point 1-14

Will the authors please comment on the one normoxic sample in Figure 1I that did not cluster with the others? Did this meet the standards to merit exclusion as an outlier?

We regret that the default scale of our plot of the principal component analysis is a bit misleading. This is the case because x-axis accounts for 80.3% of variance and y-axis only accounts for 6.1%. Therefore, the sample that might seem as an outlier actually met our standards. Nevertheless, we decided to keep the default scaling as is, in order not to embellish the graph (Fig. 1M).

Point 1-15

The authors refer to DEG as deregulated genes; while not strictly incorrect, the more standard usage is "differentially expressed genes." Please address.

We apologize for the incorrect explanation of the acronym. Of course, this was corrected in the revised manuscript.

Significance

This work on neural crest cells and hypoxia are biologically and clinically significant.

We are deeply grateful to the reviewer for considering our manuscript significant for both biologists and clinicians. We are convinced that the additional data we gathered in the course of the revision has significantly increased the importance of our work. Therefore, we once again express our gratitude to the reviewer for the valuable suggestions.

Response to reviewer 2 comments

Major comments

Point 2-1

The conclusions drawn from the experimental data are carefully formulated for the most part. One of the main concerns is that the cells were subjected to extreme hypoxic conditions, while it may be more biologically relevant to include a condition representing more mild hypoxia (e.g. 10%).

Please refer to the response to point 1-2.

Point 2-2

One of the opening claims regarding severe hypoxia only mildly affecting cell proliferation is not shown clearly, since no mitotic markers have been analyzed (i.e. KI67 or PCNA staining or a simple EdU incorporation assay). Thus, the claim that they assessed cell proliferation is not very convincing, even though cell death was analyzed.

We appreciate the reviewer’s suggestion to include a more thorough analysis of proliferation rates. We followed the advice and performed immunofluorescent stainings against Ki67 (accounting for cells in proliferative state) and phospho-histone H3 (accounting for cells undergoing mitosis). We performed this assay at different time points of culture in order to address the question if cell density might influence proliferation rates (Figs. 1F-H). Neither for Ki67 nor for pHH3 a difference was detected between 21% and 0.5% O2.

We are convinced that these analyses strengthened our initial findings and provide strong evidence that hypoxia does not influence proliferation rates of CNCC.

Point 2-3

Additionally, cellular morphology of the cells could be assessed (brightfield images), since previous studies observed that hypoxia can be an inducive factor in cranial neural crest and driving EMT (Scully et al. 2016; Barriga et al. 2013).

We thank the reviewer’s hint and followed the advice. We analyzed cellular morphology by the parameters cell length, total number of pseudopodia, number of filopodia and number of lobopodia (Figs. S1C-F). As outlined in the results section, we did not detect a difference in these parameters between 21% and 0.5% O2.

We included the second reference mentioned by the reviewer (Barriga et al, 2013) additionally to Scully et al. 2016 that had already been cited.

Point 2-4

Furthermore, in the RNA seq analysis of chondrogenic fate biased cells the authors draw a conclusion based on the proximity of the samples on the PCA plot, which is not very convincing. More careful analysis of the bulk RNA seq data sets they have generated for key marker genes will be more convincing (for example, a heatmap with selected genes would be a helpful representation).

We apologize for the rash and inaccurate conclusion based on proximity on PCA plots. We are grateful to the reviewer for the suggestion to include heatmaps with selected marker genes. Following this advice, we generated heatmaps on our bulk RNA-seq data with the GO terms specific for each differentiation paradigm (Figs. S2F, S3F, S4F).

We are convinced that these maps are perfect additions to the heatmaps of the 200 top differentially-expressed genes that already had been included in the manuscript (Figs. 2K, 3J, 4J) and helped to strengthen our findings. For chondrocytes and smooth muscle cells, the new, GO-specific heatmaps perfectly recapitulated the phenomenon of hypoxia-attenuated induction. Interestingly, for osteoblasts, about half of the induced genes were hypoxia-attenuated, while the other half was induced stronger than under normoxia. This pointed to gene-specific mechanisms of hypoxia-dependent attenuation of transcription. Moreover, it shed light on a hypoxia-evoked complete dysregulation of transcriptional induction in osteoblasts, as nearly none of the genes was induced similar to normoxia.

__ __

Point 2-5

As mentioned above, a straight-forward and not time consuming experiment (given that it was assessed for a maximum of 72 hrs) would be to repeat the culture of NCCs and stain for mitotic markers, and quantify the number of positively stained cells over total cell numbers. Furthermore, it is not that demanding to add an experimental condition of less severe hypoxia in this assay.

We thank the reviewer for the suggestion and followed the advice (cf. point 2-2). The conducted experiments straightened our results, because the initially detected slight tendency to lower cell numbers at 0.5% O2 could thus be falsified: We did not detect any difference for Ki67 and pHH3 between 0.5% and 21% O2 at any analyzed time point (Figs. 1F-H). Moreover, percentages of dead or apoptotic cells at 0.5% O2 did not vary from 21% (Figs. 1I-L, S1B). As we could not detect any difference in proliferation between 21% and 0.5% O2, we skipped the analysis of proliferating cells at 2% O2.

Point 2-6

Without underestimating how time consuming this would be, a major lack of experimental validation of the key genes they identify as important across all conditions may be the limitation of the study (this would be the difference between correlation and a probable underlying mechanism). This can be circumvented by more extensive reference to in situ data sets from mouse or existing data sets of single cell and spatial transcriptomics. A suggested targeted knock-down (for example with siRNA, shRNA or CRISPR) to validate a few of the key genes revealed as important could take a few months, with an estimated cost up to 5,000 euros per targeted gene and replicate.

We thank the reviewer for the notion that targeted knockdowns are beyond the scope of our manuscript. We are deeply grateful for the reviewer’s constructive criticism and for the suggestion to analyze publicly available data sets in order to gather data depicting in vivo relevance of our identified central hypoxia-attenuated OFC risk genes Boc, Cdo1 and Actg2 (cf. point 1-4). We detected robust expression of Boc and Cdo1 during human craniofacial development (Fig. 7A) and we identified enhancers that are active in embryonic craniofacial mouse tissue (Fig. 7B). Moreover, we detected expression of both genes during murine craniofacial development in undifferentiated mesenchymal cells, osteoblasts, chondrocytes and smooth muscle cells by reanalysis of a scRNA-seq dataset (Figs. 7C-E, S6B). This data comprised scRNA-seq of mouse embryonic maxillary prominence at stages E11.5 and E14.5 (Sun et al, 2023).

Thus, we found evidence for the in vivo relevance of Boc and Cdo1 and could rule out a possible important role of Actg2, the third gene we had identified. We therefore are deeply grateful for the suggestion, as we think these data strongly emphasize the importance of our findings.

Point 2-7

On methods, replicates and statistics: The experimental methods and approach are described efficiently and seem reproducible. All biological and technical replicates are of a minimum of N=3 from independent experiments and statistical tests have been run in all cases.

We thank the reviewer for the appreciation of our methodology, descriptions and statistical analyses.

Minor points

Point 2-8

One of the key implications of NCCs in palate formation is interaction with orofacial epithelial cells, which the authors also mention. It may be interesting to check if any signaling pathways involved in this crosstalk are affected under hypoxic conditions in their existing data sets of bulk RNA SEQ. This can be done by using available algorithms such as CellChat (Jin et al. 2021; Jin, Plikus, and Nie 2023), which has been reported to work also in bulk RNA seq data analysis (according to GitHub). The authors could mine the literature for existing RNA sequencing data that include osteoblasts, chondrocytes and epithelial cells (Ozekin, O'Rourke, and Bates 2023; Piña et al. 2023).

We are very grateful to the reviewer for this suggestion. Moreover, we like to thank the reviewer for mentioning exemplary references. We followed the advice by the methodology lined out in results and materials and methods sections: we applied the CellChat algorithm on a scRNA-seq dataset (Pina et al, 2023; Sun et al., 2023) to identify pathways containing components that are hypoxia-attenuated (and associated with a risk for OFC) in our bulk RNA-seq dataset (Figs. 7F-I). We did not use the datasets the reviewer had suggested, because the data were not available for us or the file format was not well-suited for the analysis with CellChat. Importantly, the dataset from Sun et al. has the following advantages over the suggested references: the complete maxillary prominence was used (instead of palatal shelves only), and different time points were included. Thus, we were able to follow the expression of genes of interest at different developmental stages before the onset of differentiation and after (Figs. 7C-E and S6B). By our approach, we identified several OFC-related pathways that contain hypoxia-attenuated components such as BMP and FGF signaling and deposition of collagen and fibronectin (Figs. 7F-I). Importantly, the named pathways (and others) send outgoing communication patterns to epithelial cells. Therefore, hypoxia-attenuated gene induction in CNCC could influence epithelial cells via these pathways.

We believe that the use of the CellChat algorithm has brought a deeper understanding of how hypoxia can have indirect consequences on the important topic of epithelial cells and thus could also evoke OFC. We therefore once again like to express our gratitude to the reviewer.

Point 2-9

Additionally, another process that may be affected is EMT (epithelial-to-mesenchymal-transition) and is possible to assess by re-analysis of bulk RNA-seq data while focusing on key genes implicated in this process (i.e. E-cadherin, vimentin, EpCAM, Snail, Twist, PRRX1).

We thank the reviewer for the advice. We followed the advice and analyzed cellular morphology by the parameters cell length, total number of pseudopodia, number of filopodia and number of lobopodia (Figs. S1C-F) (cf. point 2-3). As we did not detect any differences between 21% and 0.5% O2, and because the cells we used for our analyses represent mesenchymal cells, i.e. cells that had already undergone EMT, we did not re-analyze our dataset with the focus on EMT.

Point 2-10

Lastly, when the authors report on the significantly up- or down-regulated genes, it may be interesting to categorize them by ligands, receptors, intracellular molecules and transcription factors (and use separate plots to visualize them). While a big focus of the manuscript are down-regulated genes, less emphasis was given in upregulated genes (other than the response to hypoxia gene module).

We thank the reviewer for the advice. Following this advice, we categorized genes according to Panther protein classes "intercellular signal molecule" (PC00207), "transmembrane signal receptor" (PC00197) and "gene-specific transcriptional regulator" (PC00264) and depicted the results with violin plots (Fig. S5B). We could not analyze intracellular molecules, because this protein class does not exist in the Panther database. We had not focused on the genes with stronger induction in hypoxic condition, because the number of genes was low in each differentiation paradigm (7 in chondrocytes, less than 30 in osteoblasts, none in smooth muscle cells) and the transcriptional changes were mostly not as drastic as for the attenuated genes. In order to achieve a broader overview of deregulated processes, we now included GO term analyses of genes downregulated during the differentiation regimes both at 21% and 0.5% O2 (Figs. S2D,E, S3D,E, S4D,E).

Point 2-11

The authors are referencing extensively and accurately existing studies in the field and the manuscript is exceptionally well-written, with only a few points of limited clarity or increased complexity. Such an example is when the authors refer to OFC risk genes, because it is not clearly stated how the referenced studies reached their conclusions (for example, are they mouse studies, do they involve mutants, are any of these studies based on GWAS on human cohorts). This matter would significantly improve the flow of the text and highlight the importance of the study and their findings.

We would like to thank the reviewer very much for the appreciation of our scientific writing. We apologize for not explaining exactly how our OFC risk gene lists had been curated. We included this information for both non-syndromic and other OFC risk genes at the respective sites in the results section. Moreover, we included the Human Phenotype Ontology terms that had been used in the search in the materials and methods section.

We thank the reviewer for this suggestion, as we agree that this information significantly highlights the importance of our findings.

Point 2-12

The figures could be redesigned to be more intuitive to interpret. For example, using violin plots and heatmaps, as discussed, and including references or re-analysis/re-use of existing spatial transcriptomics and in situs for marker genes.

In all cases where there is a comparison of gene expression levels, violin plots would be a better representation of up- and down-regulated genes (i.e. selected genes from Fig1K, comparison of gene expression between normoxic and hypoxic NCCs, Fig 2G when analyzing chondrogenesis and the respective analysis for osteoblasts and smooth muscle cells, as well as when comparing the three fate-biasing conditions to identify common genes that are misregulated).

We thank the reviewer for the advice and for the appreciation of the usage of heatmaps (Figs. 2K, 3J, 4J, 6F). Unfortunately, as the number of biological replicates is only three to four, the visualization of gene expression data from our bulk RNA-seq data with violin plots was not intuitive. We therefore retained the heatmaps rather than choosing bar graphs, because they are much clearer when presenting expression data of several to many genes. We included violin plots whenever possible due to high numbers of data points (Figs. S1C, S1D, S1E, S1F, S5B). Moreover, we added additional heatmaps to depict transcriptional changes of genes associated with GO terms with the various differentiation regimes (Figs. S2F, S3F, S4F). Unfortunately, we did not detect the three central hypoxia-attenuated genes in spatial transcriptomics data on craniofacial development. But we used scRNA-seq data of different stages of orofacial mouse tissue where we could identify expression of Boc and Cdo1 (cf. points 1-4 and 2-6). These data helped, together with other in vivo data to gain evidence for the in vivo function of Boc and Cdo1 during CNCC differentiation and helped to dismiss Actg2 as another central player.

Significance

Several pieces of evidence have pointed to hypoxia as an environmental factor contributing to congenital orofacial clefts, ranging from studies in mouse to observations in human. The authors are doing an excellent job in putting this information together and the question they are trying to answer is of high importance, given the prevalence of such congenital syndromes.

We are deeply grateful to the reviewer for the appreciation of our work and for classifying our research topic as highly important.

In terms of the methods and model employed, there are some limitations, related to the choice of a mouse cell line over one from human, the severe hypoxia induced (over a more mild), and the conditions of directed differentiation not allowing for simultaneous examination of more complex lineage transitions. The methods as a whole are not that up-to-date, given the single cell and multiplexed transcriptomic advances the last couple of decades, advanced bioinformatics that could be used in combination with in vitro lineage tracing methods.

We thank the reviewer for the honest evaluation of our methods, especially for the constructive suggestions that were given to address our hypotheses with more up-to-date methods and at milder hypoxic conditions. As outlined above, we followed the advice and re-analyzed existing scRNA-seq datasets (cf. points 2-6 and 2-8) and checked our central hypotheses at milder hypoxic conditions (cf. response to point 1-3).

We are deeply convinced that both significantly increased the biological relevance of our results, because we thus (1) gathered evidence for the in vivo function of Boc and Cdo1 and (2) were able to show that the phenomenon of hypoxia-attenuated gene induction still holds true at biologically relevant hypoxic conditions.

The audience this work will reach are neural crest experts, developmental biologists, and potentially clinical doctors. The general public outreach of such a paper is also diverse, as more focus and visibility is required for the individuals affected by those syndromes and their families.

We thank the reviewer for the judgement that our manuscript will not only reach neural crest experts, but also developmental biologists in general and potentially also clinicians. We are very much pleased that the reviewer shares our opinion that affected individuals should be more in the focus of public attention. We like to express our gratitude for the judgement that our manuscript might help to increase focus and visibility for them.

References

Barriga EH, Maxwell PH, Reyes AE, Mayor R (2013) The hypoxia factor Hif-1α controls neural crest chemotaxis and epithelial to mesenchymal transition. The Journal of cell biology 201: 759-776, 10.1083/jcb.201212100.

Forman TE, Sajek MP, Larson ED, Mukherjee N, Fantauzzo KA (2024) PDGFRα signaling regulates Srsf3 transcript binding to affect PI3K signaling and endosomal trafficking. Elife 13, 10.7554/eLife.98531.

Funato N, Nakamura M, Yanagisawa H (2015) Molecular basis of cleft palates in mice. World journal of biological chemistry 6: 121-138, 10.4331/wjbc.v6.i3.121.

Gehlen-Breitbach S, Schmid T, Fröb F, Rodrian G, Weider M, Wegner M, Gölz L (2023) The Tip60/Ep400 chromatin remodeling complex impacts basic cellular functions in cranial neural crest-derived tissue during early orofacial development. International Journal of Oral Science 15: 16, 10.1038/s41368-023-00222-7.

Hansen JM, Jones DP, Harris C (2020) The Redox Theory of Development. Antioxid Redox Signal 32: 715-740, 10.1089/ars.2019.7976.

Li D, Tian Y, Vona B, Yu X, Lin J, Ma L, Lou S, Li X, Zhu G, Wang Y et al (2025) A TAF11 variant contributes to non-syndromic cleft lip only through modulating neural crest cell migration. Hum Mol Genet 34: 392-401, 10.1093/hmg/ddae188.

Ng KYB, Mingels R, Morgan H, Macklon N, Cheong Y (2017) In vivo oxygen, temperature and pH dynamics in the female reproductive tract and their importance in human conception: a systematic review. Human Reproduction Update 24: 15-34, 10.1093/humupd/dmx028.

Pina JO, Raju R, Roth DM, Winchester EW, Chattaraj P, Kidwai F, Faucz FR, Iben J, Mitra A, Campbell K et al (2023) Multimodal spatiotemporal transcriptomic resolution of embryonic palate osteogenesis. Nature communications 14: 5687, 10.1038/s41467-023-41349-9.

Sun J, Lin Y, Ha N, Zhang J, Wang W, Wang X, Bian Q (2023) Single-cell RNA-Seq reveals transcriptional regulatory networks directing the development of mouse maxillary prominence. J Genet Genomics 50: 676-687, 10.1016/j.jgg.2023.02.008.

Ulschmid CM, Sun MR, Jabbarpour CR, Steward AC, Rivera-González KS, Cao J, Martin AA, Barnes M, Wicklund L, Madrid A et al (2024) Disruption of DNA methylation-mediated cranial neural crest proliferation and differentiation causes orofacial clefts in mice. Proc Natl Acad Sci U S A 121: e2317668121, 10.1073/pnas.2317668121.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Schmidt and colleagues are addressing the effects of severe hypoxia on proliferation and differentiation potential of (mouse) cranial neural crest, using a neural crest cell line subjected to hypoxic conditions, assessed by transcriptomics analysis (quantitative reverse transcription PCR, bulk RNA sequencing and bioinformatics). They are reporting a mild effect of cell proliferation and an extensive inhibition of differentiation towards osteoblasts, chondrocytes and smooth muscle cells. They reveal affected biological processes shared between the three fate biasing conditions related to cytoskeleton organization and amino acid metabolism. Lastly, among affected genes upon hypoxic conditions in vitro, they authors identified risk genes linked to non-syndromic (non-genetic) orofacial clefts exclusively downregulated in osteoblasts and smooth muscle cells, namely Fgfr2, Gstt1 and Tbxa2. Similarly, hypoxia-driven downregulation of genes implicated in syndromic orofacial clefts was observed in all three chondrocyte, osteoblast and smooth muscle differentiation scenarios. Lastly, STRING analysis of downregulated genes cross-validated their findings related to affected differentiation.

Major comments:

The conclusions drawn from the experimental data are carefully formulated for the most part. One of the main concerns is that the cells were subjected to extreme hypoxic conditions, while it may be more biologically relevant to include a condition representing more mild hypoxia (e.g. 10%). One of the opening claims regarding severe hypoxia only mildly affecting cell proliferation is not shown clearly, since no mitotic markers have been analyzed (i.e. KI67 or PCNA staining or a simple EdU incorporation assay). Thus, the claim that they assessed cell proliferation is not very convincing, even though cell death was analyzed. Additionally, cellular morphology of the cells could be assessed (brightfield images), since previous studies observed that hypoxia can be an inducive factor in cranial neural crest and driving EMT (Scully et al. 2016; Barriga et al. 2013).

Furthermore, in the RNA seq analysis of chondrogenic fate biased cells the authors draw a conclusion based on the proximity of the samples on the PCA plot, which is not very convincing. More careful analysis of the bulk RNA seq data sets they have generated for key marker genes will be more convincing (for example, a heatmap with selected genes would be a helpful representation). As mentioned above, a straight-forward and not time consuming experiment (given that it was assessed for a maximum of 72 hrs) would be to repeat the culture of NCCs and stain for mitotic markers, and quantify the number of positively stained cells over total cell numbers. Furthermore, it is not that demanding to add an experimental condition of less severe hypoxia in this assay. Without underestimating how time consuming this would be, a major lack of experimental validation of the key genes they identify as important across all conditions may be the limitation of the study (this would be the difference between correlation and a probable underlying mechanism). This can be circumvented by more extensive reference to in situ data sets from mouse or existing data sets of single cell and spatial transcriptomicsA suggested targeted knock-down (for example with siRNA, shRNA or CRISPR) to validate a few of the key genes revealed as important could take a few months, with an estimated cost up to 5,000 euros per targeted gene and replicate. On methods, replicates and statistics: The experimental methods and approach are described efficiently and seem reproducible.All biological and technical replicates are of a minimum of N=3 from independent experiments and statistical tests have been run in all cases.

Minor comments:

One of the key implications of NCCs in palate formation is interaction with orofacial epithelial cells, which the authors also mention. It may be interesting to check if any signaling pathways involved in this crosstalk are affected under hypoxic conditions in their existing data sets of bulk RNA SEQ. This can be done by using available algorithms such as CellChat (Jin et al. 2021; Jin, Plikus, and Nie 2023), which has been reported to work also in bulk RNA seq data analysis (according to GitHub). The authors could mine the literature for existing RNA sequencing data that include osteoblasts, chondrocytes and epithelial cells (Ozekin, O'Rourke, and Bates 2023; Piña et al. 2023).

Additionally, another process that may be affected is EMT (epithelial-to-mesenchymal-transition) and is possible to assess by re-analysis of bulk RNA-seq data while focusing on key genes implicated in this process (i.e. E-cadherin, vimentin, EpCAM, Snail, Twist, PRRX1). Lastly, when the authors report on the significantly up- or down-regulated genes, it may be interesting to categorize them by ligands, receptors, intracellular molecules and transcription factors (and use separate plots to visualize them). While a big focus of the manuscript are down-regulated genes, less emphasis was given in upregulated genes (other than the response to hypoxia gene module).

The authors are referencing extensively and accurately existing studies in the field and the manuscript is exceptionally well-written, with only a few points of limited clarity or increased complexity. Such an example is when the authors refer to OFC risk genes, because it is not clearly stated how the referenced studies reached their conclusions (for example, are they mouse studies, do they involve mutants, are any of these studies based on GWAS on human cohorts). This matter would significantly improve the flow of the text and highlight the importance of the study and their findings. The figures could be redesigned to be more intuitive to interpret. For example, using violin plots and heatmaps, as discussed, and including references or re-analysis/re-use of existing spatial transcriptomics and in situs for marker genes.

In all cases where there is a comparison of gene expression levels, violin plots would be a better representation of up- and down-regulated genes (i.e. selected genes from Fig1K, comparison of gene expression between normoxic and hypoxic NCCs, Fig 2G when analyzing chondrogenesis and the respective analysis for osteoblasts and smooth muscle cells, as well as when comparing the three fate-biasing conditions to identify common genes that are misregulated).

References:

Barriga, Elias H., Patrick H. Maxwell, Ariel E. Reyes, and Roberto Mayor. 2013. "The Hypoxia Factor Hif-1α Controls Neural Crest Chemotaxis and Epithelial to Mesenchymal Transition." The Journal of Cell Biology 201 (5): 759-76. https://doi.org/10.1083/jcb.201212100.

Jin, Suoqin, Christian F. Guerrero-Juarez, Lihua Zhang, Ivan Chang, Raul Ramos, Chen-Hsiang Kuan, Peggy Myung, Maksim V. Plikus, and Qing Nie. 2021. "Inference and Analysis of Cell-Cell Communication Using CellChat." Nature Communications 12 (1). https://doi.org/10.1038/s41467-021-21246-9.

Jin, Suoqin, Maksim V. Plikus, and Qing Nie. 2023. "CellChat for Systematic Analysis of Cell-Cell Communication from Single-Cell and Spatially Resolved Transcriptomics." bioRxiv. https://doi.org/10.1101/2023.11.05.565674.

Ozekin, Yunus H., Rebecca O'Rourke, and Emily Anne Bates. 2023. "Single Cell Sequencing of the Mouse Anterior Palate Reveals Mesenchymal Heterogeneity." Developmental Dynamics : An Official Publication of the American Association of Anatomists 252 (6): 713-27. https://doi.org/10.1002/dvdy.573.

Piña, Jeremie Oliver, Resmi Raju, Daniela M. Roth, Emma Wentworth Winchester, Parna Chattaraj, Fahad Kidwai, Fabio R. Faucz, et al. 2023. "Multimodal Spatiotemporal Transcriptomic Resolution of Embryonic Palate Osteogenesis." Nature Communications 14 (September):5687. https://doi.org/10.1038/s41467-023-41349-9.

Scully, Deirdre, Eleanor Keane, Emily Batt, Priyadarssini Karunakaran, Debra F. Higgins, and Nobue Itasaki. 2016. "Hypoxia Promotes Production of Neural Crest Cells in the Embryonic Head." Development 143 (10): 1742-52. https://doi.org/10.1242/dev.131912.

Significance

Several pieces of evidence have pointed to hypoxia as an environmental factor contributing to congenital orofacial clefts, ranging from studies in mouse to observations in human. The authors are doing an excellent job in putting this information together and the question they are trying to answer is of high importance, given the prevalence of such congenital syndromes. In terms of the methods and model employed, there are some limitations, related to the choice of a mouse cell line over one from human, the severe hypoxia induced (over a more mild), and the conditions of directed differentiation not allowing for simultaneous examination of more complex lineage transitions. The methods as a whole are not that up-to-date, given the single cell and multiplexed transcriptomic advances the last couple of decades, advanced bioinformatics that could be used in combination with in vitro lineage tracing methods.

The audience this work will reach are neural crest experts, developmental biologists, and potentially clinical doctors. The general public outreach of such a paper is also diverse, as more focus and visibility is required for the individuals affected by those syndromes and their families.

Reviewer's expertise: mouse neural crest lineage and multipotency, lineage tracing, single cell transcriptomics, NGS, immunofluorescence, molecular methods (RNA, DNA based). Limited expertise with in vitro studies.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Title: Hypoxia impedes differentiation of cranial neural crest cells into derivatives relevant for craniofacial development

Synopsis: Cleft lip w/ or w/o cleft palate is the second-most common birth defect worldwide. Defects are often traceable to cranial neural crest cells through genetics or environmental factors. Schmid and coauthors focus on the environmental factor of hypoxia and investigate the effects of hypoxic conditions on the ability of CNCCs to differentiate and migrate. They performed RNA-seq analysis with qRT-PCR validation for specific markers and show that hypoxia appears to repress differentiation without markedly affecting proliferation. Hypoxic conditions did not demonstrated significant perturbations in cell proliferation; however, chondrocyte, osteoblast, and smooth muscle differentiation was significantly reduced for cell lines cultured under hypoxia. Bulk RNA-seq and PCA revealed dysregulation of genes implicated in cytoskeletal integrity (such as actin γ-2), neural crest cell migration (hedgehog co-receptor brother of CDO) and amino acid metabolism (cysteine dioxygenase), which Schmid and colleagues termed OFC risk genes.

Major comments

• The authors performed qRT-PCR validation for markers of differentiation and hypoxia, with a major absence of VEGF and HIF1a. The paper would be strengthened by mention of these factors, especially by qRT-PCR or Western blot.
• Please provide justification of selection 0.5% as their hypoxic condition or perhaps repeat experiments in a less extreme environment to see if their conclusions still hold true.
• standard immunohistochemistry or histology of differentiated cells would strengthen the authors' claims of reduced differentiation under hypoxic conditions, e.g., Alcian blue, alk-phos or Alizarin red, and smooth muscle actin or other indicator.
• The authors identify a few genes that appear down-regulated in all three differentiation conditions. If it is within the scope of the study, it would strengthen the claim of these genes' function to show the effect of knock-down or knock-out for validation.
• Another major critique lies in the initial claim that proliferation of O9-1 cells is not significantly impacted by hypoxia. In figures 1E-H, photograms of the cells cultured 24 -72 hours and quantifications of live vs dead cells are shown as evidence for this argument. However, the increased density of cells in normoxic conditions may be a confounding variable in this assay. It would be interesting for the researchers to assess the percent of dead vs alive cells between normoxic and hypoxic conditions when the plates reach equivalent densities.
• At end of Fig 1 section authors attempt to tie phenotypes observed in a cell line in vitro to the complex biological processes. They are not comparable and in vivo models would be better suited for these types of comparisons.
• Fig 2: if qRT-PCR did not show statistically different results between experimental and control groups why move on to bulk RNA seq?
• Fig 5: hypoxia this intense is going to affect broad range of biological processes and genes. Finding a few genes that are affected in extreme hypoxia that are also risk genes is highly unlikely. How can the authors be assured that these overlaps are actually significant and not just by chance?
• Would appreciate discussion on how examination of neural crest is relevant for OFC, as most animal models of OFC demonstrate the pathogenesis in embryonic epithelium or periderm, not in the neural crest. Defects in neural crest are associated with other congenital craniofacial anomalies such as craniosynostosis or complex (Tessier) clefts, not the typical orofacial cleft. Please revise rationale of study, interpretation of data and Discussion to specifically state how neural crest cells are involved in the pathogenesis of orofacial cleft.

Minor comments

• The author should replace "Final proof" in the introduction with "further evidence supporting."
• Authors are inconsistent when referring to Figures- sometimes they capitalize (i.e. 1J) and other times they leave lower case (i.e. 1i). Needs to be consistent throughout. Figures are not numbered
• In figures authors would sometimes list 21% O2 first then 0.5% O2 or vice versa. (i.e. Fig on page 21 panels I, J, K). Needs to be consistent.
• Figures on pages 28, 29, 30 panel J and page 31 panel F: there is no legend on what the scale/measurement is for the difference in expression level other than it ranges from -1 to +3.
• Will the authors please comment on the one normoxic sample in Figure 1I that did not cluster with the others? Did this meet the standards to merit exclusion as an outlier?
• The authors refer to DEG as deregulated genes; while not strictly incorrect, the more standard usage is "differentially expressed genes." Please address.

Significance

This work on neural crest cells and hypoxia are biologically and clinically significant.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements

We thank the editor for handling our manuscript and the reviewers for their constructive critiques. We are deeply convinced that the reviewers’ suggestions have substantially raised the quality and possible impact of our manuscript. We also like to thank the reviewers for their judgements that the subject of our manuscript is biologically and clinically significant and of high importance, and that our manuscript might help to increase focus and visibility for affected individuals.

New text passages in the manuscript are colored in red. Below is a point-by-point response to the reviewers’ comments.

2. Point-by-point description of the revisions

Response to reviewer 1 comments

Major comments

---

Point 1-1

The authors performed qRT-PCR validation for markers of differentiation and hypoxia, with a major absence of VEGF and HIF1a. The paper would be strengthened by mention of these factors, especially by qRT-PCR or Western blot.

We thank the reviewer for the suggestion to include the bona fide hypoxia markers Vegfa and Hif1-alpha. We followed the suggestion and performed qRT-PCR on Vegfa transcripts at each tested condition (Figs. 1A,2A,3A,4A,5A,5D,5I,5N). As Hif1α is rather regulated on protein than on transcript level, we followed the advice to perform Western blots. We analyzed Hif1α protein levels on proliferating cells and quantified by normalization to actin (Figs. 1B,C and 5 B,C).

Point 1-2

Please provide justification of selection 0.5% as their hypoxic condition or perhaps repeat experiments in a less extreme environment to see if their conclusions still hold true.

We admit that our approach to use 0.5% hypoxia was a drastic challenge for the cells. It should be noted, however, that physiologic oxygen levels during pregnancy at times drop to lower than 1% (Hansen et al, 2020; Ng et al, 2017). In the first place, we had used oxygen levels lower than this, because we had wanted to ensure that we can detect responses by bulk RNA-seq with a limited number of samples. As we had many conditions to compare, we did not want to use more than 3-4 samples per condition. The fact that the cells showed normal proliferation underscores the fact that 0.5% O2 per se was not so low that it would be overly stressful to the cells.

Nevertheless, we are very grateful to the reviewer for the suggestion to include a milder hypoxic condition. We chose 2% O2, because this equals the physiological oxygen concentration shortly before the onset of cranial neural crest cell (CNCC) differentiation. We could recapitulate the phenomenon of impaired differentiation to chondrocytes, osteoblasts and smooth muscle cells at these mild hypoxic conditions, as shown by qRT-PCR and immunofluorescence of typical markers (Figs. 5D-R). Moreover, the differentiation-specific induction of the two central hypoxia-attenuated risk genes associated with orofacial clefts that we had identified by our bioinformatic analyses at 0.5% O2 (Boc and Cdo1), was still observable at 2% O2 (Figs. EV6C,D). Interestingly, in some rare cases, the attenuation of induction was lost or not as drastic as in 0.5% O2.

We are convinced that the experiments at 2% O2 strongly increased the relevance of our manuscript, because we thus detected that oxygen levels prevailing shortly before the onset of CNCC differentiation still can influence their differentiation. This leads to the conclusion that only slight decreases of intra-uterine oxygen levels indeed might interfere with correct differentiation of CNCC.

Point 1-3

Standard immunohistochemistry or histology of differentiated cells would strengthen the authors' claims of reduced differentiation under hypoxic conditions, e.g., Alcian blue, alk-phos or Alizarin red, and smooth muscle actin or other indicator.

We are grateful to the reviewer for the suggestion to include stainings of cells, as these stainings visualized the drastic effects of hypoxia on the cells. We performed immunofluorescent stainings against at least one marker protein for each differentiation paradigm. At 0.5% O2, each protein signals were nearly completely absent and cell morphology was disrupted (Figs. 2E,F, 3E, 4E). At 2% O2, we detected some more protein deposition than at 0.5%. Importantly, cells had retained their normal shape at mild hypoxia (Figs. 5H,M,R, EV5A).

Point 1-4

The authors identify a few genes that appear down-regulated in all three differentiation conditions. If it is within the scope of the study, it would strengthen the claim of these genes' function to show the effect of knock-down or knock-out for validation.

We thank the reviewer for the suggestion of gene knock-down or knock-out in order to prove functional relevance of our findings. As this would have been too much effort and beyond the scope of our study, we rather followed the suggestion of reviewer 2 (cf. points 2-6, and 2-8) that headed to the same direction: we mined publicly available sequence data on orofacial development for gene expression or marks of active enhancers. We found robust expression of the two central hypoxia-attenuated OFC risk genes Boc and Cdo1 during human craniofacial development (Fig. 7A) and we identified enhancers that are active in embryonic craniofacial mouse tissue (Fig. 7B). Moreover, we detected expression of both genes during murine craniofacial development in undifferentiated mesenchymal cells, osteoblasts, chondrocytes and smooth muscle cells with the help of a single cell RNA-seq dataset (Figs. 7C-E, EV6B).

Thus, we found evidence for the in vivo relevance of Boc and Cdo1 and could rule out a possible important role of Actg2, the third gene we had identified. We therefore are grateful for the suggestion to circumvent gene knockouts by reviewer 2, as we think these data strongly emphasized the importance of our findings.

Point 1-5

Another major critique lies in the initial claim that proliferation of O9-1 cells is not significantly impacted by hypoxia. In figures 1E-H, photograms of the cells cultured 24 -72 hours and quantifications of live vs dead cells are shown as evidence for this argument. However, the increased density of cells in normoxic conditions may be a confounding variable in this assay. It would be interesting for the researchers to assess the percent of dead vs alive cells between normoxic and hypoxic conditions when the plates reach equivalent densities.

We apologize for the use of image sections from photographs with different cell densities. Of course, as demonstrated by our quantification, cell densities between 0.5% and 21% O2 in total were equal (cf. Figs. 1D,E). We therefore replaced the formerly used sections with new image sections with equal cell numbers.

We thank the reviewer for the suggestion to examine if cell numbers influence cell death rates. We followed this advice by several approaches: first, we seeded cells at different densities, incubated them for 72 h (the same time span where a minimal difference had been detected) and performed live/dead stainings (Fig. EV1B). The seeding density did not affect percentages of dead cells and the values were in the same range as in our initial experiment (Fig. 1J). Moreover, we performed TUNEL stainings of apoptotic cells at different time points to have an additional readout of cell death (Figs. 1K,L). As expected, the percentages of TUNEL-positive cells were identical between hypoxic and normoxic cells at all analyzed time points.

We therefore concluded that hypoxia does not influence the rate of cell death of proliferating CNCC and accordingly specified our wording in the results section.

Point 1-6

At end of Fig 1 section authors attempt to tie phenotypes observed in a cell line in vitro to the complex biological processes. They are not comparable and in vivo models would be better suited for these types of comparisons.

We apologize for the overconfident wording in our manuscript. Of course, our in vitro experiments cannot fully simulate the complex developmental processes taking place in vivo. We therefore changed the text to a more careful formulation. Moreover, we kept the wording in the discussion section that we cannot exclude that in the in vivo situation proliferation of CNCC is also affected by low oxygen levels because nutrients might not be available in such excess as they are in cell culture.

---

Point 1-7

Fig 2: if qRT-PCR did not show statistically different results between experimental and control groups why move on to bulk RNA seq?

We apologize that the sentence about statistical significance was misleading. What we wanted to express is that there was only a little difference (if any at all) between differentiated cells at 0.5% O2 and proliferating cells at 0.5% O2 or 21% O2. For the sake of clarity and readability, we deleted this misleading sentence.

Point 1-8

Fig 5: hypoxia this intense is going to affect broad range of biological processes and genes. Finding a few genes that are affected in extreme hypoxia that are also risk genes is highly unlikely. How can the authors be assured that these overlaps are actually significant and not just by chance?

We thank the reviewer for the suggestion to test for statistical significance. We tested significance of the overlap of respective gene sets (nsOFC vs. hyp-a; OFC vs. hyp-a) by Fisher’s exact test. We included Venn diagrams depicting the overlap and present the exact p-values (Figs. EV5C,D). In each case where overlap of genes occurred, p-values indicated significance.

Point 1-9

Would appreciate discussion on how examination of neural crest is relevant for OFC, as most animal models of OFC demonstrate the pathogenesis in embryonic epithelium or periderm, not in the neural crest. Defects in neural crest are associated with other congenital craniofacial anomalies such as craniosynostosis or complex (Tessier) clefts, not the typical orofacial cleft. Please revise rationale of study, interpretation of data and Discussion to specifically state how neural crest cells are involved in the pathogenesis of orofacial cleft.

We apologize for not pointing out enough the role of epithelial cells in the emergence of orofacial clefts. We revised our introduction, results and discussion sections in this regard and emphasized the role of epithelial cells. Importantly, we addressed the possible influence of the results gained in CNCC on epithelial cells by analyzing scRNA-seq data with the algorithm CellChat, as suggested by reviewer 2 (cf. point 2-8). We detected several cell communication pathways from CNCC to epithelial cells which contain components that are misexpressed upon hypoxia in our dataset (Figs. 7F-I). Therefore, during hypoxia, these pathways might influence epithelial cells and therefore indirectly cause orofacial clefts. We outlined this possible interplay in the discussion and briefly mentioned it in the abstract.

We have not discussed more strongly the role of CNCC in the emergence of OFC in the revised manuscript, because we did not want to put even more emphasis on this matter. Numerous studies have proven the contribution of cranial neural crest tissue to the emergence of orofacial clefts. This fact is also pointed out in several review articles about orofacial clefts. In most cases, this knowledge was achieved by mouse models, because tissue-specific conditional knockouts are feasible (in contrast to genetic studies on patients), usually via deletion with the Wnt1-Cre driver. Funato et al. give an excellent (but quite old) overview of mouse models in which the neural crest-specific knockout of a gene leads to emergence of OFC and lists 17 genes for which this is the case (Funato et al, 2015). Moreover, several recent studies also report on the emergence of orofacial clefts upon neural crest-specific deletion (Forman et al, 2024; Li et al, 2025). These include genes responsible for DNA methylation (Ulschmid et al, 2024), and a study on subunits of chromatin remodeling complexes that are necessary for correct transcription of their target genes, which was conducted by our group (Gehlen-Breitbach et al, 2023).

Minor comments

__Point 1-10 __

The author should replace "Final proof" in the introduction with "further evidence supporting."

We apologize for the incorrect wording. Of course, it is highly questionable if there is such a thing as final proof in life sciences. We re-phrased the text according to the reviewer’s suggestion.

Point 1-11

Authors are inconsistent when referring to Figures- sometimes they capitalize (i.e. 1J) and other times they leave lower case (i.e. 1i). Needs to be consistent throughout. Figures are not numbered.

We apologize for the inconsistency. We corrected the references to figures. Moreover, we apologize for the missing figure numbers. We also corrected this and included figure numbers.

Point 1-12

In figures authors would sometimes list 21% O2 first then 0.5% O2 or vice versa. (i.e. Fig on page 21 panels I, J, K). Needs to be consistent.

We again apologize for being inconsistent. We corrected the inconsistency in Fig. 1D. Now, 21% O2 is presented before/above 0.5% O2.

Point 1-13

Figures on pages 28, 29, 30 panel J and page 31 panel F: there is no legend on what the scale/measurement is for the difference in expression level other than it ranges from -1 to +3.

We thank the reviewer for the hint. We are aware that from the heatmaps we used one cannot infer relative expression rates of different genes or similar. If we would have considered expression strength of single genes, many of the gene-specific differing expression rates under the different conditions would have been hard to detect, as presentation would have been dominated by the differences in expression rates between genes. We therefore plotted gene-wise scaled expression.

We included an explanation of the procedure in the materials and methods section.

Point 1-14

Will the authors please comment on the one normoxic sample in Figure 1I that did not cluster with the others? Did this meet the standards to merit exclusion as an outlier?

We regret that the default scale of our plot of the principal component analysis is a bit misleading. This is the case because x-axis accounts for 80.3% of variance and y-axis only accounts for 6.1%. Therefore, the sample that might seem as an outlier actually met our standards. Nevertheless, we decided to keep the default scaling as is, in order not to embellish the graph (Fig. 1M).

Point 1-15

The authors refer to DEG as deregulated genes; while not strictly incorrect, the more standard usage is "differentially expressed genes." Please address.

We apologize for the incorrect explanation of the acronym. Of course, this was corrected in the revised manuscript.

Significance

This work on neural crest cells and hypoxia are biologically and clinically significant.

We are deeply grateful to the reviewer for considering our manuscript significant for both biologists and clinicians. We are convinced that the additional data we gathered in the course of the revision has significantly increased the importance of our work. Therefore, we once again express our gratitude to the reviewer for the valuable suggestions.

Response to reviewer 2 comments

Major comments

---

Point 2-1

The conclusions drawn from the experimental data are carefully formulated for the most part. One of the main concerns is that the cells were subjected to extreme hypoxic conditions, while it may be more biologically relevant to include a condition representing more mild hypoxia (e.g. 10%).

Please refer to the response to point 1-2.

Point 2-2

One of the opening claims regarding severe hypoxia only mildly affecting cell proliferation is not shown clearly, since no mitotic markers have been analyzed (i.e. KI67 or PCNA staining or a simple EdU incorporation assay). Thus, the claim that they assessed cell proliferation is not very convincing, even though cell death was analyzed.

We appreciate the reviewer’s suggestion to include a more thorough analysis of proliferation rates. We followed the advice and performed immunofluorescent stainings against Ki67 (accounting for cells in proliferative state) and phospho-histone H3 (accounting for cells undergoing mitosis). We performed this assay at different time points of culture in order to address the question if cell density might influence proliferation rates (Figs. 1F-H). Neither for Ki67 nor for pHH3 a difference was detected between 21% and 0.5% O2.

We are convinced that these analyses strengthened our initial findings and provide strong evidence that hypoxia does not influence proliferation rates of CNCC.

Point 2-3

Additionally, cellular morphology of the cells could be assessed (brightfield images), since previous studies observed that hypoxia can be an inducive factor in cranial neural crest and driving EMT (Scully et al. 2016; Barriga et al. 2013).

---

We thank the reviewer’s hint and followed the advice. We analyzed cellular morphology by the parameters cell length, total number of pseudopodia, number of filopodia and number of lobopodia (Figs. EV1C-F). As outlined in the results section, we did not detect a difference in these parameters between 21% and 0.5% O2.

We included the second reference mentioned by the reviewer (Barriga et al, 2013) additionally to Scully et al. 2016 that had already been cited.

Point 2-4

Furthermore, in the RNA seq analysis of chondrogenic fate biased cells the authors draw a conclusion based on the proximity of the samples on the PCA plot, which is not very convincing. More careful analysis of the bulk RNA seq data sets they have generated for key marker genes will be more convincing (for example, a heatmap with selected genes would be a helpful representation).

We apologize for the rash and inaccurate conclusion based on proximity on PCA plots. We are grateful to the reviewer for the suggestion to include heatmaps with selected marker genes. Following this advice, we generated heatmaps on our bulk RNA-seq data with the GO terms specific for each differentiation paradigm (Figs. EV2F, EV3F, EV4F).

We are convinced that these maps are perfect additions to the heatmaps of the 200 top differentially-expressed genes that already had been included in the manuscript (Figs. 2K, 3J, 4J) and helped to strengthen our findings. For chondrocytes and smooth muscle cells, the new, GO-specific heatmaps perfectly recapitulated the phenomenon of hypoxia-attenuated induction. Interestingly, for osteoblasts, about half of the induced genes were hypoxia-attenuated, while the other half was induced stronger than under normoxia. This pointed to gene-specific mechanisms of hypoxia-dependent attenuation of transcription. Moreover, it shed light on a hypoxia-evoked complete dysregulation of transcriptional induction in osteoblasts, as nearly none of the genes was induced similar to normoxia.

__ __

---

Point 2-5

As mentioned above, a straight-forward and not time consuming experiment (given that it was assessed for a maximum of 72 hrs) would be to repeat the culture of NCCs and stain for mitotic markers, and quantify the number of positively stained cells over total cell numbers. Furthermore, it is not that demanding to add an experimental condition of less severe hypoxia in this assay.

We thank the reviewer for the suggestion and followed the advice (cf. point 2-2). The conducted experiments straightened our results, because the initially detected slight tendency to lower cell numbers at 0.5% O2 could thus be falsified: We did not detect any difference for Ki67 and pHH3 between 0.5% and 21% O2 at any analyzed time point (Figs. 1F-H). Moreover, percentages of dead or apoptotic cells at 0.5% O2 did not vary from 21% (Figs. 1I-L, EV1B). As we could not detect any difference in proliferation between 21% and 0.5% O2, we skipped the analysis of proliferating cells at 2% O2.

Point 2-6

Without underestimating how time consuming this would be, a major lack of experimental validation of the key genes they identify as important across all conditions may be the limitation of the study (this would be the difference between correlation and a probable underlying mechanism). This can be circumvented by more extensive reference to in situ data sets from mouse or existing data sets of single cell and spatial transcriptomics. A suggested targeted knock-down (for example with siRNA, shRNA or CRISPR) to validate a few of the key genes revealed as important could take a few months, with an estimated cost up to 5,000 euros per targeted gene and replicate.

We thank the reviewer for the notion that targeted knockdowns are beyond the scope of our manuscript. We are deeply grateful for the reviewer’s constructive criticism and for the suggestion to analyze publicly available data sets in order to gather data depicting in vivo relevance of our identified central hypoxia-attenuated OFC risk genes Boc, Cdo1 and Actg2 (cf. point 1-4). We detected robust expression of Boc and Cdo1 during human craniofacial development (Fig. 7A) and we identified enhancers that are active in embryonic craniofacial mouse tissue (Fig. 7B). Moreover, we detected expression of both genes during murine craniofacial development in undifferentiated mesenchymal cells, osteoblasts, chondrocytes and smooth muscle cells by reanalysis of a scRNA-seq dataset (Figs. 7C-E, EV6B). This data comprised scRNA-seq of mouse embryonic maxillary prominence at stages E11.5 and E14.5 (Sun et al, 2023).

Thus, we found evidence for the in vivo relevance of Boc and Cdo1 and could rule out a possible important role of Actg2, the third gene we had identified. We therefore are deeply grateful for the suggestion, as we think these data strongly emphasize the importance of our findings.

Point 2-7

On methods, replicates and statistics: The experimental methods and approach are described efficiently and seem reproducible. All biological and technical replicates are of a minimum of N=3 from independent experiments and statistical tests have been run in all cases.

---

We thank the reviewer for the appreciation of our methodology, descriptions and statistical analyses.

Minor points

Point 2-8

One of the key implications of NCCs in palate formation is interaction with orofacial epithelial cells, which the authors also mention. It may be interesting to check if any signaling pathways involved in this crosstalk are affected under hypoxic conditions in their existing data sets of bulk RNA SEQ. This can be done by using available algorithms such as CellChat (Jin et al. 2021; Jin, Plikus, and Nie 2023), which has been reported to work also in bulk RNA seq data analysis (according to GitHub). The authors could mine the literature for existing RNA sequencing data that include osteoblasts, chondrocytes and epithelial cells (Ozekin, O'Rourke, and Bates 2023; Piña et al. 2023).

We are very grateful to the reviewer for this suggestion. Moreover, we like to thank the reviewer for mentioning exemplary references. We followed the advice by the methodology lined out in results and materials and methods sections: we applied the CellChat algorithm on a scRNA-seq dataset (Pina et al, 2023; Sun et al., 2023) to identify pathways containing components that are hypoxia-attenuated (and associated with a risk for OFC) in our bulk RNA-seq dataset (Figs. 7F-I). We did not use the datasets the reviewer had suggested, because the data were not available for us or the file format was not well-suited for the analysis with CellChat. Importantly, the dataset from Sun et al. has the following advantages over the suggested references: the complete maxillary prominence was used (instead of palatal shelves only), and different time points were included. Thus, we were able to follow the expression of genes of interest at different developmental stages before the onset of differentiation and after (Figs. 7C-E and EV6B). By our approach, we identified several OFC-related pathways that contain hypoxia-attenuated components such as BMP and FGF signaling and deposition of collagen and fibronectin (Figs. 7F-I). Importantly, the named pathways (and others) send outgoing communication patterns to epithelial cells. Therefore, hypoxia-attenuated gene induction in CNCC could influence epithelial cells via these pathways.

We believe that the use of the CellChat algorithm has brought a deeper understanding of how hypoxia can have indirect consequences on the important topic of epithelial cells and thus could also evoke OFC. We therefore once again like to express our gratitude to the reviewer.

Point 2-9

Additionally, another process that may be affected is EMT (epithelial-to-mesenchymal-transition) and is possible to assess by re-analysis of bulk RNA-seq data while focusing on key genes implicated in this process (i.e. E-cadherin, vimentin, EpCAM, Snail, Twist, PRRX1).

We thank the reviewer for the advice. We followed the advice and analyzed cellular morphology by the parameters cell length, total number of pseudopodia, number of filopodia and number of lobopodia (Figs. EV1C-F) (cf. point 2-3). As we did not detect any differences between 21% and 0.5% O2, and because the cells we used for our analyses represent mesenchymal cells, i.e. cells that had already undergone EMT, we did not re-analyze our dataset with the focus on EMT.

Point 2-10

Lastly, when the authors report on the significantly up- or down-regulated genes, it may be interesting to categorize them by ligands, receptors, intracellular molecules and transcription factors (and use separate plots to visualize them). While a big focus of the manuscript are down-regulated genes, less emphasis was given in upregulated genes (other than the response to hypoxia gene module).

We thank the reviewer for the advice. Following this advice, we categorized genes according to Panther protein classes "intercellular signal molecule" (PC00207), "transmembrane signal receptor" (PC00197) and "gene-specific transcriptional regulator" (PC00264) and depicted the results with violin plots (Fig. EV5B). We could not analyze intracellular molecules, because this protein class does not exist in the Panther database. We had not focused on the genes with stronger induction in hypoxic condition, because the number of genes was low in each differentiation paradigm (7 in chondrocytes, less than 30 in osteoblasts, none in smooth muscle cells) and the transcriptional changes were mostly not as drastic as for the attenuated genes. In order to achieve a broader overview of deregulated processes, we now included GO term analyses of genes downregulated during the differentiation regimes both at 21% and 0.5% O2 (Figs. EV2D,E, EV3D,E, EV4D,E).

Point 2-11

The authors are referencing extensively and accurately existing studies in the field and the manuscript is exceptionally well-written, with only a few points of limited clarity or increased complexity. Such an example is when the authors refer to OFC risk genes, because it is not clearly stated how the referenced studies reached their conclusions (for example, are they mouse studies, do they involve mutants, are any of these studies based on GWAS on human cohorts). This matter would significantly improve the flow of the text and highlight the importance of the study and their findings.

We would like to thank the reviewer very much for the appreciation of our scientific writing. We apologize for not explaining exactly how our OFC risk gene lists had been curated. We included this information for both non-syndromic and other OFC risk genes at the respective sites in the results section. Moreover, we included the Human Phenotype Ontology terms that had been used in the search in the materials and methods section.

We thank the reviewer for this suggestion, as we agree that this information significantly highlights the importance of our findings.

Point 2-12

The figures could be redesigned to be more intuitive to interpret. For example, using violin plots and heatmaps, as discussed, and including references or re-analysis/re-use of existing spatial transcriptomics and in situs for marker genes.

In all cases where there is a comparison of gene expression levels, violin plots would be a better representation of up- and down-regulated genes (i.e. selected genes from Fig1K, comparison of gene expression between normoxic and hypoxic NCCs, Fig 2G when analyzing chondrogenesis and the respective analysis for osteoblasts and smooth muscle cells, as well as when comparing the three fate-biasing conditions to identify common genes that are misregulated).

We thank the reviewer for the advice and for the appreciation of the usage of heatmaps (Figs. 2K, 3J, 4J, 6F). Unfortunately, as the number of biological replicates is only three to four, the visualization of gene expression data from our bulk RNA-seq data with violin plots was not intuitive. We therefore retained the heatmaps rather than choosing bar graphs, because they are much clearer when presenting expression data of several to many genes. We included violin plots whenever possible due to high numbers of data points (Figs. EV1C, EV1D, EV1E, EV1F, EV5B). Moreover, we added additional heatmaps to depict transcriptional changes of genes associated with GO terms with the various differentiation regimes (Figs. EV2F, EV3F, EV4F). Unfortunately, we did not detect the three central hypoxia-attenuated genes in spatial transcriptomics data on craniofacial development. But we used scRNA-seq data of different stages of orofacial mouse tissue where we could identify expression of Boc and Cdo1 (cf. points 1-4 and 2-6). These data helped, together with other in vivo data to gain evidence for the in vivo function of Boc and Cdo1 during CNCC differentiation and helped to dismiss Actg2 as another central player.

Significance

Several pieces of evidence have pointed to hypoxia as an environmental factor contributing to congenital orofacial clefts, ranging from studies in mouse to observations in human. The authors are doing an excellent job in putting this information together and the question they are trying to answer is of high importance, given the prevalence of such congenital syndromes.

We are deeply grateful to the reviewer for the appreciation of our work and for classifying our research topic as highly important.

In terms of the methods and model employed, there are some limitations, related to the choice of a mouse cell line over one from human, the severe hypoxia induced (over a more mild), and the conditions of directed differentiation not allowing for simultaneous examination of more complex lineage transitions. The methods as a whole are not that up-to-date, given the single cell and multiplexed transcriptomic advances the last couple of decades, advanced bioinformatics that could be used in combination with in vitro lineage tracing methods.

We thank the reviewer for the honest evaluation of our methods, especially for the constructive suggestions that were given to address our hypotheses with more up-to-date methods and at milder hypoxic conditions. As outlined above, we followed the advice and re-analyzed existing scRNA-seq datasets (cf. points 2-6 and 2-8) and checked our central hypotheses at milder hypoxic conditions (cf. response to point 1-3).

We are deeply convinced that both significantly increased the biological relevance of our results, because we thus (1) gathered evidence for the in vivo function of Boc and Cdo1 and (2) were able to show that the phenomenon of hypoxia-attenuated gene induction still holds true at biologically relevant hypoxic conditions.

The audience this work will reach are neural crest experts, developmental biologists, and potentially clinical doctors. The general public outreach of such a paper is also diverse, as more focus and visibility is required for the individuals affected by those syndromes and their families.

We thank the reviewer for the judgement that our manuscript will not only reach neural crest experts, but also developmental biologists in general and potentially also clinicians. We are very much pleased that the reviewer shares our opinion that affected individuals should be more in the focus of public attention. We like to express our gratitude for the judgement that our manuscript might help to increase focus and visibility for them.

References

---

Barriga EH, Maxwell PH, Reyes AE, Mayor R (2013) The hypoxia factor Hif-1α controls neural crest chemotaxis and epithelial to mesenchymal transition. The Journal of cell biology 201: 759-776, 10.1083/jcb.201212100.

Forman TE, Sajek MP, Larson ED, Mukherjee N, Fantauzzo KA (2024) PDGFRα signaling regulates Srsf3 transcript binding to affect PI3K signaling and endosomal trafficking. Elife 13, 10.7554/eLife.98531.

Funato N, Nakamura M, Yanagisawa H (2015) Molecular basis of cleft palates in mice. World journal of biological chemistry 6: 121-138, 10.4331/wjbc.v6.i3.121.

Gehlen-Breitbach S, Schmid T, Fröb F, Rodrian G, Weider M, Wegner M, Gölz L (2023) The Tip60/Ep400 chromatin remodeling complex impacts basic cellular functions in cranial neural crest-derived tissue during early orofacial development. International Journal of Oral Science 15: 16, 10.1038/s41368-023-00222-7.

Hansen JM, Jones DP, Harris C (2020) The Redox Theory of Development. Antioxid Redox Signal 32: 715-740, 10.1089/ars.2019.7976.

Li D, Tian Y, Vona B, Yu X, Lin J, Ma L, Lou S, Li X, Zhu G, Wang Y et al (2025) A TAF11 variant contributes to non-syndromic cleft lip only through modulating neural crest cell migration. Hum Mol Genet 34: 392-401, 10.1093/hmg/ddae188.

Ng KYB, Mingels R, Morgan H, Macklon N, Cheong Y (2017) In vivo oxygen, temperature and pH dynamics in the female reproductive tract and their importance in human conception: a systematic review. Human Reproduction Update 24: 15-34, 10.1093/humupd/dmx028.

Pina JO, Raju R, Roth DM, Winchester EW, Chattaraj P, Kidwai F, Faucz FR, Iben J, Mitra A, Campbell K et al (2023) Multimodal spatiotemporal transcriptomic resolution of embryonic palate osteogenesis. Nature communications 14: 5687, 10.1038/s41467-023-41349-9.

Sun J, Lin Y, Ha N, Zhang J, Wang W, Wang X, Bian Q (2023) Single-cell RNA-Seq reveals transcriptional regulatory networks directing the development of mouse maxillary prominence. J Genet Genomics 50: 676-687, 10.1016/j.jgg.2023.02.008.

Ulschmid CM, Sun MR, Jabbarpour CR, Steward AC, Rivera-González KS, Cao J, Martin AA, Barnes M, Wicklund L, Madrid A et al (2024) Disruption of DNA methylation-mediated cranial neural crest proliferation and differentiation causes orofacial clefts in mice. Proc Natl Acad Sci U S A 121: e2317668121, 10.1073/pnas.2317668121.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Schmidt and colleagues are addressing the effects of severe hypoxia on proliferation and differentiation potential of (mouse) cranial neural crest, using a neural crest cell line subjected to hypoxic conditions, assessed by transcriptomics analysis (quantitative reverse transcription PCR, bulk RNA sequencing and bioinformatics). They are reporting a mild effect of cell proliferation and an extensive inhibition of differentiation towards osteoblasts, chondrocytes and smooth muscle cells. They reveal affected biological processes shared between the three fate biasing conditions related to cytoskeleton organization and amino acid metabolism. Lastly, among affected genes upon hypoxic conditions in vitro, they authors identified risk genes linked to non-syndromic (non-genetic) orofacial clefts exclusively downregulated in osteoblasts and smooth muscle cells, namely Fgfr2, Gstt1 and Tbxa2. Similarly, hypoxia-driven downregulation of genes implicated in syndromic orofacial clefts was observed in all three chondrocyte, osteoblast and smooth muscle differentiation scenarios. Lastly, STRING analysis of downregulated genes cross-validated their findings related to affected differentiation.

Major comments:

The conclusions drawn from the experimental data are carefully formulated for the most part. One of the main concerns is that the cells were subjected to extreme hypoxic conditions, while it may be more biologically relevant to include a condition representing more mild hypoxia (e.g. 10%). One of the opening claims regarding severe hypoxia only mildly affecting cell proliferation is not shown clearly, since no mitotic markers have been analyzed (i.e. KI67 or PCNA staining or a simple EdU incorporation assay). Thus, the claim that they assessed cell proliferation is not very convincing, even though cell death was analyzed. Additionally, cellular morphology of the cells could be assessed (brightfield images), since previous studies observed that hypoxia can be an inducive factor in cranial neural crest and driving EMT (Scully et al. 2016; Barriga et al. 2013).

Furthermore, in the RNA seq analysis of chondrogenic fate biased cells the authors draw a conclusion based on the proximity of the samples on the PCA plot, which is not very convincing. More careful analysis of the bulk RNA seq data sets they have generated for key marker genes will be more convincing (for example, a heatmap with selected genes would be a helpful representation). As mentioned above, a straight-forward and not time consuming experiment (given that it was assessed for a maximum of 72 hrs) would be to repeat the culture of NCCs and stain for mitotic markers, and quantify the number of positively stained cells over total cell numbers. Furthermore, it is not that demanding to add an experimental condition of less severe hypoxia in this assay. Without underestimating how time consuming this would be, a major lack of experimental validation of the key genes they identify as important across all conditions may be the limitation of the study (this would be the difference between correlation and a probable underlying mechanism). This can be circumvented by more extensive reference to in situ data sets from mouse or existing data sets of single cell and spatial transcriptomicsA suggested targeted knock-down (for example with siRNA, shRNA or CRISPR) to validate a few of the key genes revealed as important could take a few months, with an estimated cost up to 5,000 euros per targeted gene and replicate. On methods, replicates and statistics: The experimental methods and approach are described efficiently and seem reproducible.All biological and technical replicates are of a minimum of N=3 from independent experiments and statistical tests have been run in all cases.

Minor comments:

One of the key implications of NCCs in palate formation is interaction with orofacial epithelial cells, which the authors also mention. It may be interesting to check if any signaling pathways involved in this crosstalk are affected under hypoxic conditions in their existing data sets of bulk RNA SEQ. This can be done by using available algorithms such as CellChat (Jin et al. 2021; Jin, Plikus, and Nie 2023), which has been reported to work also in bulk RNA seq data analysis (according to GitHub). The authors could mine the literature for existing RNA sequencing data that include osteoblasts, chondrocytes and epithelial cells (Ozekin, O'Rourke, and Bates 2023; Piña et al. 2023).

Additionally, another process that may be affected is EMT (epithelial-to-mesenchymal-transition) and is possible to assess by re-analysis of bulk RNA-seq data while focusing on key genes implicated in this process (i.e. E-cadherin, vimentin, EpCAM, Snail, Twist, PRRX1). Lastly, when the authors report on the significantly up- or down-regulated genes, it may be interesting to categorize them by ligands, receptors, intracellular molecules and transcription factors (and use separate plots to visualize them). While a big focus of the manuscript are down-regulated genes, less emphasis was given in upregulated genes (other than the response to hypoxia gene module).

The authors are referencing extensively and accurately existing studies in the field and the manuscript is exceptionally well-written, with only a few points of limited clarity or increased complexity. Such an example is when the authors refer to OFC risk genes, because it is not clearly stated how the referenced studies reached their conclusions (for example, are they mouse studies, do they involve mutants, are any of these studies based on GWAS on human cohorts). This matter would significantly improve the flow of the text and highlight the importance of the study and their findings. The figures could be redesigned to be more intuitive to interpret. For example, using violin plots and heatmaps, as discussed, and including references or re-analysis/re-use of existing spatial transcriptomics and in situs for marker genes.

In all cases where there is a comparison of gene expression levels, violin plots would be a better representation of up- and down-regulated genes (i.e. selected genes from Fig1K, comparison of gene expression between normoxic and hypoxic NCCs, Fig 2G when analyzing chondrogenesis and the respective analysis for osteoblasts and smooth muscle cells, as well as when comparing the three fate-biasing conditions to identify common genes that are misregulated).

References:

Barriga, Elias H., Patrick H. Maxwell, Ariel E. Reyes, and Roberto Mayor. 2013. "The Hypoxia Factor Hif-1α Controls Neural Crest Chemotaxis and Epithelial to Mesenchymal Transition." The Journal of Cell Biology 201 (5): 759-76. https://doi.org/10.1083/jcb.201212100.

Jin, Suoqin, Christian F. Guerrero-Juarez, Lihua Zhang, Ivan Chang, Raul Ramos, Chen-Hsiang Kuan, Peggy Myung, Maksim V. Plikus, and Qing Nie. 2021. "Inference and Analysis of Cell-Cell Communication Using CellChat." Nature Communications 12 (1). https://doi.org/10.1038/s41467-021-21246-9.

Jin, Suoqin, Maksim V. Plikus, and Qing Nie. 2023. "CellChat for Systematic Analysis of Cell-Cell Communication from Single-Cell and Spatially Resolved Transcriptomics." bioRxiv. https://doi.org/10.1101/2023.11.05.565674.

Ozekin, Yunus H., Rebecca O'Rourke, and Emily Anne Bates. 2023. "Single Cell Sequencing of the Mouse Anterior Palate Reveals Mesenchymal Heterogeneity." Developmental Dynamics : An Official Publication of the American Association of Anatomists 252 (6): 713-27. https://doi.org/10.1002/dvdy.573.

Piña, Jeremie Oliver, Resmi Raju, Daniela M. Roth, Emma Wentworth Winchester, Parna Chattaraj, Fahad Kidwai, Fabio R. Faucz, et al. 2023. "Multimodal Spatiotemporal Transcriptomic Resolution of Embryonic Palate Osteogenesis." Nature Communications 14 (September):5687. https://doi.org/10.1038/s41467-023-41349-9.

Scully, Deirdre, Eleanor Keane, Emily Batt, Priyadarssini Karunakaran, Debra F. Higgins, and Nobue Itasaki. 2016. "Hypoxia Promotes Production of Neural Crest Cells in the Embryonic Head." Development 143 (10): 1742-52. https://doi.org/10.1242/dev.131912.

Significance

Several pieces of evidence have pointed to hypoxia as an environmental factor contributing to congenital orofacial clefts, ranging from studies in mouse to observations in human. The authors are doing an excellent job in putting this information together and the question they are trying to answer is of high importance, given the prevalence of such congenital syndromes. In terms of the methods and model employed, there are some limitations, related to the choice of a mouse cell line over one from human, the severe hypoxia induced (over a more mild), and the conditions of directed differentiation not allowing for simultaneous examination of more complex lineage transitions. The methods as a whole are not that up-to-date, given the single cell and multiplexed transcriptomic advances the last couple of decades, advanced bioinformatics that could be used in combination with in vitro lineage tracing methods.

The audience this work will reach are neural crest experts, developmental biologists, and potentially clinical doctors. The general public outreach of such a paper is also diverse, as more focus and visibility is required for the individuals affected by those syndromes and their families.

Reviewer's expertise: mouse neural crest lineage and multipotency, lineage tracing, single cell transcriptomics, NGS, immunofluorescence, molecular methods (RNA, DNA based). Limited expertise with in vitro studies.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Title: Hypoxia impedes differentiation of cranial neural crest cells into derivatives relevant for craniofacial development

Synopsis: Cleft lip w/ or w/o cleft palate is the second-most common birth defect worldwide. Defects are often traceable to cranial neural crest cells through genetics or environmental factors. Schmid and coauthors focus on the environmental factor of hypoxia and investigate the effects of hypoxic conditions on the ability of CNCCs to differentiate and migrate. They performed RNA-seq analysis with qRT-PCR validation for specific markers and show that hypoxia appears to repress differentiation without markedly affecting proliferation. Hypoxic conditions did not demonstrated significant perturbations in cell proliferation; however, chondrocyte, osteoblast, and smooth muscle differentiation was significantly reduced for cell lines cultured under hypoxia. Bulk RNA-seq and PCA revealed dysregulation of genes implicated in cytoskeletal integrity (such as actin γ-2), neural crest cell migration (hedgehog co-receptor brother of CDO) and amino acid metabolism (cysteine dioxygenase), which Schmid and colleagues termed OFC risk genes.

Major comments

• The authors performed qRT-PCR validation for markers of differentiation and hypoxia, with a major absence of VEGF and HIF1a. The paper would be strengthened by mention of these factors, especially by qRT-PCR or Western blot.
• Please provide justification of selection 0.5% as their hypoxic condition or perhaps repeat experiments in a less extreme environment to see if their conclusions still hold true.
• standard immunohistochemistry or histology of differentiated cells would strengthen the authors' claims of reduced differentiation under hypoxic conditions, e.g., Alcian blue, alk-phos or Alizarin red, and smooth muscle actin or other indicator.
• The authors identify a few genes that appear down-regulated in all three differentiation conditions. If it is within the scope of the study, it would strengthen the claim of these genes' function to show the effect of knock-down or knock-out for validation.
• Another major critique lies in the initial claim that proliferation of O9-1 cells is not significantly impacted by hypoxia. In figures 1E-H, photograms of the cells cultured 24 -72 hours and quantifications of live vs dead cells are shown as evidence for this argument. However, the increased density of cells in normoxic conditions may be a confounding variable in this assay. It would be interesting for the researchers to assess the percent of dead vs alive cells between normoxic and hypoxic conditions when the plates reach equivalent densities.
• At end of Fig 1 section authors attempt to tie phenotypes observed in a cell line in vitro to the complex biological processes. They are not comparable and in vivo models would be better suited for these types of comparisons.
• Fig 2: if qRT-PCR did not show statistically different results between experimental and control groups why move on to bulk RNA seq?
• Fig 5: hypoxia this intense is going to affect broad range of biological processes and genes. Finding a few genes that are affected in extreme hypoxia that are also risk genes is highly unlikely. How can the authors be assured that these overlaps are actually significant and not just by chance?
• Would appreciate discussion on how examination of neural crest is relevant for OFC, as most animal models of OFC demonstrate the pathogenesis in embryonic epithelium or periderm, not in the neural crest. Defects in neural crest are associated with other congenital craniofacial anomalies such as craniosynostosis or complex (Tessier) clefts, not the typical orofacial cleft. Please revise rationale of study, interpretation of data and Discussion to specifically state how neural crest cells are involved in the pathogenesis of orofacial cleft.

Minor comments

• The author should replace "Final proof" in the introduction with "further evidence supporting."
• Authors are inconsistent when referring to Figures- sometimes they capitalize (i.e. 1J) and other times they leave lower case (i.e. 1i). Needs to be consistent throughout. Figures are not numbered
• In figures authors would sometimes list 21% O2 first then 0.5% O2 or vice versa. (i.e. Fig on page 21 panels I, J, K). Needs to be consistent.
• Figures on pages 28, 29, 30 panel J and page 31 panel F: there is no legend on what the scale/measurement is for the difference in expression level other than it ranges from -1 to +3.
• Will the authors please comment on the one normoxic sample in Figure 1I that did not cluster with the others? Did this meet the standards to merit exclusion as an outlier?
• The authors refer to DEG as deregulated genes; while not strictly incorrect, the more standard usage is "differentially expressed genes." Please address.

Significance

This work on neural crest cells and hypoxia are biologically and clinically significant.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Major comments

Unfortunately the major conclusions of the article are not well supported by the provided data. Including:

1. That interhemispheric remodelling occurs in non-mammalian amniotes. It would not surprise me that this may be the case, however, the major evidence for this is a series of horizontal insets that do not evidence this point well. There are broad morphological changes during development that can change the proportions and regionalisation of tissue, and therefore the IHF becoming apparently smaller as development progresses (qualitatively, in single sectioning planes, and without clear n numbers) may easily be explained by sutble differences in sectioning planes, or, for example, more caudal territories of the brain expanding at faster rates than the rostral territories. Quantification of the ratio between the IHF and total midline length across ages and between species may go some way to helping to clarify the degree of potential midline remodelling. Very high quality live imaging of the process would be the definitive way to evidence the claim, although I appreciate this is highly technically difficult and may not be possible. A key opportunity seems to be missed in the Satb2 knockout geckoes, where midline remodelling is purported to not occur. This is shown only qualitatively in a single plane of sectioning and again is not convincing. If the IHF length in these animals was quantified to be longer than wildtype at a comparable age, this would help to evidence the claim that remodelling occurs in these species.

Our responses

We take seriously the critique that the series of horizontal section images in the figures do not sufficiently substantiate our claim that interhemispheric remodeling occurs in non-mammalian amniotes. To address this, we plan to create a simplified atlas composed of adjacent sections of various wild-type amniotes as well as Satb2-knockout geckos.

Additionally, in response to the suggestion that the IHF (interhemispheric fissure) should be quantified relative to the total midline length across developmental stages and species, we note that Figure 1 already presents such an analysis. Specifically, we have quantified changes in the midline collagen content using Principal Component Analysis (PCA) in Satb2 Crispants in geckos(FigureS4). However, if necessary, we also plan to perform a similar analysis on wild-type soft-shelled turtles at developmental stages before and after interhemispheric remodeling.

That similar cell types contribute to remodelling in non-mammalian amniotes as mice/eutherian mammals. The microphotographs presented are not of very high quality, and it is often difficult to be convinced that the data is showing the strong claims made in the paper. For instance the "MZG-like cells" may in fact be astrocytes or another cell type as it is hard to visualise morphology, and the "intercalation of GFP-positive radial glial fibres" is very unclear from the photos. The colocalization of MMPsense with laminin positive cells is very hard to appreciate from the figure, and again not quantified. Similarly, there is a claim that there was degeneration of laminin-positive leptomeninges during astroglial intercalation, which is an active process that is difficult to infer from a single microphotograph. From the data, I can appreciate that some of the similar broad categories of cell types that exist at the mouse midline (glia, radial glia) are also present in non-mammalian amniote midlines, but it is difficult to be convinced of much more than this from the data presented.

Our responses

We take seriously the critique that the degeneration of Laminin-positive leptomeninges close to astroglial components is not accepted and that the evidence for glial fiber intercalation is insufficient.

Verifying the degeneration of Laminin-positive leptomeninges is highly challenging. However, we have recently developed a method to visualize collagen in the pia mater using μCT and a CHP probe (3Helix Inc.). Preliminary experiments have already revealed pan-collagen deposition in the midline of the telencephalon (with lower amounts in the fusion region) and degeneration of the collagen composing the pia mater. We plan to incorporate these findings into the revised manuscript.

---

That the gecko RPC and CPC connect distinct parts of the brain (rostral and caudal). These tracer injections lacked visualisation of the deposition site to confirm specificity, as well as appropriate quantification. Importantly, the absence of axons in the CPC following the rostral dye deposition (and vice versa) was not shown, which is essential to make the claim that these commissures carry axons from specific parts of the brain. The alternative hypothesis is that all axons are intermixed and traverse both commissures, independent of brain area of origin, which is not at all tested or disproved by the data presented.

Our responses

Thank you for the valuable critique suggestion. To support our claim that the pallial commissure in geckos consists of axons derived from specific brain regions, we should carefully eliminate the possibility that all axons are intermixed and cross both RPC and CPC regardless of brain region.

To address this, we are planning additional experiments and will include a schematic diagram clearly indicating the labeling sites.

---

Overall, the major conclusions of the study are not well supported by the data. A major effort to quantify phenomena and/or dramatically soften conclusions would be needed in order to make the conclusions well supported.

Our responses

We will thoroughly reconsider our conclusions and make significant efforts to revise the manuscript.

Minor comments

1. The n numbers are not always clearly reported

Our responses

We plan to address the clarification of quantitative data and the exact number of replicates.

At times important points reference reviews or articles that do not support the statements as well as the most important primary articles might.

Our responses

We plan to carefully review the manuscript and, in addition to citing the most important primary papers, revise any descriptions that are not sufficiently supported by the cited reviews or articles, as per the suggestions.

Figures showing the entire section that insets were taken from would help to convince that sectioning planes were equivalent, and also show the deposition site of neurovue experiments.

Our response

We will add a schematic showing the locations labeled in NeuroVue and additional experiments as a similar point made in Major comment 3.

The fibre direction of GFAP+ fibres in figure 6 is confusing - It seems from the labelling on the figures as if red is used for the WT condition in mouse, but for the Satb2del condition in Gecko? If this is the case, then it would appear that the fibres are more specifically oriented in the del condition in mice, but in the WT condition of geckoes? There are several instances of this where clearer description and labelling would help the reader to interpret the results.

Our response

We plan to add clarification and indication of the direction of GFAP+ fibers in Figure 6 to make it easier to understand.

Reviewer #1 (Significance (Required)):

This study attempts to address a highly significant, novel and important question, that, if well achieved, would be publishable at a high degree of interest and impact to the basic research fields of brain development and evolution. Unfortunately the major conclusions made by the study are stronger than the data provided is able to evidence, and I remain unconvinced by many of them.

Our responses

We take seriously the suggestion that the major claims made by this study are excessive and so strong that they cannot be proven with the data provided. We will revise the manuscript as necessary.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary

The authors provide a comparative analysis of interhemispheric (IHF) remodeling and its potential role in the generation of commissural axons. Based on histological material from mice, chickens, turtles, and geckos, the IHF remodeling of the midline is divided in two events: caudal and rostral. It is suggested that the rostral event is a preliminary step to the crossing of commissural axons, as it is characteristic of eutherian mammals with a corpus callosum (CC). However, the authors describe similar histologic features in other amniotes during development, particularly reptiles. This is in contrast with the case of the chick, which does not show signs of IHF remodeling nor a rostral pallial commissure. Additionally, deficient transgenic mice and geckos illustrate a potential role of Satb2 in rostral IHF remodeling and subsequent commissural formation. Whereas the topic and the conclusions of the analysis are interesting and provide new knowledge to the evo-devo field, several issues should be addressed prior to publication, such as data precision and presentation to support the main statements in the manuscript.

Major comments:

____-A central point of this article is the splitting of the IHF into rostral and caudal events. The authors suggest that each one can be regulated differentially, and they attribute the rostral remodeling as a step prior to corpus callosum (CC) formation, in contrast to the caudal remodeling. In my opinion, these two events are not sufficiently characterized either in the figures or the manuscript. It is necessary to better describe these two processes that the authors mention. For instance, the authors could add or re-organize information in Figures 1-3 to include wide-field images showing the whole septum from rostral to caudal, and representative dorsoventral sections at important stages (with insets pointing at specific features). Otherwise, a table summarizing the rostral and caudal events would also be helpful to the reader.

Our responses

We take the suggestion seriously that the distinction between rostral and caudal remodeling may not be clear, especially regarding rostral remodeling, which is prior to the stage of corpus callosum (CC) formation, in contrast to caudal remodeling. Specifically, we plan to add or restructure the information from Figures 1 to 3 by including wide-field images that show the entire septum from rostral to caudal, as well as representative sagittal sections along the dorsal-ventral axis at key stages, with insets highlighting specific features. These will be added to the Supplementary data. Additionally, a table summarizing the events in both the rostral and caudal regions will also be created and included in the revised manuscript.

When the authors refer to the reptilian rostral pallial commissure (RPC) and caudal pallial commissure (CPC), are these the same structures as the pallial commissure and anterior commissure described by Lanuza and Halpern (1997), Butler and Hodos (2005) and Puelles et al. (2019)? It is necessary to clarify the nomenclature, given that they are providing data from several species. Also, structures with the same names among species may not be truly homologous. A simple atlas with some horizontal and transverse planes highlighting anatomical landmarks and important structures (commissural tracts in this case) of the non-mammalian species would be extremely useful for the reader.

Our responses

As suggested by the reviewer, we are considering to provide a more detailed definition of the nomenclature of the pallial commissure in the revised manuscript, specifically in the introduction. Additionally, as mentioned earlier, we plan to create a simplified atlas with several horizontal and transverse sections, emphasizing anatomical landmarks and important structures (in this case, the commissural pathways) in species other than mammals.

____I wonder if the authors tested Fgf8 as marker on any of their sauropsidian tissue samples, as this gene has a known role in murine MZG development, which is required for IHF remodeling (Gobius et al. 2016, already cited in the manuscript). It would be beneficial to test this marker for the study, and if positive, it would open the possibility of designing loss-of-function experiments in avian or reptilian development models to identify mechanisms common to eutherians and support the statements of this work.

Our responses

We plan to verify the gene expression necessary for mouse MZG development and IHF remodeling, including Fgf8, DCC, and MMP2, through immunohistochemical staining as suggested.

It would be really interesting to provide a more elaborate discussion on whether authors consider the sauropsidian IHF as a homologous process to eutherian IHF, and the reptilian RPC as an homologous of the CC.

Our responses

Since 3 out of the 4 reviewers consider IHF remodeling in sauropods to be homologous to that in placental mammals, we plan to further emphasize this claim in the revised manuscript. Additionally, we will expand on the discussion regarding whether the process of RPC formation in reptiles is considered homologous to that of the corpus callosum, and I will approach this from the context of character identity mechanisms claimed by Dr. Günter Wagner.

Data and methods are presented in such a way that, in principle, they could be reproduced. Authors should indicate the number of animals/replicates of each species used in each experiment.

---

Our responses

As suggested, we plan to provide more detailed descriptions of the methods to ensure reproducibility. This will include adding the number of samples and trial repetitions for each animal species used in the experiments, including those for the additional experiments, in the revised manuscript.

---

Minor comments:

In the results section, paragraph 2, line 3: "We detected the accumulation of GFAP-positive cells and phosphorylated vimentin (Ser55) -positive mitotic radial glia in the IHF and telencephalic hinge in developing turtles, geckoes and chicks (Figure 2A)". Figure 2A shows sections from the four analyzed species labeled with radial glia markers at the end of the IHF remodeling. It would be beneficial to have analogous sections at several time points (perhaps before or after the process) to compare and show more clearly the accumulation of glial cells at that location.

Our responses

We have prepared serial sections before and after the developmental stages when interhemispheric remodeling occurs, in order to compare and more clearly show the accumulation of glial cells at their respective locations in mice, geckos, and soft-shelled turtles. I plan to add these results to Figure 2A in the revised manuscript.

The article will improve its quality by adding more comparative information in the introduction about the analyzed sauropsidian structures (rostral pallial commissure and caudal pallial commissure), their relations with the pallial and anterior commissures, the structures/cells connected by them, and homologies previously proposed.

Our responses

We will add comparative information regarding the brain structures in sauropod, including the rostral and caudal pallial commissures and their relationship to the pallial commissure and anterior commissure, and the structures they connect, such as pyramidal cells, along with previously proposed homologies. This information will be included in the introduction and summarized in a table.

---

In Figure 1 panels A-D, there is a lot of disparity in brain sizes and scales both between sections of the same species and between species. Placing the insets next to their source images is very necessary for clarity.

---

Our responses

As mentioned earlier, I will create a simplified atlas using adjacent sections and continuous μCT tomography images. Additionally, I will adjust the placement of the inset images in the revised manuscript to more visually accessible positions, improving their visibility.

In the results section, paragraph 2, line 11: "In addition, it was suggested that astroglial intercalation occurs in conjunction with the aforementioned regression of the IHF from st.21 to st.26 in the developing turtle (Figure 2C)." In Figure 2C, all images are at different scales,

which makes it very hard to properly compare between stages.

Our responses

By creating inset images based on the low-magnification images in the upper panel, we will enhance the visibility of GFAP intercalation. Additionally, we will improve the visibility in the revised manuscript by adding scale bars, referencing the simplified atlas in the figure legends, and standardizing the tissue specimen scale. we also plan to correct any typographical errors in the figures.

In Figure 2D, the authors show the presence of MMP around the leptomeninges, suggesting MMP-mediated degradation. In the images, MMP labeling is revealed in dark blue, which is largely invisible against the black background. Colors should be used properly to allow visualization of this MMP labeling.

Our responses

In Figure 2D, we will reconsider the selection of pseudo-colors and use cyan to represent MMPsense.

In Figure 4, it would really help if the authors provided wide-field images and DAPI counterstaining of the anterograde and retrograde tracings, to provide anatomical landmarks that help readers to identify the midline and understand the orientation of images.

---

Our responses

In addition to the previously mentioned schematic diagram of the gecko's pallial commissure and the additional experiments, we plan to include wide-field images along with forward and retrograde tracing using Hoechst counterstaining.

In Figure 5B, I understand that the images in the red and blue squares correspond to brain areas in the squares in A. However, some confusion remains, especially with the image in B, which does not seem to be at the same angle as in the diagram representation. This makes it difficult to understand the results.

Our responses

According to the comment, we will revise the design of the Figure 5B to be more easily understand, and modify the scheme to match the angle of sections with actual figures.

In Figure 6D, to better visualize defects in the RPC formation, the asterisk in the middle of the deficient structure needs to be replaced with a more lateral arrow pointing to the malformation.

---

Our responses

To better visualize the absence of RPC formation in Figure 6D, we will replace the asterisk in the center of the missing structure with a horizontal arrow indicating the malformation.

In Figure S5, violin plots in panel C do not correspond with data in A and B. This needs correction or clarification.

Our responses

In Figure S5, the inconsistency between the violin plot in panel C and the data in panels A and B is a clear error, and we will correct this in the revised manuscript.

In the article, a section appears solely to explain spatial transcriptomics results in a chick coronal section. The conclusion of this experiment is that three markers associated with midline remodeling are present in chick, suggesting that interhemispheric remodeling is conserved between mouse and chick. As these are complementary results and are not deeply analyzed in this manuscript, I think it would be better to summarize these findings in a dedicated paragraph and transfer some of the key images from Figure S2 to one of the main figures. Other problems with Figure S2: color contrast between clusters in the tSNE projection in B is very poor, should be enhanced; color intensity in FeaturePlots of panels D-F is too weak, and it seems that there is not really much expression at all in any cluster for any of these genes.

Our responses

In the revised manuscript, we will move some of the key images from Figure S2 to Main Figure 3 to demonstrate that the three markers related to midline remodeling are also present in chickens, showing that interhemispheric remodeling is conserved between mice and chickens. Additionally, we will enhance the contrast between clusters in the tSNE projection of the FeaturePlots in S2B and D-F by increasing the pseudo-color intensity or adjusting the intensity levels to emphasize the color contrast, and incorporate this updated figure into the revised manuscript.

Reviewer #2 (Significance (Required)):

The authors identify in the developing brain of sauropsids an event similar to IHF remodeling in eutherians, and suggest a causal relation between the rostral IHF remodeling and the formation of the pallial commissure in reptilian brains. This implies a potential homology between the pallial commissure and the corpus callosum of placental mammals. If this is the intention of the authors, this conclusion should be addressed explicitly and at length in the Discussion section. Whereas the results and conclusions described in the manuscript will be valuable in the field, the data presented in the manuscript needs quite some improvement, particularly for some of the images in the previously mentioned figures. Otherwise, the original data cannot be properly judged and may set reasonable doubt to readers.

Advance: The findings described in this report are new to my knowledge. The description of the IHF remodeling event prior to corpus callosum development in mice has been published (Gobius et al. 2016, Cell Reports), but not in other mammalian branches or non-mammalian vertebrates. For this reason, the data in this report should be very convincing and better presented.

Audience: This research will be interesting for a specialized and basic research audience, particularly for researchers in the evo-devo fields.

My expertise: neuroanatomy, development, evolution, brain, cerebral cortex

Our responses

Thank you for your positive feedback on the novelty and high evaluation of identifying phenomena in reptilian development that resemble interhemispheric fissure (IHF) remodeling in placental mammals and demonstrating a causal relationship between rostral IHF remodeling and the formation of the reptilian pallial commissure. we will incorporate the concept of the potential homology between the corpus callosum in placental mammals and the brain commissures in reptiles into the revised manuscript, reflecting this in the context of character Identity mechanisms claimed by Dr. Günter Wagner. This will be clearly and thoroughly discussed in the discussion section. Additionally, we sincerely appreciate the constructive comment about the room for significant improvement, particularly in some of the figures, and we will address these points in the revised manuscript.

---

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Conserved interhemispheric morphogenesis in amniotes preceded the evolution

of the corpus callosum. Noji Kaneko et al., 2025

The CC is formed exclusively in placental mammals. In other amniotes species, the communication of the two hemispheres is mediated by other structures such as the anterior commissure or the hippocampal commissure. The authors perform anatomical comparisons between species to conclude that interhemispheric fissure remodeling, a prior developmental step for CC formation, is highly conserved in non-mammalian amniotes, such as reptiles and birds. They suggest that might have contributed to the evolution of eutherian-specific CC formation. In an attempt to test their hypothesis, the authors investigate the role of Satb2 in interhemispheric fissure remodeling. They show IH fissure defects in both mice and geckoes. This is a nice manuscript that bridges a gap in the current understanding of CC formation. The study is mostly anatomical and directed at a specialized community.

Our response

We appreciate for positive comments on the manuscript.

I suggest some changes that might contribute to improving the manuscript.

Main

1. Much of the most important conclusions are extracted from the anatomical observation of the dynamics of IHF closure and the emergence of the Hinge. It is very clear that the researchers are specialists in the field but for a broader audience, the images they provide are not always easy to interpret. It takes a lot of effort to visualize the anatomical data they use for their conclusions. As an example, perhaps the authors can find ways to explain how to identify the hinge specifically. It is very clear what the hinge is in the schemes (drawings)but forms one picture to the other at different developmental stages neither in the same animal species nor from different species. In Figure 1, it is difficult to see how the hinge in the mouse is similar (i.e. the same structure) to the hinge in the Gecko and chick. Moreover, in panels C , chick brain sections are shown at much greater magnification than the gecko, and thus is very difficult. In addition, in the manuscript text, the authors refer to sequential sectioning, but only one image for each stage is shown. They can show more images in supplementary Figures, otr they can just explain that they show the relevant images of the sectioning. As another example, in Fig2A, in the text, the authors explain that they detect the same specific glial components, but the images show very different co-localizations and distributions. In Figures 1 and 3, there are lines indicating Dorsal to ventral. This refers to the sectioning but in reality, what the sections are illustrating is the anterior-to-posterior differences in the IHF. maybe they can clarify it, because at quick sight it can be confusing.

---

Our responses

We sincerely appreciate the feedback regarding the interpretation of images that show the dynamics of interhemispheric remodeling and the emergence of the hinge, which is central to the most important conclusions of this study, as it may not always be easy to interpret. In the revised manuscript, we plan to address this by making the following revisions. For example, to clarify how the hinge corresponds across different species, we will create a simplified atlas to explain that the sections from the main figure are at the same level within the continuous slices.

The authors have to revise the manuscript text to be more precise. For example, In the result section quote "To address whether the interhemispheric remodeling in non-mammalian amniotes is dependent on midline glial activities, we next examined the expression of several glial markers in the reptilian and avian midline regions". the anatomical comparison does not address the role of glial.

Our responses

Thank you for your feedback. I will correct the expression "midline glial activities" to "midline glial components" and incorporate this more accurate terminology into the revised manuscript.

---

As an option to increase the relevance of their work, the authors might want to consider to describe in more detail and moving the results of the RNAseq and the analysis of the Stab2 mutants to the main figures.

---

Our responses

Thank you for your feedback. we will move the RNAseq results and the analysis of Satb2 mutants to the main figures and will describe them in more detail to enhance the relevance of the study. Specifically, we plan to separate Figure 6A-C as independent figures and add Supplementary Figure 5, corresponding to mice and geckos, to the main figures in the revised manuscript.

---

Minor:

Please indicate the length of the scale bars in the figure legends, and not only in the figure panels Fig5. Indicate the animal model in the panel Perhaps they can draw a model of the different mechanisms of caudal and anterior remodeling.

---

Our responses

Thank you for your feedback. I plan to revise the figure legend for Figure 5 by clearly indicating the scale bar length and increasing the font size, as well as including the information in each panel. Additionally, I will add a graphical abstract that illustrates the different mechanisms of caudal and rostral remodeling to enhance visual comprehension.

---

Reviewer #3 (Significance (Required)):

The study addresses a gap in knowledge from an evolutionary perspective. It provides novel hypotheses and an innovative framework for the understanding of cortical development and evolution. however, most of the conclusions are inferred from anatomical observations and the experimental testing of the hypothesis (Mutants and RNAseq analysis) are minor part of the study that could be further developed. The study is interesting for investigators with expertise in brain development and evolution but requires familiarity with comparative anatomy and even then it is difficult to go through the work.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

Overall, this is a well-written manuscript focusing on the evolution of mid-line interhemispheric fusion related to corpus callosum development and evolution from amniotes to eutherian species. The authors also demonstrated that Satb2 plays a critical role in interhemispheric remodeling, which is essential for corpus callosum development. This is a nicely organized and interesting study and the data are compelling. The following are suggestions for improvement, mostly for clarity:

Minor comments:

1. Figure 1A: While the E14 and E17 horizontal sections are informative, the addition of the E12 horizontal section does not provide further information. It would be better to place the inset and the whole image side by side, rather than having them far apart across the whole figure. For Figures 1C-D, is it possible to include horizontal sections for chick at

E14 and Gecko at 45 dpo, as shown in the subsequent images?

Our responses

In Figure 1A, we will replace the current figure with a new one that visually enhances the comparison by placing the inset and the full image side by side. we will also add new horizontal sections of the whole image for chicken E14 and gecko 45 dpo, obtained from μCT tomography images and HE staining, to improve visibility between the images.

---

When comparing across species it is sometimes helpful to use a standard staging system so that different developmentally staged tissue can be compared. A timeline of how embryonic day or dpo equates to stage might be helpful.

Our response

To clarify the developmental stages, I plan to incorporate a time scale into the graphical abstract in the revised manuscript.

---

Figure 2B: It is difficult to discern the perspective without a full, lower power section of Gecko at 45 dpo. Adding a full image with an inset would be helpful. In Figure 1C, it would be helpful to define the magnified area by placing a box on the low magnification image.

Our responses

We plan to add a low-magnification μCT tomography image or HE-stained whole image of the gecko at 45 dpo in the revised manuscript. As for Figure 1C, it has already been included in the preprint.

---

Figures 3B-E: Include the staining methods used for these sections.

Our response

We plan to add a note specifying that the image is stained with HE.

---

Figure 4B: Add a low magnification image with an inset. The current image is a bit confusing as it is unclear what is being shown.

Our responses

We plan to add a low-magnification image showing the entire section and use an inset to indicate the positional relationship of the section's plane in a schematic diagram.

Figures 6A-E: It would be helpful to denote the genotype as Satb2+/- or heterozygous, rather than Satb2 WT/del, which can be confusing. Ensure consistency in genotyping notation throughout all figures. It is noted that some are CRISPR knockdown and could be denoted as such.

Our responses

For CRISPR knockdown, I will adopt the term "CRISPANT" in the revised manuscript. This terminology will be used consistently throughout all figures to avoid confusion in genotype notation.

---

Reviewer #4 (Significance (Required)):

The corpus callosum evolved only in eutherian mammals and its development relies critically on an earlier developmental process known as interhemispheric remodeling. Nomura and colleagues investigate the evolution of these processes and identify that interhemispheric remodeling occurs in reptiles and birds and was therefore already present in the common ancestor of amniotes. This highly conserved developmental process likley evolved early and provided a substrates for major commissures to form throughout evolution.

3.____Description of the revisions that have already been incorporated in the transferred manuscript.

Currently we do not incorporate the revision in the transferred manuscript.

---

__ Description of analyses that authors prefer not to carry out__

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Major

That similar cell types contribute to remodelling in non-mammalian amniotes as mice/eutherian mammals. The microphotographs presented are not of very high quality, and it is often difficult to be convinced that the data is showing the strong claims made in the paper. For instance the "MZG-like cells" may in fact be astrocytes or another cell type as it is hard to visualise morphology, and the "intercalation of GFP-positive radial glial fibres" is very unclear from the photos. The colocalization of MMPsense with laminin positive cells is very hard to appreciate from the figure, and again not quantified. Similarly, there is a claim that there was degeneration of laminin-positive leptomeninges during astroglial intercalation, which is an active process that is difficult to infer from a single microphotograph. From the data, I can appreciate that some of the similar broad categories of cell types that exist at the mouse midline____　____(glia, radial glia) are also present in non-mammalian amniote midlines, but it is difficult to be convinced of much more than this from the data presented.

Our responses

We are confident that this paper provides sufficient evidence that cell types similar to those in non-mammalian amniotes, mice, and placental mammals contribute to interhemispheric remodeling and that glial fiber intercalation occurs. This point is also supported by other reviewers.

In the present study, we have not conducted the MMPsense experiments with the aim of showing the co-localization of MMPsense and laminin-positive cells or pia mater. Contrary to the reviewer's claim, it is important that the non-continuous regions of MMPsense and laminin-positive areas (pia mater), which are extracellular components, are adjacent to each other. Unfortunately, establishing a quantification system using MMPsense is practically impossible.

Major

The implication that Satb2 expression at the midline is necessary for appropriate interhemispheric remodeling. Alternative hypotheses for an inappropriately remodeled midline upon whole-brain Satb2 knockout is that it is not dependent on expression at the midline region. Rather, it could be that, for example, the appropriately timed interaction between ingrowing callosal axons and the midline territory is needed for the timely differentiation and/or behavior of midline cells. Other alternatives include that the lack of axonal midline crossing changes the morphology of the midline territory, including potentially "unfusing" the midline. Given the high prevalence of midline remodelling defects concomitant with callosal agenesis referred to be the authors in the literature, it seems like these alternatives would be worth considering. Indeed, the only article the authors reference in their statement that "several studies implicated that agenesis of CC in Satb2-deficient mice is also associated with defects in midline fusion" is an article where Satb2 was knocked out specifically in the cortex and hippocampus. This result is difficult to interpret, as some Emx1 promotors do label some of the midline territory, however the point stands that it is difficult to interpret solely that Satb2 action at the midline is responsible for the effects. I understand that this is a hard question to investigate, so I would suggest allusion to the alternative hypotheses/interpretations as the main priority when interpreting the data.

Our responses

This study does not aim to demonstrate the detailed molecular function of Satb2 in the developmental processes of the corpus callosum or pallial commissure. We plan to clearly state this point in the revised manuscript and focus on the findings obtained as a result. Based on the histological relationships, we will classify interhemispheric remodeling and consider adding a section in the Discussion to identify the common character identity mechanisms underlying the development of the pallial commissure and corpus callosum. This addition will help provide a more detailed understanding of the remodeling mechanisms. As is well known, discussions of homology are complex, and we understand that providing concrete evidence is even more challenging. When discussing homology, we will emphasize that it must be handled cautiously, and that discussions on molecular features and homology will rely heavily on future research. As an alternative, we plan to position the results of Satb2 Crispants in mice and geckos as evidence of the disruption of character identity mechanisms. By incorporating this perspective into the revised manuscript, we believe it will deepen our understanding of the role of Satb2 and its molecular mechanisms.

Reviewer4

Minor comment 7. There is very valuable data in the supplementary figures. As suggestion is to incorporate Supp. figures S1, S2 and S5 in the main figures.

Our responses

Due to space constraints, we plan to move only Supplementary Figure S5 to the supplementary section, and Figures S1 and S2 will not be included in the main figures of the revised manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

Overall, this is a well-written manuscript focusing on the evolution of mid-line interhemispheric fusion related to corpus callosum development and evolution from amniotes to eutherian species. The authors also demonstrated that Satb2 plays a critical role in interhemispheric remodeling, which is essential for corpus callosum development. This is a nicely organized and interesting study and the data are compelling. The following are suggestions for improvement, mostly for clarity:

Minor comments:

1. Figure 1A: While the E14 and E17 horizontal sections are informative, the addition of the E12 horizontal section does not provide further information. It would be better to place the inset and the whole image side by side, rather than having them far apart across the whole figure. For Figures 1C-D, is it possible to include horizontal sections for chick at E14 and Gecko at 45 dpo, as shown in the subsequent images?
2. When comparing across species it is sometimes helpful to use a standard staging system so that different developmentally staged tissue can be compared. A timeline of how embryonic day or dpo equates to stage might be helpful.
3. Figure 2B: It is difficult to discern the perspective without a full, lower power section of Gecko at 45 dpo. Adding a full image with an inset would be helpful. In Figure 1C, it would be helpful to define the magnified area by placing a box on the low magnification image.
4. Figures 3B-E: Include the staining methods used for these sections.
5. Figure 4B: Add a low magnification image with an inset. The current image is a bit confusing as it is unclear what is being shown.
6. Figures 6A-E: It would be helpful to denote the genotype as Satb2 +/- or heterozygous, rather than Satb2 WT/del, which can be confusing. Ensure consistency in genotyping notation throughout all figures. It is noted that some are CRISPR knockdown and could be denoted as such.
7. There is very valuable data in the supplementary figures. As suggestion is to incorporate Supp. figures S1, S2 and S5 in the main figures.

Significance

The corpus callosum evolved only in eutherian mammals and its development relies critically on an earlier developmental process known as interhemispheric remodeling. Nomura and colleagues investigate the evolution of these processes and identify that interhemispheric remodeling occurs in reptiles and birds and was therefore already present in the common ancestor of amniotes. This highly conserved developmental process likley evolved early and provided a substrates for major commissures to form throughout evolution.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Conserved interhemispheric morphogenesis in amniotes preceded the evolution of the corpus callosum. Kaneko et al., 2025

The CC is formed exclusively in placental mammals. In other amniotes species, the communication of the two hemispheres is mediated by other structures such as the anterior commissure or the hippocampal commissure. The authors perform anatomical comparisons between species to conclude that interhemispheric fissure remodeling, a prior developmental step for CC formation, is highly conserved in non-mammalian amniotes, such as reptiles and birds. They suggest that might have contributed to the evolution of eutherian-specific CC formation. In an attempt to test their hypothesis, the authors investigate the role of Satb2 in interhemispheric fissure remodeling. They show IH fissure defects in both mice and geckoes. This is a nice manuscript that bridges a gap in the current understanding of CC formation. The study is mostly anatomical and directed at a specialized community.

I suggest some changes that might contribute to improving the manuscript.

Main

1. Much of the most important conclusions are extracted from the anatomical observation of the dynamics of IHF closure and the emergence of the Hinge. It is very clear that the researchers are specialists in the field but for a broader audience, the images they provide are not always easy to interpret. It takes a lot of effort to visualize the anatomical data they use for their conclusions. As an example, perhaps the authors can find ways to explain how to identify the hinge specifically. It is very clear what the hinge is in the schemes (drawings)but forms one picture to the other at different developmental stages neither in the same animal species nor from different species. In Figure 1, it is difficult to see how the hinge in the mouse is similar (i.e. the same structure) to the hinge in the Gecko and chick. Moreover, in panels C , chick brain sections are shown at much greater magnification than the gecko, and thus is very difficult In addition, in the manuscript text, the authors refer to sequential sectioning, but only one image for each stage is shown. They can show more images in supplementary Figures, otr they can just explain that they show the relevant images of the sectioning. As another example, in Fig2A, in the text, the authors explain that they detect the same specific glial components, but the images show very different co-localizations and distributions. In Figures 1 and 3, there are lines indicating Dorsal to ventral. This refers to the sectioning but in reality, what the sections are illustrating is the anterior-to-posterior differences in the IHF. maybe they can clarify it, because at quick sight it can be confusing.
2. The authors have to revise the manuscript text to be more precise. For example, In the result section quote "To address whether the interhemispheric remodeling in non-mammalian amniotes is dependent on midline glial activities, we next examined the expression of several glial markers in the reptilian and avian midline regions". the anatomical comparison does not address the role of glial.
3. As an option to increase the relevance of their work, the authors might want to consider to describe in more detail and moving the results of the RNAseq and the analysis of the Stab2 mutants to the main figures.

Minor:

Please indicate the length of the scale bars in the figure legends, and not only in the figure panels

Fig5 .Indicate the animal model in the panel

Perhaps they can draw a model of the different mechanisms of caudal and anterior remodeling.

Significance

The study addresses a gap in knowledge from an evolutionary perspective. It provides novel hypotheses and an innovative framework for the understanding of cortical development and evolution. however, most of the conclusions are inferred from anatomical observations and the experimental testing of the hypothesis (Mutants and RNAseq analysis) are minor part of the study that could be further developed. The study is interesting for investigators with expertise in brain development and evolution but requires familiarity with comparative anatomy and even then it is difficult to go through the work.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The authors provide a comparative analysis of interhemispheric (IHF) remodeling and its potential role in the generation of commissural axons. Based on histological material from mice, chickens, turtles, and geckos, the IHF remodeling of the midline is divided in two events: caudal and rostral. It is suggested that the rostral event is a preliminary step to the crossing of commissural axons, as it is characteristic of eutherian mammals with a corpus callosum (CC). However, the authors describe similar histologic features in other amniotes during development, particularly reptiles. This is in contrast with the case of the chick, which does not show signs of IHF remodeling nor a rostral pallial commissure. Additionally, deficient transgenic mice and geckos illustrate a potential role of Satb2 in rostral IHF remodeling and subsequent commissural formation. Whereas the topic and the conclusions of the analysis are interesting and provide new knowledge to the evo-devo field, several issues should be addressed prior to publication, such as data precision and presentation to support the main statements in the manuscript.

Major comments:

• A central point of this article is the splitting of the IHF into rostral and caudal events. The authors suggest that each one can be regulated differentially, and they attribute the rostral remodeling as a step prior to corpus callosum (CC) formation, in contrast to the caudal remodeling. In my opinion, these two events are not sufficiently characterized either in the figures or the manuscript. It is necessary to better describe these two processes that the authors mention. For instance, the authors could add or re-organize information in Figures 1-3 to include wide-field images showing the whole septum from rostral to caudal, and representative dorsoventral sections at important stages (with insets pointing at specific features). Otherwise, a table summarizing the rostral and caudal events would also be helpful to the reader.
• When the authors refer to the reptilian rostral pallial commissure (RPC) and caudal pallial commissure (CPC), are these the same structures as the pallial commissure and anterior commissure described by Lanuza and Halpern (1997), Butler and Hodos (2005) and Puelles et al. (2019)? It is necessary to clarify the nomenclature, given that they are providing data from several species. Also, structures with the same names among species may not be truly homologous. A simple atlas with some horizontal and transverse planes highlighting anatomical landmarks and important structures (commissural tracts in this case) of the non-mammalian species would be extremely useful for the reader.
• I wonder if the authors tested Fgf8 as marker on any of their sauropsidian tissue samples, as this gene has a known role in murine MZG development, which is required for IHF remodeling (Gobius et al. 2016, already cited in the manuscript). It would be beneficial to test this marker for the study, and if positive, it would open the possibility of designing loss-of-function experiments in avian or reptilian development models to identify mechanisms common to eutherians and support the statements of this work
• It would be really interesting to provide a more elaborate discussion on whether authors consider the sauropsidian IHF as a homologous process to eutherian IHF, and the reptilian RPC as an homologous of the CC.
• Data and methods are presented in such a way that, in principle, they could be reproduced. Authors should indicate the number of animals/replicates of each species used in each experiment.

Minor comments:

• In the results section, paragraph 2, line 3: "We detected the accumulation of GFAP-positive cells and phosphorylated vimentin (Ser55) -positive mitotic radial glia in the IHF and telencephalic hinge in developing turtles, geckoes and chicks (Figure 2A)". Figure 2A shows sections from the four analyzed species labeled with radial glia markers at the end of the IHF remodeling. It would be beneficial to have analogous sections at several time points (perhaps before or after the process) to compare and show more clearly the accumulation of glial cells at that location.
• The article will improve its quality by adding more comparative information in the introduction about the analyzed sauropsidian structures (rostral pallial commissure and caudal pallial commissure), their relations with the pallial and anterior commissures, the structures/cells connected by them, and homologies previously proposed.
• In Figure 1 panels A-D, there is a lot of disparity in brain sizes and scales both between sections of the same species and between species. Placing the insets next to their source images is very necessary for clarity.
• In the results section, paragraph 2, line 11: "In addition, it was suggested that astroglial intercalation occurs in conjunction with the aforementioned regression of the IHF from st.21 to st.26 in the developing turtle (Figure 2C)." In Figure 2C, all images are at different scales, which makes it very hard to properly compare between stages.
• In Figure 2D, the authors show the presence of MMP around the leptomeninges, suggesting MMP-mediated degradation. In the images, MMP labeling is revealed in dark blue, which is largely invisible against the black background. Colors should be used properly to allow visualization of this MMP labeling.
• In Figure 4, it would really help if the authors provided wide-field images and DAPI counterstaining of the anterograde and retrograde tracings, to provide anatomical landmarks that help readers to identify the midline and understand the orientation of images.
• In Figure 5B, I understand that the images in the red and blue squares correspond to brain areas in the squares in A. However, some confusion remains, especially with the image in B, which does not seem to be at the same angle as in the diagram representation. This makes it difficult to understand the results.
• In Figure 6D, to better visualize defects in the RPC formation, the asterisk in the middle of the deficient structure needs to be replaced with a more lateral arrow pointing to the malformation.
• In Figure S5, violin plots in panel C do not correspond with data in A and B. This needs correction or clarification.
• In the article, a section appears solely to explain spatial transcriptomics results in a chick coronal section. The conclusion of this experiment is that three markers associated with midline remodeling are present in chick, suggesting that interhemispheric remodeling is conserved between mouse and chick. As these are complementary results and are not deeply analyzed in this manuscript, I think it would be better to summarize these findings in a dedicated paragraph and transfer some of the key images from Figure S2 to one of the main figures. Other problems with Figure S2: color contrast between clusters in the tSNE projection in B is very poor, should be enhanced; color intensity in FeaturePlots of panels D-F is too weak, and it seems that there is not really much expression at all in any cluster for any of these genes.

Significance

The authors identify in the developing brain of sauropsids an event similar to IHF remodeling in eutherians, and suggest a causal relation between the rostral IHF remodeling and the formation of the pallial commissure in reptilian brains. This implies a potential homology between the pallial commissure and the corpus callosum of placental mammals. If this is the intention of the authors, this conclusion should be addressed explicitly and at length in the Discussion section. Whereas the results and conclusions described in the manuscript will be valuable in the field, the data presented in the manuscript needs quite some improvement, particularly for some of the images in the previously mentioned figures. Otherwise, the original data cannot be properly judged and may set reasonable doubt to readers.

Advance: The findings described in this report are new to my knowledge. The description of the IHF remodeling event prior to corpus callosum development in mice has been published (Gobius et al. 2016, Cell Reports), but not in other mammalian branches or non-mammalian vertebrates. For this reason, the data in this report should be very convincing and better presented.

Audience: This research will be interesting for a specialized and basic research audience, particularly for researchers in the evo-devo fields.

My expertise: neuroanatomy, development, evolution, brain, cerebral cortex

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The authors study several valuable developmental series of non-mammalian amniotes, reaching the conclusion that interhemispheric remodeling occurs in these species and that it is dependent on transcription factor Satb2

Major comments

Unfortunately the major conclusions of the article are not well supported by the provided data. Including:

1. That interhemispheric remodelling occurs in non-mammalian amniotes. It would not surprise me that this may be the case, however, the major evidence for this is a series of horizontal insets that do not evidence this point well. There are broad morphological changes during development that can change the proportions and regionalisation of tissue, and therefore the IHF becoming apparently smaller as development progresses (qualitatively, in single sectioning planes, and without clear n numbers) may easily be explained by sutble differences in sectioning planes, or, for example, more caudal territories of the brain expanding at faster rates than the rostral territories. Quantification of the ratio between the IHF and total midline length across ages and between species may go some way to helping to clarify the degree of potential midline remodelling. Very high quality live imaging of the process would be the definitive way to evidence the claim, although I appreciate this is highly technically difficult and may not be possible. A key opportunity seems to be missed in the Satb2 knockout geckoes, where midline remodelling is purported to not occur. This is shown only qualitatively in a single plane of sectioning and again is not convincing. If the IHF length in these animals was quantified to be longer than wildtype at a comparable age, this would help to evidence the claim that remodelling occurs in these species.
2. That similar cell types contribute to remodelling in non-mammalian amniotes as mice/eutherian mammals. The microphotographs presented are not of very high quality, and it is often difficult to be convinced that the data is showing the strong claims made in the paper. For instance the "MZG-like cells" may in fact be astrocytes or another cell type as it is hard to visualise morphology, and the "intercalation of GFP-positive radial glial fibres" is very unclear from the photos. The colocalization of MMPsense with laminin positive cells is very hard to appreciate from the figure, and again not quantified. Similarly, there is a claim that there was degeneration of laminin-positive leptomeninges during astroglial intercalation, which is an active process that is difficult to infer from a single microphotograph. From the data, I can appreciate that some of the similar broad categories of cell types that exist at the mouse midline (glia, radial glia) are also present in non-mammalian amniote midlines, but it is difficult to be convinced of much more than this from the data presented.
3. That the gecko RPC and CPC connect distinct parts of the brain (rostral and caudal). These tracer injections lacked visualisation of the deposition site to confirm specificity, as well as appropriate quantification. Importantly, the absence of axons in the CPC following the rostral dye deposition (and vice versa) was not shown, which is essential to make the claim that these commissures carry axons from specific parts of the brain. The alternative hypothesis is that all axons are intermixed and traverse both commissures, independent of brain area of origin, which is not at all tested or disproved by the data presented.
4. The implication that Satb2 expression at the midline is necessary for appropriate interhemispheric remodeling. Alternative hypotheses for an inappropriately remodeled midline upon whole-brain Satb2 knockout is that it is not dependent on expression at the midline region. Rather, it could be that, for example, the appropriately timed interaction between ingrowing callosal axons and the midline territory is needed for the timely differentiation and/or behavior of midline cells. Other alternatives include that the lack of axonal midline crossing changes the morphology of the midline territory, including potentially "unfusing" the midline. Given the high prevalence of midline remodelling defects concomitant with callosal agenesis referred to be the authors in the literature, it seems like these alternatives would be worth considering. Indeed, the only article the authors reference in their statement that "several studies implicated that agenesis of CC in Satb2-deficient mice is also associated with defects in midline fusion" is an article where Satb2 was knocked out specifically in the cortex and hippocampus. This result is difficult to interpret, as some Emx1 promotors do label some of the midline territory, however the point stands that it is difficult to interpret solely that Satb2 action at the midline is responsible for the effects. I understand that this is a hard question to investigate, so I would suggest allusion to the alternative hypotheses/interpretations as the main priority when interpreting the data.

Overall, the major conclusions of the study are not well supported by the data. A major effort to quantify phenomena and/or dramatically soften conclusions would be needed in order to make the conclusions well supported.

Minor comments

1. The n numbers are not always clearly reported
2. At times important points reference reviews or articles that do not support the statements as well as the most important primary articles might.
3. Figures showing the entire section that insets were taken from would help to convince that sectioning planes were equivalent, and also show the deposition site of neurovue experiments.
4. The fibre direction of GFAP+ fibres in figure 6 is confusing - It seems from the labelling on the figures as if red is used for the WT condition in mouse, but for the Satb2del condition in Gecko? If this is the case, then it would appear that the fibres are more specifically oriented in the del condition in mice, but in the WT condition of geckoes? There are several instances of this where clearer description and labelling would help the reader to interpret the results.

Significance

This study attempts to address a highly significant, novel and important question, that, if well achieved, would be publishable at a high degree of interest and impact to the basic research fields of brain development and evolution. Unfortunately the major conclusions made by the study are stronger than the data provided is able to evidence, and I remain unconvinced by many of them.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This study by Zimyanin et al. examines the role of the C. elegans chromokinesin KLP-19 in the formation and architecture of the anaphase central spindle in C. elegans zygotes. Through a combination of electron and light microscopy, along with RNAi-mediated perturbations, the authors propose that KLP-19 influences central spindle stiffness by regulating microtubule dynamics.

In Figure 5, the effect of KLP-19 depletion on central spindle microtubules appears unconvincing. The FRAP results show no significant difference with or without KLP-19, and overall microtubule density does not consistently respond to its depletion. Additionally, the double klp-19; gpr-1/2 (RNAi) condition does not exhibit a strong increase in microtubule density, though a statistical test is missing for this condition. Furthermore, the spd-1; gpr-1/2 double depletion produces a similar increase in microtubule density to most klp-19 depletion conditions, suggesting that the effect cannot be solely attributed to the absence of KLP-19.

Figure 5A shows that depletion of KLP-19 leads to an increase in tubulin signal in the spindle midzone. The reviewer is correct, that there are differences in the microtubule density between KLP-19 depletion alone and KLP-19 + GPR-1/2 depletion. While depletion of KLP-19 alone leads to a significant increase, co-depletion of GPR-1/2 and KLP-19 leads to a slight, but not significant increase. Along this line, we have added Supplementary Table 1 that contains all p-Values for the different conditions for Figure 5A. However, depletion of GPR-1/2 alone does not affect the microtubule density in the midzone, arguing that changes in pulling forces do not affect the microtubule density in the midzone. It is possible, that the double RNAi leads to a decrease in efficiency and thus a reduced effect on microtubule intensity. We will demonstrate the RNAi efficiency by western blot. Another possibility is that there are some feedback mechanisms that responds to presence/ absence of pulling forces and some of our data (not from this manuscript) hints in this direction, but we have not yet worked out the details of this. We are planning to publish this in a follow up publication.

• *

In response to the spd-1 + gpr-1/2 (RNAi), the reviewer is correct, that the microtubule density in the midzone is not significantly different from klp-19 (RNAi) conditions and we think it is interesting to note that spd-1 + gpr-1/2 (RNAi) leads to an increased microtubule density in the midzone. This could be, as above mentioned caused by some feedback mechanisms that responds to pulling forces, or also due to some functions of SPD-1 that affects microtubules in the midzone. Interestingly, our data also shows that metaphase spindles are significantly shorter in the absence of SPD-1 in comparison to spindles in control embryos, suggesting that SPD-1 plays a role in regulating microtubules or force transmission. We are currently working on understanding SPD-1's role in this process.

• *

We also agree that there is no significant effect on the microtubule turn-over as shown in Figure 5B and we have stated this in the text. Our data does show a trend to a decreased turn-over, but the difference is not significant. This could be due to the low sample number.

• *

Overall, we think our data, the light microscopy and even more so the EM data does show a clear effect on midzone microtubules.

• *

The use of hcp-6 depletion to argue that KLP-19 depletion affects central spindle elongation independently of stretched chromatin is problematic. hcp-6 encodes a component of the Condensin II complex in C. elegans, and its depletion leads to chromatin decompaction rather than the stretched, dense chromatin observed in the midzone during anaphase in klp-19 (RNAi) embryos. These conditions are not equivalent and do not effectively rule out the possibility that chromatin stretching contributes to the observed phenotype.

We agree with the reviewer that the HCP-6 experiments do not entirely rule out effects from lagging chromosomes. Proving that the reduced spindle and chromosome separation is not due to lagging chromosomes is challenging. Most of the depletions that lead to lagging chromosomes are based on defective kinetochore microtubule connections, such as depletion of KNL-1, NDC-80 or CLS-2 (CLASP). In C. elegans, this leads to the mass of Chromosomes staying behind in anaphase and increased spindle pole separation, which is not comparable to KLP-19 depletion. Perturbations that do not affect kinetochore microtubules but still lead to lagging chromosomes are often targeting cohesin or condensin. Ultimately none of these conditions are directly comparable.

A probably better way to test this would be to deplete KLP-19 only after metaphase to prevent its effect on chromosome alignment. However, this is currently not possible as the time window is about 1 minute or less. We currently do not have the tools to conduct this type of experiment. As other reviewers also criticized this experiment and its significance for the paper, we have removed this entirely and have added the following part to the discussion about the potential effect of lagging chromosomes:

" *We can not unambiguously rule out that failure to properly align chromosomes and the resulting lagging chromosomal material could also lead to some of the observed effects on spindle dynamics, such as slow chromosome segregation and pole separation rates as well as preventing spindle rupture in absence of SPD-1. However, several observations argue in favor of KLP-19 actively changing the midzone cytoskeleton network and thus affecting spindle dynamics. *

Most of the protein depletions in C. elegans that lead to lagging chromosomes are based on defective kinetochore microtubule connections, such as depletion of CeCENP-A, CeCENP-C, KNL-1 or NDC-80 (70-72). This mostly leads to the Chromosome mass staying behind in anaphase and increased spindle pole separation (70-72), which is not comparable to KLP-19 depletion. Perturbations that do not affect kinetochore microtubules but still lead to lagging chromosomes are often targeting cohesin or condensin, which depletion leads to chromatin decompaction (73-74) rather than the stretched, dense chromatin as observed in the midzone during anaphase in klp-19 (RNAi) embryos. Ultimately none of these conditions are directly comparable, making it difficult to completely rule out an effect of lagging chromosomes. A better way to test this would be to deplete KLP-19 only after metaphase to prevent its effect on chromosome alignment. However, this is currently not possible as the time window is about 1 minute or less and we do not have the tools to conduct this type of experiment.

*Based on our results we hypothesize that the observed spindle dynamics in absence of KLP-19 are not only caused by lagging chromosomes. Instead, KLP-19 RNAi results in a global rearrangement of the spindle and leads to a significant reduction of the spindle size, microtubule overlap, growth rate, and stability. Furthermore, the increase of microtubule interactions after klp-19 (RNAi) could also contribute to lagging of chromosomes and exacerbation of fragmented extrachromosomal material." *

Additionally, the authors report that KLP-19 influences astral microtubule dynamics (Figure 5E), yet in Figure 3E, they show that KLP-19 localizes exclusively to kinetochores and spindle microtubules, excluding astral microtubules and spindle poles. How do they reconcile this apparent contradiction?

We think that KLP-19 localizes also to astral Microtubules. Our KLP-19 GFP CRISPR line is very dim and this makes it hard to see. We are proposing to use a TIRF approach to image KLP-19 GFP on the C. elegans cortex, which we will include in the revised version. In addition, in support of our hypothesis of KLP-19 binding to astral Microtubules as well we would like to note that there is a PhD thesis available from Jack Martin in Josana Rodriguez Sanchez's Lab in Newcastle (LINK, will lead to a download of the thesis! ) that has reported KLP-19s localization to cortical Microtubules in C. elegans. In this thesis the author also reports an effect on astral microtubule growth.

Figure legends lack consistency and do not adhere to standard C. elegans nomenclature conventions (e.g., protein names should not be capitalized, and genetic perturbations should be italicized). Standardizing these elements would improve clarity and readability.

We have checked our figure legend and to our best knowledge the legends adhere to the C. elegans nomenclature. All RNAi conditions are lower case italicized and Protein names are capitalized as it is standard in other C. elegans publications. We have however noticed some variation in our Figures, i.e. EB-2 instead of EBP-2 and have corrected this in all figures.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Zimyanin et al, Chromokinesin Klp-19 regulates microtubule overlap and dynamics during anaphase in C. elegans.

The authors used a myriad of techniques, including confocal live-cell imaging, 2-photon microscopy, second harmonic generation imaging, FRAP, microfluidic-coupled TIRF, EM-tomography, to study spindle midzone assembly dynamics in C. elegans one-cell stage embryos. In particular, they illuminated the role of kinesin-4 KLP-19 in maintaining proper midzone length and organization. Inhibition of KLP-19 results in longer more stable midzones, implying KLP-19 functions in depolymerizing microtubules.

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Minor comments:

Fig 3E / There is an unusual diagonal line bisecting the embryo. Visually this does not affect viewing of the His::GFP and KLP-19::GFP signals. However, when these signals are quantified and normalized (as in Fig 3F), the diagonal bisect displaying different background signal may impact the measurements.

We are very sorry about this line in the images. The line is due to a defect in the camera chip of the spinning disc. We will acquire new images for this Figure using our new spinning disc microscope.

Fig 4B / While the kymographs clearly show KLP-19::GFP motility on microtubules, they also show that the majority of KLP(-::GFP do not move. Perhaps some quantification and discussion of this result is appropriate?

The reviewer is correct that only a small fraction small fraction of molecules, maybe ~10%, moves. We will add this quantification to the paper and discussion. This could be due to several reasons: Many of the non-moving particles are not visibly colocalized with microtubules, which could mean they are sticking non-specifically to the surface (or sticking to small tubulin aggregates that aren't long enough to support movement). In addition, as this experiment is done in a lysate it is hard to interpret if the immobile KLP-19 is not moving because other proteins are bound along the microtubule blocking its way or if the KLP-19 requires some activation (i.e. phosphorylations) to become mobiles. We think this could be very interesting and will follow up on this in the future.

• *

Reviewer #2 (Significance (Required)):

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary:

The anaphase spindle midzone is an essential structure for cell division. It consists of antiparallel overlapping microtubules organized by the antiparallel microtubule bundler PRC1, molecular motors and other regulatory proteins. This manuscript investigates the role of KLP-19 (C. elegans ortholog of human kinesin-4 KIF4A) and SPD-1 (C. elegans ortholog of PRC1) for spindle midzone organization in the C. elegans embryo and its relevance for proper spindle function. Advanced fluorescence microscopy, 3D electron tomography, and a fluorescence microscopy-based single molecule assay in embryo lysate are used in a unique combination. The authors confirm several aspects of PRC1 and KIF4A function in anaphase, as reported in previous work, mostly in human cells and Drosophila embryos and also in C. elegans embryos. Measurements are mostly very quantitative and to a high quality standard. The main difference to previous conclusions is that here, the authors propose that KLP-19 does not interact with SPD-1, in contrast to what has been established for other animal kinesin-4s and PRC1, and instead localizes to the spindle midzone independently of PRC1 by a mechanism that remains unknown. The authors provide evidence that KLP-19 nevertheless controls microtubule overlap length as in other species and that it produces outward forces sliding midzone microtubules apart a movement that SPD-1 counteracts (presumably by friction). The manuscript presents a rich resource of carefully measured quantitative structural and dynamic C. elegans anaphase spindle data.

Major comments:

Key conclusions convincing?

(1) The key conclusions that the length of the central anaphase spindle microtubule overlap remains constant as the C.elegans spindle elongates (Fig. 1), that PRC1 indeed localizes quite precisely to these overlaps as previously assumed based on its in vitro (purified protein) behavior (Fig. 3B) and that the kinesin-4 KLP-19 controls overlap length as in other species (Fig. 3B) are all convincingly shown. What's missing are quantitative KLP-19 together with microtubule polarity profiles in the presence and absence of SPD-1, leaving it unclear to which extent this kinesin localizes to microtubule overlaps in the two situations. Such data seem crucial, given the authors' claim that KLP-19 localizes to the midzone and that this localization of KLP-19 is mostly unaffected by the absence of SPD-1.

If we understand this correctly the reviewer is asking for second harmonic imaging (SHG) together with imaging of KLP-19 GFP. This is currently not possible due to the way this imaging must be done (2-photon of GFP-Tubulin followed by the SHG). The only thing we can do is provide KLP-19 GFP profiles for control and SPD-1 depleted embryos. We can also use the line co-expressing SPD-1 Halo-tag and KLP-19 GFP to plot their respective localizations in control conditions. We are happy to provide such plots. Generally, we see KLP-19 going to the midzone in absence of SPD-1 and the SHG data does show that the overlap is increased. If KLP-19 specifically localizes to microtubule overlap (rather to i.e. microtubule ends) can currently not be distinguished in the spindle midzone. In vitro data from other labs and our in vitro assay suggests that KLP-19 does not specifically bind to antiparallel overlaps but rather microtubules in general.

(2) 'Normalized KLP-19 intensities' are used to demonstrate that the total amount of this kinesin localizing to the spindle midzone does not depend on the presence of SPD-1 (Fig. 3F). Given that this claim represents a major novelty of the study, the efficiency of the SPD-1 knock-down should be documented, ideally by western blot and fluorescence microscopy.

We agree with the reviewer and will provide western blots.

(3) The authors show convincingly that the kinesin KLP-19 contributes to outward microtubule sliding (and can contribute to spindle rupture in the absence of SPD-1) (Fig. 2), which is interesting and in line with the author's main claim.

(4) The interaction between KIF4a and PRC1 is well established in other species and has been clearly demonstrated both in cells and in vitro (with purified proteins). The authors claim that this concept does not apply to the C. elegans orthologs. To show 'in vitro' (outside of the spindle) that the C. elegans homologs KLP-19 and SPD-1 do not interact, the authors use a novel microfluidic fluorescence-based single-molecule assay in lysate (Fig. 4). Although very original, these experiments do not reach the biochemical standard of previous experiments with purified proteins without appropriate controls. Given that the lysate setup is fairly novel, it's advisable to present at least one positive control demonstrating that interactions between soluble proteins can indeed be detected using this assay. It would also be useful to show the absence of interaction between KLP-19 and SPD-1 by a more conventional method like co-IP, again with a positive control, to support the authors' claim. Eventually, experiments with purified proteins will have to unequivocally demonstrate whether KLP-19 and SPD-1 indeed do not interact - something which appears, however, to be beyond the scope of this study. Without additional experimental proof, the authors may want to indicate that these results are of more preliminary nature.

*We agree with the reviewer, and we will conduct co-IPs of SPD-1 and KLP-19. We will also add CYK-4 as a positive control as previous publications have shown the interaction of CYK-4 with SPD-1. We are now generating lines co-expressing CYK-4 GFP and SPD-1 Halo-tag for the co-IP experiments. *

(5) Unfortunately, the authors do not propose an alternative mechanism for KLP-19 localization to the midzone in SPD-1 depleted embryos, limiting the conceptual advance. Does KLP-19 bind directly to antiparallel microtubules, for example in the assay presented in Fig. 4 (where signs of microtubule crosslinking are shown for SPD-1)? If not, how would it accumulate in the midzone (if it does) in the C. elegans embryo anaphase spindle? The authors do also not propose a mechanism explaining why central antiparallel microtubule overlap length does not change as the spindle elongates in anaphase. Moreover, there is no discussion regarding the potential mechanism leading to KLP-19 controlling microtubule dynamics globally instead of locally where the motor accumulates, indicating limitations in mechanistic insight.

*We agree with the reviewer and will add these points to the discussion. *

*To address some of the points: *

*How does KLP-19 end up in the midzone? : Our data shows that localization of KLP-19 does depend on AIR-2 and BUB-1 as previously reported. However, those proteins primarily affect the formation of the midzone. The in vitro assay does not suggest that KLP-19 has a preference for overlaps, unlike SPD-1, but rather binds microtubules in general. One possible mechanism of midzone localization could be microtubule end-tagging, as has been suggested for PRC1 (SPD-1 homolog). This could lead to an accumulation of KLP-19 in the midzone. *

Why does the central overlap stay constant? : This can be explained by constant microtubule growth at the plus-ends why maintaining the overlap length. Alternatively, this could be explained by some (sophisticated) rearrangements of microtubules that ensure the overlap length stays the same. Generally, this is a very interesting question, as each of this scenario still requires that the overlap length is tightly regulated. Our data suggests that this is correlated with microtubule length in the midzone, as KLP-19 depletion leads to longer microtubules and overlap. This suggests that the regulation of microtubule dynamics might be an important factor in this process. We will add this to the discussion.

• *

Potential mechanism leading to KLP-19 controlling microtubule dynamics globally: We think that KLP-19 localizes to spindle and astral microtubules and regulates the dynamics on all of those, leading to a global regulation. By increasing it's concentration locally, microtubule dynamics can be regulated in the midzone. We will add data showing the localization of KLP-19 to astral microtubules.

Claims justified/preliminary and clearly presented?

The observation that the spindle length remains constant throughout anaphase in C. elegans is based on elegant, but unconventional fluorescence microscopy data (Fig. 1A & B). It would be helpful to add images of SHG and two-photon microscopy to help the reader understand the graphs. Measurements are presented based on distances between the poles. It is unclear why the distances between 15-20 µm were chosen and how they translate to anaphase progression. Can measurements be carried out across the entire duration of cell division to demonstrate that the overlap's 'constant length' property is unique to anaphase? (This could demonstrate already in Fig. 1 that the method in principle is capable of measuring different overlap lengths.)

We agree with the reviewer and have moved the SHG images from supplementary Fig. 6 to the main Figure 1A for better visibility. In addition, we have added a plot as an inset in (now) Figure 1B and C explanation of how the used spindle pole distances related to the progression through anaphase. Unfortunately, we can only acquire a single timepoint and not a live movie during the SHG.

Even though the manuscript contains an impressive amount of data, it appears to be lengthy, the motivation for several experiments is not clearly described, and the order of data presentation can probably be improved. For example, it is unclear why SPD-1 profiles are presented late and why KLP-19 profiles are missing - one would expect to see them early on as an essential characterization of the system under study. The motivation of the paragraph investigating the relation of KLP-19 and SPD-1 to HCP-6 is especially unclear (more than 1 page of text describing supplementary material).

We will go through our text again and will revise the order of presented experiments. As stated above, we have removed the HCP-6 data.

The absence of interaction between KLP-19 and SPD-1 is not demonstrated to the same quality standard as the presence of interaction between the orthologs in the literature, which should at least be mentioned.

Additional experiments essential to support the claims of the paper?

KLP-19 profiles in the presence and absence of SPD-1 seem to be essential.

We agree with the reviewer and will add this.

A co-IP of KLP-19 and SPD-1 (including positive control) to prove that the proteins are not interacting would help to support the claim.

We agree with the reviewer and will add this

Data and methods presented so that they can be reproduced? Yes.

Experiments adequately replicated and statistical analysis adequate? Yes.

Minor comments:

Generating cellular electron tomography data is very laborious. It is a pity that no raw data is provided; for example, a slice of a reconstructed tomogram or a video of whole volumes without segmentation would be an informative addition and allow assessment of the data quality.

We agree with the reviewer and will add movies of the raw electron microscopy data.

The clear evidence for direct interaction between PRC1 and kinesin-4 in other species should be correctly acknowledged throughout the text.

We agree with the reviewer and have corrected this

The average (mean or median?) values and STDs reported in the text do not appear to match those in Fig. 1D.

*We thank the reviewer for pointing this out and have corrected the figure. The violin lot showed the mean and percentiles, we have now changed the plot to show mean and STD. *

• *

The kymograph of spd-1 RNAi in Fig. 2A seems stretched, and the size based on the scale bar does not fit the values stated in the text.

We thank the reviewer for pointing this out and have corrected the figure.

The figure numbering, as stated in the text, does not seem to agree with those in Supplementary Figure 8.

*We thank the reviewer for pointing this out and have corrected the text. *

Page numbers and/or line numbers and figure numbers on the figures would help the reader to navigate the manuscript more easily.

We agree with the reviewer and have added this.

Reviewer #3 (Significance (Required)):

The manuscript is a rich resource of quantitative measurements of C.elegans' structural and dynamic spindle properties, using advanced light microscopy and 3D electron microscopy imaging. In large parts, the work confirms previous conclusions of the function of PRC1 and kinesin-4 in the anaphase spindle, but also reports some interesting differences, namely that the C.elegans proteins differ from their orthologs in that they do not interact with each other, raising the question of how the kinesin-4 KLP-19 localizes to the central spindle in this organism. This work is of interest for researchers studying cell division, and specifically spindle architecture, dynamics, and function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The anaphase spindle midzone is an essential structure for cell division. It consists of antiparallel overlapping microtubules organized by the antiparallel microtubule bundler PRC1, molecular motors and other regulatory proteins. This manuscript investigates the role of KLP-19 (C. elegans ortholog of human kinesin-4 KIF4A) and SPD-1 (C. elegans ortholog of PRC1) for spindle midzone organization in the C. elegans embryo and its relevance for proper spindle function. Advanced fluorescence microscopy, 3D electron tomography, and a fluorescence microscopy-based single molecule assay in embryo lysate are used in a unique combination. The authors confirm several aspects of PRC1 and KIF4A function in anaphase, as reported in previous work, mostly in human cells and Drosophila embryos and also in C. elegans embryos. Measurements are mostly very quantitative and to a high quality standard. The main difference to previous conclusions is that here, the authors propose that KLP-19 does not interact with SPD-1, in contrast to what has been established for other animal kinesin-4s and PRC1, and instead localizes to the spindle midzone independently of PRC1 by a mechanism that remains unknown. The authors provide evidence that KLP-19 nevertheless controls microtubule overlap length as in other species and that it produces outward forces sliding midzone microtubules apart a movement that SPD-1 counteracts (presumably by friction). The manuscript presents a rich resource of carefully measured quantitative structural and dynamic C. elegans anaphase spindle data.

Major comments:

Key conclusions convincing?

1. The key conclusions that the length of the central anaphase spindle microtubule overlap remains constant as the C.elegans spindle elongates (Fig. 1), that PRC1 indeed localizes quite precisely to these overlaps as previously assumed based on its in vitro (purified protein) behavior (Fig. 3B) and that the kinesin-4 KLP-19 controls overlap length as in other species (Fig. 3B) are all convincingly shown. What's missing are quantitative KLP-19 together with microtubule polarity profiles in the presence and absence of SPD-1, leaving it unclear to which extent this kinesin localizes to microtubule overlaps in the two situations. Such data seem crucial, given the authors' claim that KLP-19 localizes to the midzone and that this localization of KLP-19 is mostly unaffected by the absence of SPD-1.
2. 'Normalized KLP-19 intensities' are used to demonstrate that the total amount of this kinesin localizing to the spindle midzone does not depend on the presence of SPD-1 (Fig. 3F). Given that this claim represents a major novelty of the study, the efficiency of the SPD-1 knock-down should be documented, ideally by western blot and fluorescence microscopy.
3. The authors show convincingly that the kinesin KLP-19 contributes to outward microtubule sliding (and can contribute to spindle rupture in the absence of SPD-1) (Fig. 2), which is interesting and in line with the author's main claim.
4. The interaction between KIF4a and PRC1 is well established in other species and has been clearly demonstrated both in cells and in vitro (with purified proteins). The authors claim that this concept does not apply to the C. elegans orthologs. To show 'in vitro' (outside of the spindle) that the C. elegans homologs KLP-19 and SPD-1 do not interact, the authors use a novel microfluidic fluorescence-based single-molecule assay in lysate (Fig. 4). Although very original, these experiments do not reach the biochemical standard of previous experiments with purified proteins without appropriate controls. Given that the lysate setup is fairly novel, it's advisable to present at least one positive control demonstrating that interactions between soluble proteins can indeed be detected using this assay. It would also be useful to show the absence of interaction between KLP-19 and SPD-1 by a more conventional method like co-IP, again with a positive control, to support the authors' claim. Eventually, experiments with purified proteins will have to unequivocally demonstrate whether KLP-19 and SPD-1 indeed do not interact - something which appears, however, to be beyond the scope of this study. Without additional experimental proof, the authors may want to indicate that these results are of more preliminary nature.
5. Unfortunately, the authors do not propose an alternative mechanism for KLP-19 localization to the midzone in SPD-1 depleted embryos, limiting the conceptual advance. Does KLP-19 bind directly to antiparallel microtubules, for example in the assay presented in Fig. 4 (where signs of microtubule crosslinking are shown for SPD-1)? If not, how would it accumulate in the midzone (if it does) in the C. elegans embryo anaphase spindle? The authors do also not propose a mechanism explaining why central antiparallel microtubule overlap length does not change as the spindle elongates in anaphase. Moreover, there is no discussion regarding the potential mechanism leading to KLP-19 controlling microtubule dynamics globally instead of locally where the motor accumulates, indicating limitations in mechanistic insight.

Claims justified/preliminary and clearly presented?

The observation that the spindle length remains constant throughout anaphase in C. elegans is based on elegant, but unconventional fluorescence microscopy data (Fig. 1A & B). It would be helpful to add images of SHG and two-photon microscopy to help the reader understand the graphs. Measurements are presented based on distances between the poles. It is unclear why the distances between 15-20 µm were chosen and how they translate to anaphase progression. Can measurements be carried out across the entire duration of cell division to demonstrate that the overlap's 'constant length' property is unique to anaphase? (This could demonstrate already in Fig. 1 that the method in principle is capable of measuring different overlap lengths.)

Even though the manuscript contains an impressive amount of data, it appears to be lengthy, the motivation for several experiments is not clearly described, and the order of data presentation can probably be improved. For example, it is unclear why SPD-1 profiles are presented late and why KLP-19 profiles are missing - one would expect to see them early on as an essential characterization of the system under study. The motivation of the paragraph investigating the relation of KLP-19 and SPD-1 to HCP-6 is especially unclear (more than 1 page of text describing supplementary material).

The absence of interaction between KLP-19 and SPD-1 is not demonstrated to the same quality standard as the presence of interaction between the orthologs in the literature, which should at least be mentioned.

Additional experiments essential to support the claims of the paper?

KLP-19 profiles in the presence and absence of SPD-1 seem to be essential.

A co-IP of KLP-19 and SPD-1 (including positive control) to prove that the proteins are not interacting would help to support the claim.

Data and methods presented so that they can be reproduced? Yes.

Experiments adequately replicated and statistical analysis adequate? Yes.

Minor comments:

Generating cellular electron tomography data is very laborious. It is a pity that no raw data is provided; for example, a slice of a reconstructed tomogram or a video of whole volumes without segmentation would be an informative addition and allow assessment of the data quality.

The clear evidence for direct interaction between PRC1 and kinesin-4 in other species should be correctly acknowledged throughout the text.

The average (mean or median?) values and STDs reported in the text do not appear to match those in Fig. 1D.

The kymograph of spd-1 RNAi in Fig. 2A seems stretched, and the size based on the scale bar does not fit the values stated in the text.

The figure numbering, as stated in the text, does not seem to agree with those in Supplementary Figure 8.

Page numbers and/or line numbers and figure numbers on the figures would help the reader to navigate the manuscript more easily.

Significance

The manuscript is a rich resource of quantitative measurements of C.elegans' structural and dynamic spindle properties, using advanced light microscopy and 3D electron microscopy imaging. In large parts, the work confirms previous conclusions of the function of PRC1 and kinesin-4 in the anaphase spindle, but also reports some interesting differences, namely that the C.elegans proteins differ from their orthologs in that they do not interact with each other, raising the question of how the kinesin-4 KLP-19 localizes to the central spindle in this organism. This work is of interest for researchers studying cell division, and specifically spindle architecture, dynamics, and function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Zimyanin et al, Chromokinesin Klp-19 regulates microtubule overlap and dynamics during anaphase in C. elegans.

The authors used a myriad of techniques, including confocal live-cell imaging, 2-photon microscopy, second harmonic generation imaging, FRAP, microfluidic-coupled TIRF, EM-tomography, to study spindle midzone assembly dynamics in C. elegans one-cell stage embryos. In particular, they illuminated the role of kinesin-4 KLP-19 in maintaining proper midzone length and organization. Inhibition of KLP-19 results in longer more stable midzones, implying KLP-19 functions in depolymerizing microtubules.

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Minor comments:

Fig 3E / There is an unusual diagonal line bisecting the embryo. Visually this does not affect viewing of the His::GFP and KLP-19::GFP signals. However, when these signals are quantified and normalized (as in Fig 3F), the diagonal bisect displaying different background signal may impact the measurements.

Fig 4B / While the kymographs clearly show KLP-19::GFP motility on microtubules, they also show that the majority of KLP(-::GFP do not move. Perhaps some quantification and discussion of this result is appropriate?

Significance

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study by Zimyanin et al. examines the role of the C. elegans chromokinesin KLP-19 in the formation and architecture of the anaphase central spindle in C. elegans zygotes. Through a combination of electron and light microscopy, along with RNAi-mediated perturbations, the authors propose that KLP-19 influences central spindle stiffness by regulating microtubule dynamics.

In Figure 5, the effect of KLP-19 depletion on central spindle microtubules appears unconvincing. The FRAP results show no significant difference with or without KLP-19, and overall microtubule density does not consistently respond to its depletion. Additionally, the double klp-19; gpr-1/2 (RNAi) condition does not exhibit a strong increase in microtubule density, though a statistical test is missing for this condition. Furthermore, the spd-1; gpr-1/2 double depletion produces a similar increase in microtubule density to most klp-19 depletion conditions, suggesting that the effect cannot be solely attributed to the absence of KLP-19.

The use of hcp-6 depletion to argue that KLP-19 depletion affects central spindle elongation independently of stretched chromatin is problematic. hcp-6 encodes a component of the Condensin II complex in C. elegans, and its depletion leads to chromatin decompaction rather than the stretched, dense chromatin observed in the midzone during anaphase in klp-19 (RNAi) embryos. These conditions are not equivalent and do not effectively rule out the possibility that chromatin stretching contributes to the observed phenotype.

Additionally, the authors report that KLP-19 influences astral microtubule dynamics (Figure 5E), yet in Figure 3E, they show that KLP-19 localizes exclusively to kinetochores and spindle microtubules, excluding astral microtubules and spindle poles. How do they reconcile this apparent contradiction?

Edit: In the sentence: "Similar, 60s after anaphase onset, spindles of klp-19 (RNAi) (19.2 μm {plus minus} 0.5 μm) and klp-19/spd-1 (RNAi) treated spindles (16.2 μm {plus minus} 0.6 μm) were significantly shorter in comparison to control (20.6 μm {plus minus} 0.2 μm),".

Figure legends lack consistency and do not adhere to standard C. elegans nomenclature conventions (e.g., protein names should not be capitalized, and genetic perturbations should be italicized). Standardizing these elements would improve clarity and readability.

Significance

The experiments are generally well executed and provide convincing data. However, a key concern is that the role of chromokinesins-particularly Kif4, the vertebrate homolog of KLP-19-in central spindle assembly and microtubule regulation has already been demonstrated (Hu et al., CB 2011). Additionally, the function of KLP-12, a C. elegans paralog of KLP-19, in inhibiting microtubule dynamics was more recently reported and the structural details of this inhibition have been dissected (Taguchi et al., eLife 2022, this prior work should be cited and discussed). Given these considerations, and despite the extensive array of approaches used in this paper, the novelty of the current study appears rather limited and may be of interest for C. elegans researchers mainly.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Revision Plan (Response to Reviewers)

1. General Statements [optional]

Response: We are pleased the reviewers appreciate the power of this novel proteomics methodology that allowed us to uncover new depths on the complexity of the ribosome ubiquitination code in response to stress. We also appreciate that the reviewers think that this is a “very timely” study and “interesting to a broad audience” that can change the models of translation control currently adopted in the field. Characterizing complex cellular processes is critical to advance scientific knowledge and our work is the first of its kind using targeted proteomics methods to unveil the integrated complexity of ribosome ubiquitin signals in eukaryotic systems. We also appreciate the fairness of the comments received and below we offer a comprehensive revision plan substantially addressing the main points raised by the reviewers. According to the reviewers’ suggestions, we will also expand our studies to two additional E3 ligases (Mag2 and Not4) known to ubiquitinate ribosomes, which will create an even more complete perspective of ubiquitin roles in translation regulation.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

        __Response: __We appreciate the evaluation of our work and that the power of our proteomics method established a good foundation for the study. We also understand the reviewer’s concerns and we will detail below a plan to enhance quantification and increase systematic comparisons. The experiments presented here were conducted with biological replicates, but in several instances, we focused on presence and absence of bands, or their pattern (mono vs poly-ub) because of the semi-quantitative nature of immunoblots. We will revise the figures and present their quantification and statistical analyses. In additional, we did not intend to use these stressors interchangeably, but instead, to use select conditions to highlight the complexity the stress response. In particular, we followed up with H2O2 *versus* 4-NQO because both chemicals are considered sources of oxidative stress. Even though it is unfeasible to compare every single stress condition in every strain background, in the revised version, we will include additional controls to increase the cohesion of the narrative, and expand the comparison between MMS, H2O2, and 4-NQO, as suggested. Details below.

To strengthen the work, the following revisions are essential:

R1.1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.

        __Response: __As requested, we will display quantification and statistical analysis of the suggested and new immunoblots that will be conducted during the revision period.

---

R1.3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.

        __Response: __We will follow the reviewers’ suggestion and redesign the analysis to increase consistency and prioritize data under identical conditions. To increase confidence in the mRNA data analysis, we intend to perform follow up experiments and analyze protein abundance of *ARG proteins* and *CTT1 *under different conditions. The remaining data using non-parallel comparisons will be moved to supplemental material and de-emphasized in the final version of the manuscript.

R1.4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

        __Response: __To ensure a better comparison across strains and conditions, we will re-run several experiments and focus on our main stress conditions. Specifically:

• 3D: We plan to re-run this experiment and include MMS

• 3E: We plan to perform the same panel of experiments in rad6D ,and display WT data as main figure.

• 4A-B: We plan to perform translation output (HPG incorporation) experiments with MMS as suggested

• 4C: We plan to re-run blots for p-eIF2a under MMS for improved comparison.

Reviewer #1 (Significance (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript the authors use a new target proteomics approach to quantify site-specific ubiquitin modification across the ribosome before and after oxidative stress. Then they validate their findings following in particular ubiquitination of Rps20 and Rps3 and extend their analysis to different forms of oxidative stress. Finally they question the relevance of two known actors of ribosome ubiquitination, Hel2 and Rad6. It is not easy to summarize the observations because in fact the major finding is that the patterns of ribosome ubiquitination occur in a stresser and enyzme specific manner (even when considering only oxidative stress). However, the complexity revealed by this study is very relevant for the field, because it underlies that the ubiquitination code of ribosomes is not easy to interpret with regard to translation dynamics and responses to stress or players involved. It suggests that some of the models that have generally been adopted probably need to be amended or completed. I am not a proteomics expert, so I cannot comment on the validity of the new proteomics approach, of whether the methods are appropriately described to reproduce the experiments. However, for the follow up experiments, the results following Rps20 and Rps3 ubiquitination are well performed, nicely controlled and are appropriately interpreted.

Maybe what one can regret is that the authors have limited their analysis to the study of Hel2 and Rad6, and not included other enyzmes that have already been associated with regulation of ribosome ubiquitination, to get a more complete picture. It may not take that much time to test more mutants, but of course there is the risk that rather than enable to make a working model it might make things even more complex.

        __Response: __We value the positive evaluation of our work. We also appreciate the notion that it meaningfully expands the knowledge on the complexity of the ribosome ubiquitination code, challenges the current models of translation control, and conducted well-performed, and nicely controlled experiments. To address the main concern of the reviewer, we will expand our work by studying two additional enzymes involved in ribosome ubiquitination (Mag2 and Not4) and provide a more comprehensive picture of this integrated system. Specifically, we will generate yeast strains deleted for *MAG2* and *NOT4*, and evaluate their impact in ribosome ubiquitination under our main conditions of stress. We will investigate the role of these additional E3s in translation output (HPG incorporation), and in inducing the integrated stress response via phosphorylated eIF2α and Gcn4 expression. Additional follow up experiments will be performed according to our initial results.

Reviewer #2 (Significance (Required)):

In recent years, regulation of translation elongation dynamics has emerged as a much more relevant site of control of gene expression that previously envisonned. The ribosome has emerged as a hub for control of stress responses. Therefore this study is certainly very timely and interesting for a broad audience. However, it does fall short of giving any simple picture, and maybe the only point one can question is whether it is interesting to publish a manuscript that concludes that regulation is complicated, without really being able to provide any kind of suggestive model.

My feeling is nevertheless that it will impact how scientists in the field design their experiments and what they will conclude. It will certainly also drive new experiments and approaches, and lead to investigations on how all the different players in regulation of ribosome modification talk to each other and signal to signaling pathways.

        __Response: __We appreciate the comments and the balanced view that studies like ours will still be impactful and contribute to a number of fields in multiple and meaningful ways. With the new experiments proposed here, and used of additional mutants and strains, we intend to propose and provide a more unified model that explain this complex and dynamic relationship.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

R3.1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.

        __Response: __We understand the importance of the suggested experiment. We have already requested and kindly received strains expressing these mutations, which will reduce the time required to successfully address this point. We will perform our translation and ISR assays such as the one referred by the reviewer in Figs. 4A-C and 5E, and results will determine the role of individual ribosome ubiquitination sites in translation control.

R3.2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

        __Response: __As was also requested by reviewer 1 and discussed above (point R1.1), we will conduct quantification and display statistical analyses for our immunoblots. In addition, we will re-run the aforementioned experiments to improve quantification following the reviewers’ request (same gel & diluted control samples).

Reviewer #3 (Significance (Required)):

• General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

• Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

• Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

• The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

__ Response:__ We appreciate that our work will be valuable to a number of fields in protein dynamics and that our method advances the field by measuring ribosome ubiquitination relatively and stoichiometrically in response to stress.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Response: All requested changes require experiments and data analyses, and a complete revision plan is delineated above in section #2.

• *

4. Description of analyses that authors prefer not to carry out

• *

R1.2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.

        __Response: __Although we understand the interest on the proposed result for consistency, this is the only requested experiment that we do not intend to conduct. Because of the lack of overall ubiquitination of ribosomal proteins in *rad6**D* in response to H2O2 (e.g., Silva et al., 2015, Simoes et al., 2022), we believe that this PRM experiment in unlikely to produce meaningful insight on the ubiquitination code. In this context, we expected that sites regulated by Hel2 will be the ones largely modified in rad6*D *and we followed up on them via immunoblot. Moreover, this experiment would not be time or cost-effective, and resources and efforts could be used to strengthen other important areas of the manuscript, such as including the E3’s Mag2 and Not4 into our work.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.
2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

Significance

General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript the authors use a new target proteomics approach to quantify site-specific ubiquitin modification across the ribosome before and after oxidative stress. Then they validate their findings following in particular ubiquitination of Rps20 and Rps3 and extend their analysis to different forms of oxidative stress. Finally they question the relevance of two known actors of ribosome ubiquitination, Hel2 and Rad6.

It is not easy to summarize the observations because in fact the major finding is that the patterns of ribosome ubiquitination occur in a stresser and enyzme specific manner (even when considering only oxidative stress). However, the complexity revealed by this study is very relevant for the field, because it underlies that the ubiquitination code of ribosomes is not easy to interpret with regard to translation dynamics and responses to stress or players involved. It suggests that some of the models that have generally been adopted probably need to be amended or completed. I am not a proteomics expert, so I cannot comment on the validity of the new proteomics approach, of whether the methods are appropriately described to reproduce the experiments. However, for the follow up experiments, the results following Rps20 and Rps3 ubiquitination are well performed, nicely controlled and are appropriately interpreted. Maybe what one can regret is that the authors have limited their analysis to the study of Hel2 and Rad6, and not included other enyzmes that have already been associated with regulation of ribosome ubiquitination, to get a more complete picture. It may not take that much time to test more mutants, but of course there is the risk that rather than enable to make a working model it might make things even more complex.

Significance

In recent years, regulation of translation elongation dynamics has emerged as a much more relevant site of control of gene expression that previously envisonned. The ribosome has emerged as a hub for control of stress responses. Therefore this study is certainly very timely and interesting for a broad audience.

However, it does fall short of giving any simple picture, and maybe the only point one can question is whether it is interesting to publish a manuscript that concludes that regulation is complicated, without really being able to provide any kind of suggestive model.

My feeling is nevertheless that it will impact how scientists in the field design their experiments and what they will conclude. It will certainly also drive new experiments and approaches, and lead to investigations on how all the different players in regulation of ribosome modification talk to each other and signal to signaling pathways.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

To strengthen the work, the following revisions are essential:

1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.
2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.
3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.
4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

Significance

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Using human pluripotent stem cell-based Gel-3D model of amniogenesis this study investigates the transcriptional dynamics of amnion differentiation at single cell level. Seven cell clusters are identified that emerge over four days of differentiation, including progressive amniotic fates precursors, among them CLDN10 progenitor located on the boundary of amnion and epiblast, and primordial germ cell-like fate. Mutational studies support the role of CLDN10 in promoting amniotic but limiting primordial germ cell-like.

Major Comments:

This generally clearly presented study significantly advances our understanding of human/NHP amniogenesis and should be of broad interest with relevance to human reproduction. However, the there are several questions about how experiments were performed, analyzed, presented and interpreted that need to be answered.

1. The presented antibody stainings while beautiful are presented without sufficient quantification. A single representative (?) cyst is shown. Please provide information about how many cysts on average have been analyzed in how many experiments, expression levels of should be quantified to support conclusions such as: Line 116-118 "a subset of cells within the cysts displays reduced expression of NANOG, while TFAP2A expression becomes weakly activated"; or Line 129 "the transition from pluripotent to amnion cell types occurs progressively over the cyst, starting from focal initiation sites".
2. As the the scRNA-seq experiment is one of the main advances of this study and it explores the temporal dynamics and transitional cell populations during amniogenesis this experiment should be performed with two independent biological replicates to investigate the variability of the amniogenesis in this model in terms of the proportion of the 7 distinct cell populations the authors identified in this analysis.
3. Another interesting parallel between the amnion model and the CS7 human gastrula is most Tyser "Epiblast" cells are seen in the "pluripotency-exiting" population of the amnion model. However, pluripotency exit is a hallmark of epiblast as it initiates gastrulation and primitive streak formation/mesendoderm differentiation. This should be analyzed and discussed further, especially that the authors see in the amnion model some cells expressing TBXT at low level.
4. How do the authors explain/interpret the difference in CLDN10 expression at RNA and protein level?
5. Two hESC CLDN10 mutant lines are presented in Figure S4, which are transheterozygous for framesfhit mutations. However, it is not clear how (guideRNAs), in which position of the gene these mutations were generated and what is predicted mutant protein product of each allele. Please provide, gene structure, gRNA position and predicted protein product cartoons. As we do not know the antigen recognized by CLDN10 antibody, these are critical considerations.
6. What are the consequences of these mutations on CLDN10 transcript? qPCR and also scRNA-seq data the authors have.
7. Please indicate in the experiments using CLDN10 mutant lines, which KO line has been used for specific experiment and whether same/different results have been obtained with the two lines.
8. The excess of PGCL cells in CLDN10 KO Gel-3D amnion model is an important observation, but not fully supported by the data. We are presented with single images of mutant cysts at different stages of amniogenesis. Additional data and the number of SOX17+ cells in WT and mutant cysts at should be provided.
9. The authors propose an interesting concept of CLDN10 at the boundary between the amnion and the epiblast promoting amniogenesis and limiting hHPGLC formation. They speculate about the role of tight junction in this process in agreement with increased hHPGLC formation upon ZO1 reduction in another hPSC model. However, surprisingly little discussion is provided about signaling implications of the reported amniogenic transcriptional cascade, and signals emanating from the different amnion progression cell types. Given the important role of BMP in the formation of amnion and hPGCs, notable is increasing expression of BAMBI in progenitor cell types and high expression in specified and maturing clusters. The expression of signaling pathway components should be analyzed and discussed in more depth.

Additional comments:

1. It is not easy to discern the numbers of the seven populations that are detected at D1-D4 from Figure 1C. A panel in Figure 1 illustrating this would be informative.
2. The similarity of the "Ectoderm" cluster from the CS7 human gastrula Tyser et al., 2021 to extraembryonic cell type with amnion/trophectoderm characteristics in hESC 2D-gastruloid model has been reported by Minn et al., Stem Cell Reports, 2021 and this should be acknowledged.

Referees cross-commenting

There is consensus among the reviewers that this is a novel and important work, but additional experiments and their rigorous quantification is needed. Attending to the reviewers comments will significantly elevate this exciting work.

Significance

Occurring upon implantation of human blastocyst, amniogenesis, or formation of the amniotic sac from the pluripotent epiblast, is still poorly understood but essential process of human embryogenesis. The key morphogenetic aspects of amniogenesis, i.e. epithelial polarization of epiblast into a cyst and subsequently differentiation of the portion of the cyst abutting the trophectoderm proximal to the uterus into squamous epithelium is in part modeled by the hESC-based amnion models in which BMP stimulation plays a crucial role. In the Gel-3D amnion model model deployed here, no exogenous BMP is added, however, BMP signaling is activated in the cells by a mechanosensitive cue provided by the soft substrate; hESCs initially form a cyst of epithelial cells expressing pluripotent markers that initiate transcriptional cascade and within 4 days of culture differentiate into a cyst of squamous-amnion-like epithelium.

This work expands on the previous studies by investigating the transcriptional dynamics of amnion differentiation at single cell level combined with additional antibody stainings and compare their findings to distinct cell types in a Carnegie stage 7 human embryo (Tyser et al., 2021) and relevant non-human primate datasets. Based on the resulting data the authors posit contiguous amniogenic cell states: pluripotency-exiting, early progenitor, late progenitor, specified and maturing. Moreover, they also uncover that this model of amniogenesis also produces primordial germ cell-like (hPGC-L) and mesoderm-like cells. A notable finding is that high levels of CLDN10 mark a later transient progenitor state, but CLDN10 expression is downregulated more differentiated cells. Moroever, the authors posit that CLDN10 is a marker of the progenitor population, expression of which is restricted to the boundary between the amnion and the epiblast of the cynomolgus macaque peri-gastrula. Functional interrogation of CLDN10 using hESC mutant lines in the Gel-3D amnion model shows reduced amniogenesis and excess of hPGC-L cells. The authors propose that the CLDN10 the amnion-epiblast boundary is a site of active amniogenesis but limits hPGC-L. This work advances our understanding of amniogenesis, strengthens the concept that amnion and PGC progressing cells initially share acommon intermediate lineage, provides a valuable transcriptomic dataset and should be of broad interest with relevance to human development and reproduction.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Sekulovski et al characterize the transcriptome of human pluripotent stem cells differentiating to an amnion fate in 3D, using single-cell RNA sequencing. This leads to the identification of CLDN10+ cells as amnion progenitors. When CLDN10 is eliminated, amniogenesis is compromised. Moreover, analysis of CLDN10 localization in cynomolgus macaque embryos reveals that this progenitor population is located at the boundary between the epiblast and the amnion.

Major comments:

The key results are convincing and supported by clear experiments. However, additional controls, quantifications, and clarifications are needed as follows:

• The authors identify five amnion-progressing states in vitro and mention that each of these states also shows transcriptional similarities to cell types in a CS7 embryo (Tyser et al, 2021). How do the authors interpret this result? Would this mean that there are amnion cells at all different maturation stages present at a specific time point in development? Given that the available in vivo reference is derived from a single human embryo, it is more likely that the true in vivo counterpart of these states is not captured in the embryo data.
• The authors stain the 3D amnion model at different stages and conclude that "amniogenesis initiates focally and spreads laterally". This cannot be concluded from the data provided. The images in Figure 1 simply show heterogeneity in the levels of TFAP2A. To support their claims, the authors would need to perform time-lapse experiments using a TFAP2A reporter line.
• The authors conclude that CLDN10+ cells give rise to amnion during gastrulation of cynomolgus macaque embryos. The data provided does not prove that CLDN10+ cells are the amnion progenitors in vivo.
• CLDN10 KO cells form amnion cysts like control cells by day 3. However, by day 4 the cysts lose expression of the amnion marker ISL1 and become disorganized. To characterize the epithelial (or lack of) phenotype, the authors should include membrane/polarity/adhesion immunostainings. Is the disorganization observed at day 4 associated with the progressive changes in cell identity, or is it a time-dependent phenotype? The authors should include human PSC cysts as a control. This would allow them to determine whether the role of CLDN10 is specific to amnion cells.
• Figure 2: is there a correlation between the levels of CLDN10 and TFAP2A based on the scRNAseq data and the immunofluorescence stainings? The IF data would benefit from quantifications.
• Figure 4: the experiment has not been quantified. What is the % of PGCLCs in WT and KO cells? What are the levels of ISL1 in WT and KO cells? What is the localization of epithelial determinants in WT and KO cells? Is there an anti-correlation between CLDN10 and ISL1?

Referees cross-commenting

I think there is a general consensus that additional quantifications and careful analyses are needed before this paper is accepted for publication. I agree with the comments raised by the other reviewers.

Significance

This manuscript is a follow-up work of Sekulovski et al, 2024. In this recent manuscript, the authors already provided a temporally resolved transcriptomic characterization of in vitro amniogenesis. The key difference between the two articles is that while Sekulovski et al, 2024 performed a bulk RNAseq experiment, in the current manuscript a single-cell RNAseq experiment has been done. It is fundamental to clearly define what new findings have been obtained thanks to the single-cell experiment, which could not have been obtained using the bulk transcriptomics data. This is a particularly important point given the robustness and synchrony of the model. For example, the authors had already identified five amnion states in vitro in their previous publication. Is CLDN10 differentially expressed in the progenitor population based on the bulk RNAseq data? Are the same dynamics of expression recapitulated? The title of the manuscript does not mention CLDN10 but rather focuses on transcriptional profiling at the single-cell level. In my opinion, the key novelty of this manuscript is the identification of CLDN10 and the role it plays during amniogenesis. Focusing the manuscript on the dynamic transcriptional profile diminishes the novelty, as this had already been done by the authors at the bulk level. Globally, this manuscript provides additional information of the poorly understood process of amniogenesis that will be interesting for those working on early human embryogenesis.

My area of expertise is early mammalian embryo development and stem cells. I do not have the computational background to evaluate the bioinformatic analyses of the manuscript in-depth.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This is a novel and important and interesting study that uses one of the best amniogeneisis form PSCs int he field. The authors do scRNA-seq during 4 day course to understand different populations emerging during amniogeneisis, and they identify CLDN10 as a marker for newly emerging new amion cells, and then use their model and monkey real embryos to prove the CLDN10+ population at the amnion-epiblast border. In the final part, the authors knockout CLDN10 and claim it compromises amniogenesis and favours formation.

Significance

This is a well conducted study, and conclusions are novel and super exciting and IMPORTANT!!!. I have one-2 major comments to strengthen conclusions in the last part, and will help make this excellent study become superb and a landmark study.

1. it is not really clear what is the phenotype of CLDN10 KO cells. is amniogenesis totally inhibited? can the authors do scRNA-seq on the KO cells and compare them to WT cells? There is no quantitation to amnion or PGC formation efficiency ? how many structures where analyzed?
2. in continuation with the above The claim that PGC formation is enhanced in KO is not strong. PGCs should be stained for NANOS3 and blimp1 specific marker and not only SOX17 which can also be a Pre marker. Then quantification should be properly done.

Note: This response was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Dear Review Commons editorial team,

Thank you for coordinating the thorough and careful review of our manuscript. We are especially grateful to the four anonymous reviewers for recognizing the value of our work and for their constructive suggestions on how to improve it.

We are encouraged by the positive reception of our main conclusions on the robustness of adaptation to DNA replication stress and its relevance to multiple fields. All reviewers provided insightful comments, with reviewers #2 and #4 emphasizing that further experimental validation of the hypothesized role of reduced dNTPs in alleviating fitness during constitutive DNA replication stress would strengthen the paper. While the precise molecular mechanisms underlying this suppression are not the primary focus of this manuscript, we are eager to perform additional experiments based on the reviewers’ suggestions.

Below, we present a detailed revision plan in the form of a point-by-point response to their comments.

Reviewer #1 (Evidence, reproducibility and clarity):

This study investigates the compensatory evolutionary response of Saccharomyces cerevisiae to DNA replication stress, focusing on the influence of genotype-environment interactions (GXE). The authors used a range of experimental conditions with varying nutrient levels to assess evolutionary outcomes under replication stress. Their genomic analysis reveals that while glucose levels affect initial adaptation rates, the genetics of adaptation remain robust across all nutritional environments. The research offers new insights into the adaptability of S. cerevisiae, emphasizing the role of the nutritional environment in evolutionary processes related to DNA replication stress. It identifies recurrent advantageous mutations under different macronutrient availabilities and uncovers a novel role for the RNA polymerase II mediator complex in adaptation to replication stress. Overall, this well-designed study adds to the growing recognition of the complexity and robustness of evolutionary responses to environmental stressors. It provides strong evidence that compensatory evolution to replication stress is robust across varying nutritional conditions. It both challenges and reinforces previous findings regarding the resilience of the yeast genetic interaction network to environmental perturbations. The detailed analysis of specific compensatory mutations and their fitness impacts across different conditions offers valuable insights into adaptive dynamics over 1000 generations, contributing a clear empirical framework for understanding how replication-associated stress shapes evolutionary outcomes in diverse environments.

Based on the analysis:

1) The conclusions are generally well-supported by the presented data. The evolution experiments and genomic analyses are robust and provide convincing evidence for the study's main claims. The authors took steps to eliminate bias, such as maintaining an adequate Ne, which, if not done, could have compromised their conclusions by affecting genetic drift and limiting the population's access to beneficial mutations.

2) The figures are well-designed and easy to understand.

3) The methodology is well-described and appears reproducible. The authors provide sufficient details on experimental procedures. Experimental replication is adequate, with multiple evolutionary lines.

4) They also made efforts to validate their observations, such as the validation of mutations, the prediction of interactions in the Med14 structure, and its potential implication in gene regulation, as well as the analysis of the cumulative fitness benefit and the reconstruction of the quadruple mutant.

There are, however, a few results that would benefit from further clarification:

1) The experimental design is strong, offering a diverse range of conditions. However, the high glucose condition (8%) stands out as significantly different from the neutral 2% condition, both in range and margin, compared to the low glucose conditions (0.25-0.5%). While this mainly affects growth profiles and evolvability in the early generations, a brief explanation in the discussion would strengthen the conclusions. Specifically, addressing:

1. a) The rationale behind selecting these particular glucose concentrations.

2. b) How other glucose concentrations might influence the outcomes. Providing this additional context would enhance the reader's understanding of the experimental setup and its potential implications, while also offering insights into the broader applicability of the findings and possible directions for future research.

We thank the reviewer for pointing out the need to clarify the rationale behind the glucose concentrations used in our study, an aspect we agree should have been better explained. In response, we have added the following text detailing the chosen conditions and their established effects on cellular metabolism.

Line 67: “Glucose is the most abundant monosaccharide in nature, and represents the preferred source of energy for most cells.”

Line 110: “...we grew WT and ctf4Δ cells in varying glucose concentrations to induce distinct physiological states. Low glucose levels (0.25% and 0.5%) induce caloric restriction and ultimately glucose starvation (Lin et al 2000, Smith et al. 2009). These conditions elicit increased respiration (Lin et al., 2002), sirtuins expression (Guarente, 2013), autophagy (Bagherniya et al. 2018), DNA repair (Heydari et al., 2007), and reduced recombination at the ribosomal DNA locus (Riesen and Morgan, 2009) ultimately extending lifespan in several organisms (Kapahi et al., 2016). In contrast, standard laboratory conditions typically use 2% glucose, promoting a rapid proliferation environment to which strains have been adapted since laboratory domestication (Lindergren, 1949). Finally, elevated glucose concentrations (such as 8%) result in higher ethanol production (Lin et al., 2012) and reactive oxygen species (ROS) levels (Maslanka et al., 2017).

2) In the discussion section, a more explicit comparison with similar studies in other model organisms would help contextualize the findings within the broader field of evolutionary biology. While the results appear robust, it would be beneficial to explore how they align with or contrast to previous studies on DNA damage, particularly in bacteria or highly complex eukaryotes.

We appreciate this suggestion to better contextualize our findings within the broader literature, as it provides an opportunity to highlight the unique aspects of our work. While many studies have explored how environmental factors shape fitness landscapes and influence evolutionary strategies, to our knowledge, only a few have addressed this in the context of compensatory evolution, where cells must recover fitness lost due to intracellular perturbations. To address this point, we have added a discussion of additional examples involving other model organisms, highlighting their difference with the question asked in this work.

Line 34: “Genotype-by-environment (GxE) interactions are well-documented. For example, several studies on E. coli have demonstrated how different environments influence fitness and epistatic interactions among adaptive mutations in the Lenski Long-Term Evolution Experiment (Ostrowski et al., 2005, 2008; Flynn et al., 2012; Hall et al., 2019). Adaptive mutations in viral genomes similarly exhibit variable fitness effects across different hosts (Lalic and Elena, 2012; Cervera, 2016). Furthermore, interactions between mutations in the Plasmodium falciparum dihydrofolate reductase gene have been shown to predict distinct patterns of resistance to antimalarial drugs (Ogbunugafor et al., 2016). However, the role of environmental factors in shaping evolution within the context of compensatory adaptation, when fitness defects primarily arise from intracellular perturbations, remains much less explored.”

However, if the reviewer have particular additional studies in mind, we welcome further suggestions to include in the final manuscript.

Minor comments:

1) The presentation of data in the figures is clear and informative. However, some figure legends could benefit from more detailed explanations. For example, although the statistical tests used are mentioned in the methods section, it would be helpful to also include them in the figure legends, such as in legend 1acde, as well as in all other figures.

We are now reporting the statistical test used for each comparison also in figure legends.

2) In terms of broader conclusions, here are a few suggestions, though they are, of course, optional:

a) The study could benefit from exploring the potential trade-offs of adaptive mutations in the hypothetical return to environments without replication stress, at least theoretically. This would provide a more comprehensive understanding of the evolutionary constraints.

We thank the reviewer for the suggestion, we had performed the measurements but did not comment on them explicitly. We are now commenting on them as follows:

Line 310: “In the WT background, all mutations were nearly neutral, with only minimal deleterious or advantageous effects on fitness depending on glucose concentrations (Fig S4A).”

Line 468: “The nearly neutral effects on fitness of the core adaptive mutations in WT suggest that they are likely to persist even after the initial replication stress is resolved.”

b) A brief discussion of the potential limitations of using lab strains versus wild isolates of S. cerevisiae would offer valuable context for the generalizability of the findings.

This is an excellent point. While addressing it fully would warrant a separate manuscript, we provide our comments here, along with similar observations raised by this and other reviewers, as follows:

Line 450: “How generalizable are our conclusions about the reproducibility of evolutionary repair to DNA replication stress across other organisms, species, or replication challenges? While dedicated future studies are needed to fully address these important questions, several lines of evidence are encouraging. A recent report demonstrated that the identity of suppressor mutations of lethal alleles was conserved when introduced into highly divergent wild yeast isolates (Paltenghi and van Leeuwen, 2024). Similarly, earlier work showed that even ploidy, which significantly alters the target size for loss- and gain-of-function mutations, affected only the identity of the genes targeted by selection, while the broader cellular modules involved remained consistent (Fumasoni and Murray, 2021). Moreover, divergent organisms experiencing different types of DNA replication stress exhibit some of the adaptive responses described here. For example, the yeast genus Hanseniaspora, which lacks the Pol32 subunit of the replisome, has also been reported to have lost the DNA damage checkpoint (Steenwyk et al., 2019). Human Ewing sarcoma cells carrying the fusion oncogene EWS-FLI1 frequently exhibit adaptive amplification of the cohesin subunit RAD21 (Su et al., 2021). Together, these findings suggest that while the specific details of DNA replication perturbations and the genomic features of organisms may shape the precise targets of compensatory evolution, the overarching principles and cellular modules affected are broadly conserved.”

Furthermore, we plan to search a recently published database of variants found in natural isolates of S. cerevisiae to assess whether similar evolutionary processes to those described in this study may have occurred in wild strains.

c) It would be valuable to present the differences in ploidy in the context of other studies, such as the nutrient-limitation hypothesis (e.g., 'The Evolutionary Advantage of Haploid Versus Diploid Microbes in Nutrient-Poor Environments' by Bessho, 2015), since, as previously demonstrated by the authors of this article that is being reviewed, ploidy may influence the evolutionary trajectories of DNA repair.

d) Interrelating these three terms: nutrient-limitation, ploidy, and DNA repair could be an interesting avenue to explore in the discussion.

In response to comments c and d, we have now commented on the intersection between ploidy and other types of DNA perturbation in the paragraph starting in line 491 (see response above)

3) Specific details:

a) Line 116: To improve clarity, it would be beneficial to refer to the figure right after the statement: 'However, their relative fitness improved compared to the WT reference as the initial glucose levels (Figure X).'

b) Line 404: The statement about antibiotics and cancer progression is somewhat brief here; it might be helpful to provide more context on why this mechanism influences these processes (here or before).

c) Line 418: "were re-suspended in water containing zymolyase (Zymo Research, Irvine, CA, US, 0.025 μ/μL), incubated at". Something is missing in the units.

d) Line 459: "and G2 phases for each genotype was estimated by deriving the the relative cell distribution". The article "the" is repeated.

e) 1a: The x-axis ticks appear misaligned, which makes it difficult to interpret the boxplots. For example, at 0.25, the tick is closer to the orange boxplot than to the black one. In contrast, at 2%, the tick seems well-centered."

f) Figure 3 could benefit from a general legend at the top regarding the colors, as finding it in 2c was not intuitively easy.

The typos and suggestions raised in points 3a-f have now been corrected in the manuscript.

g) I didn't review the code on GitHub.

Reviewer #1 (Significance):

The main strength of the study is that it shows robustness of compensatory evolution across varying nutrient conditions. The study adds to the growing body of literature on DNA replication stress and evolutionary adaptation by showing that compensatory evolution can occur regardless of nutrient availability. This fundamental finding challenges prior assumptions that nutrient conditions significantly alter evolutionary outcomes, contributing to a more nuanced understanding of how cells respond to stress. Furthermore, the discovery of the RNA polymerase II mediator complex's role in this process is particularly novel and opens new lines of investigation.

Advance in the field: The results advance our understanding of evolutionary biology, particularly in the context of DNA replication stress and compensatory evolution. The study demonstrates that evolutionary repair mechanisms are predictable, even under variable environmental conditions, which has key implications for evolutionary biology and therapeutic applications.

Audience:

This paper will be of interest to a specialized audience in evolutionary biology, genomics, and cell biology, particularly those interested in DNA replication stress and adaptive evolution. Researchers studying stress responses in model organisms, such as S. cerevisiae, will find the findings valuable, as will those working in applied fields where stress adaptation is a critical factor (e.g., industrial yeast fermentation, drug development, disease resistance, cancer research, or aging studies).

Expertise:

Evolutionary biology, genomic analysis, and cellular stress responses, with a particular focus on experimental evolution under DNA damage stress in Saccharomyces cerevisiae. Recently graduated and beginner reviewer.

Reviewer #2 (Evidence, reproducibility and clarity):

The paper addresses the effect of sugar availability in shaping compensatory evolution. The first observation of the paper is that cell physiology changes by modulating glucose availability also in strains that come with defective DNA replication (ctf4-null previously studied by the authors). An intriguing result is that ctf4-null grows comparatively better in low concentrations of glucose. This is hypothesized to be a consequence of both the decrease in dNTPs in low glucose, which causes slow down of fork progression, and/or reduced fork collapse at rDNA locus. Hence, wild types and ctf4-null show an opposite trend: in the mutant, the lowest concentration of glucose is the least affected by the mutation; in wild type, the highest concentration is the least affected. Adaptation rate is inversely related with the initial fitness. The effect on physiology and adaptation rate is a starting point for asking the key question: are evolutionary trajectories influnced by the growth conditions? The answer is negative: evolution experiments show the very same core of genetic changes at all sugar concentrations. The result is apparently at odds with previous publications, and the authors conclude that, in this particular setting, availability of carbon sources plays a minor role compared to impaired DNA replication. The different rates of adaptation in WT and mutant is rather explained by the initial fitness at the different glucose concentrations, which, as mentioned, is opposite in WT and ctf4-null mutants. The paper also reports a new mutation in MED14, component of the transcription mediator complex, which rescues the lack of Ctf4 activity. The study is interesting and asks a relevant question. The experiments are well executed and convincing, but the paper can be strengthened by testing some of the hypotheses which are put forward.

Main points

1- The raw data for evolutionary dynamics (Figure S2C) are fitted with the power law suggested by Wiser and Lenski, and return different values of the parameter 'b'. The authors say that the result depends greatly on the initial conditions ("due to the varying initial fitness of ctf4Δ cells across different glucose environments, they display an opposite trend to WT"). Around the initial values, however, the curves are non-monotonic, especially for low glucose availability. Both for WT and ctf4-null there is an initial drop in fitness, after which fitness increases. If one would neglect this initial dynamics, the value of the parameter 'b' would likely be different.

The non-monotonic trend in fitness highlighted by the reviewer is likely due to technical factors: Fitness at Generation 0 was measured with high precision in a low-throughput manner early in the project. In contrast, fitness from Generation 100 to 1000 was measured later in the study in a high-throughput fashion, necessitated by the large number of competitions conducted (96 wells × 4 time points × 6 replicates = 2304 assays). This difference in methodologies may have introduced a slight offset when the datasets were combined at Generation 100. Following the reviewer’s suggestion, we have excluded the data point at Generation 100 responsible for this non-monotonic behavior and re-fitted the curves. While this adjustment has caused minor changes in the parameter ‘b’, the qualitative trends, particularly the opposing trends between WT and ctf4Δ as glucose increases, remain consistent (Figure_rev_only 1). To ensure transparency, we have retained all recorded fitness values in the original figure for reference.

In general, one can question whether curves with this shape are best fitted by the power law proposed by Wiser and Lenski. For example, for the WT 0.25% glucose the linear fit gives a better R2 (why do the authors show the linear fit anyway?). This impression is further reinforced by the observation that Wiser and Lenski fit dynamics that last 50.000 generation, here the curves last 1/50th of it. In conclusion, I would question whether the parameter 'b' is a solid measurement of 'rate of adaptation'. Also, normalizations makes it difficult to appreciate the result shown in Figure 2B. I think the authors should look for a different way to show the different trend in adaptation dynamics for different glucose concentrations between wild types and mutants. For example, they could move Figure S2C in the main text to stress the result shown in Figure 2C, which already shows the difference between WT and mutant. This is especially true if what Figure 2C shows is (evo-anc)/evo. This is not fully clear to me: in the legend it refers to the delta, in the label of the y-axis I read that this is a percentage.

We thank the reviewer for prompting us to clarify our methods for reporting fitness changes over time. The fitness values are reported, throughout the paper, as a percentage change relative to the reference WT strain. The gain in fitness during evolution (reported as Δ) represents the difference between the evolved strain (evo%) and the ancestral strain (anc%), calculated as Δ = evo% - anc%. This represents the absolute gain, rather than the relative gain. This value is still reported as a percentage as it’s the same scale and unit as the two values being subtracted. We have included additional details to clarify this aspect in the figure legend.

“(C) Absolute fitness gains (Δ) at generation 1000 for evolved WT (upper panel, black) and ctf4Δ (lower panel, orange) populations. Box plots show median, IQR, and whiskers extending to 1.5×IQR, with individual data points beyond whiskers considered outliers. Absolute fitness gains were calculated by subtracting the ancestral relative fitness from the relative fitness of the evolved (Δ = evo% - anc%), both calculated as percentages relative to the same reference strain in the same glucose concentration.”

To conclude: the data show a different trend between wild types and mutants, which is interesting. Fitting it with the power law seems to be neither required nor appropriate. I suggest the authors to show the WT vs mutant pattern differently.

We followed the reviewer’s suggestion and moved Figure S2C, which depicts the detailed fitness trajectories over time, into the main manuscript as Figure 2D. We agree that presenting these trajectories alongside the absolute fitness gains (now in Figure S2C) provides a more intuitive and effective depiction of the evolutionary dynamics of WT and ctf4Δ strains without relying solely on the power-law fit. Additionally, we quantified the mean adaptation rate, calculated as the absolute fitness gain (Δ) divided by the total number of generations (now Figure 2B). While no individual method definitively captures the adaptation rates across the experiment, these complementary analyses consistently highlight the same trends noted by the reviewer. We have re-written the main text as follows:

Line 171: “By generation 1000, both WT and ctf4Δ evolved lines achieved, on average, slightly higher fitness in low glucose compared to high glucose conditions (Fig S2B). However, due to the varying initial fitness of ctf4Δ cells across different glucose environments, they recovered the same extent of the original defect (Fig S2C). ctf4Δ lines displayed an opposite trend to WT, with increasing absolute fitness throughout the experiment as glucose concentration rose (Fig S2B vs S2D). The differint absolute fitness gains over the same number of generations highlight distinct mean adaptation rates (Fig 2B). These differences are evident when examining the evolutionary dynamics of the evolved lines over time (Fig 2C). Additionally, we approximated the fitness trajectories using the power law function (Fig 2C, dashed purple lines), previously proposed to describe long-term evolutionary dynamics in constant environments (Wiser et al., 2013). The parameter b in this formula determines the curve's steepness, and can be used to quantify the global adaptation rate over generations (Fig S2E). Collectively, these analyses demonstrate that, unlike WT cells, ctf4Δ lines adapt faster in the presence of high glucose. This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

Overall, these results demonstrate that cells can recover from fitness defects caused by constitutive DNA replication stress regardless of the glucose environment. However, adaptation rates under DNA replication stress exhibit opposing trends compared to WT cells, with faster adaptation yielding greater fitness gains in higher glucose conditions.”

2- In Figure S2C, the individual trajectories for WT at 2% glucose are strangely variable. In this case, plotting the average does not make too much sense. This result is strange, since this is the default condition, where cells are grown without any change of sugar concentration. Can the authors give any rationale? Are there other available results to replace those published in Figure S2C?

We agree with the reviewer that the individual trajectories for WT at 2% glucose are intriguing. However, we do not find these results necessarily “strange” as they could be explained by the following rationale: WT cells have been cultivated in 2% glucose since the 1950s, likely fixing most beneficial mutations for this condition. When many isogenic strains are evolved in parallel, (a) some lines show no improvement due to the scarcity of available beneficial mutations, (b) others exhibit slight decreases in fitness due to genetic drift fixing deleterious mutations, and (c) a few lines discover rare beneficial mutations, leading to fitness increases. In contrast, other conditions represent “newer” environments with larger mutational target sizes, resulting in more consistent outcomes.

Prompted by the reviewer’s comment, we look for other studies reporting detailed fitness measurements of evolved WT strains in standard laboratory media. We downloaded and plotted the fitness data from Johnson et al. 2021, where authors studied the evolution of WT strains over 10.000 generations. Interestingly, we see that in the early phase of the evolution (generations 500-1400) evolved lines show similar levels of variability in fitness as the one reported in our study (Figure_rev_only 2). Of note is that in Johnson et al. 2021 most of the adaptive mutations alleviate the toxicity of the ade2-1 allele. In our WT strain the gene was preemptively restored, furter reducing the target size for adaptation in YPD.

We believe it is important to report these measurements and decided to leave the original data, with the appropriate quantifications of variability, in Figure 2.

3- The molecular explanation given for the rescue of ctf4-null proposes a very relevant role for dNTPs downregulation. Particularly, both for Irx1 and med14-H919P, the authors propose that this happens via Rnr1 downregulation. At this stage, this is only a hypothesis. The molecular verification of the central role of Rnr1 downregulation would make the conclusion much stronger. For example, a preliminary test would imply that duplicating RNR1 in ctf4-null irx1-null and/or ctf4-null med14-H919P would revert the rescue. Any other experiment addressing this point would be useful to improve the paper.

We agree that the experiment suggested by the reviewer, or similar tests, would substantiate our hypotheses and strengthen the paper. Specifically, we plan to perturb dNTP production in both ctf4Δ ixr1Δ and ctf4Δ med14-H919P mutants through genetic manipulation of known factors involved in dNTP synthesis. We will then compare the resulting fitness to the expectations based on our hypotheses: reduced fitness benefits of the double mutants upon increasing dNTP levels and/or increased fitness in ctf4Δ mutants by decreasing dNTP levels through alternative mechanisms.

4- The authors propose from Figure S4B that the rescue of ixr1-null is less evident at low sugar concentration since both conditions trigger a reduction of dNTPs. I think this is interesting, since it would provide a link between glucose concentration and evolutionary trajectories to adaptation, which is what the authors wanted to study. In particular, one would predict that 0.25% glucose would see less ixr1-null than the other glucose conditions. I could not (was not able to) confute this hypothesis from the data shown in the paper. Likewise, for med14-H919P. If the authors have not tested it, it would be worth trying.

We had reported the appearance and frequency of all ‘core adaptive mutations’ (Figure S6C) but did not explicitly test the likelihood of their appearance under different glucose conditions. Following the reviewer’s suggestion, we have now performed χ2 tests (on the presence or absence of mutations) and ANOVA tests (on their mean frequency) to determine whether any mutation is particularly enriched or depleted in a given glucose environment. At first glance, the results do not support the hypothesis proposed by the reviewer. However, we note that although ixr1 mutants are less beneficial in low glucose than in high glucose, they still confer an 8% fitness advantage, which is likely sufficient to drive clones to fixation. We believe the reviewer’s reasoning is correct but is potentially masked by the still elevated fitness advantage of ixr1 in low glucose.

To better convey the results of this analysis, we have included a visual representation of the presence and frequency of the mutations in Figure 6A, and the results of the χ2 and ANOVA tests in Supplementary File 5. We also comment on the analysis as follows:

Line 314: “Similarly, we did not detect differences in the frequency of occurrence (χ2 tests) or average fractions (ANOVA test) achieved by the mutations in the populations evolved under different glucose environments (Fig 6A, Fig S4C and Supplementary File 5. The presence of all mutations in the final evolved lines correlated with their fitness benefits, suggesting how their selection in all glucose conditions was mostly dictated by their relative fitness benefits, rather than the environment (Fig 6A).”

5- The combination of the four genetic adaptation (Fig 6B) would benefit from an experimental verification to show that the different solutions are not mutually exclusive. This is not obvious: if more than one solution acts by reducing dNTPs, maybe their combined effect is less strong than what measured theoretically. The authors could derive some clones at the end of the experiment and Sanger sequencing some of the four genes, to confirm the co-presence of some of them in the same cell.

The co-occurrence of nearly every combination of the four core adaptive mutations we identified can be inferred from their relative frequencies, as revealed by deep whole-genome sequencing of the evolved populations (Fig. S4C). In these data, we observe populations carrying each pairwise combination of mutations at frequencies exceeding 50%, implying their coexistence. Moreover, many combinations of mutations approach or reach fixation. A particularly striking example is ctf4Δ Population 11, evolved in 8% glucose, where all core adaptive mutations are present at 100% frequency. These findings provide robust evidence that the different adaptive solutions are not mutually exclusive and can coexist within the same genetic background.

Nevertheless, we agree that experimentally verifying the compatibility and fitness of the four genetic adaptations described in Figure 6B (now Fig 6C) would further strengthen our conclusions. To this end, we plan to reconstruct all combinations of mutations observed at high frequency in the final evolved populations. We will then measure their fitness and compare it to that of the evolved populations, as well as to the theoretical expectations based on additivity currently presented in Figure 6C.

Minor points

Figures

• S4B: in the legend it should be explained that it is compared to ctf4D

We now report how the values were obtained in the figure legend:

(D = |anc%|-|reconstraucted%|)

-2A: the color code is not fully clear to me: what does green and blue indicate? higher and lower than 2%?

We apogise for not having included an explicit description of the color code in Figure 2A. Throughout the paper blue refers to glucose starvation (light blue for 0,25%, dark blue for 0,5%), while green refers to glucose abundance (light blue for 2%, dark blue for 8%). We now include a detailed description of the color code when it first appears (Fig 1B) and make sure is properly reported in all figure legends.

• S3A: the authors should show the statistical difference between WT and ctf4-null, which is mentioned as non-existent in p.6

The p value is now represented in Fig S3A

Text

• RNR1 is not really the gene with the highest score in Figure 5D, not even close: can you give a rationale for pin-pointing it (see also main point 3)?

The reviewer is correct. Perturbations of the mediator complex, which regulate the expression of most of RNA PolII transcripts, is expected to result in changes in the expression of a large set of genes. However, our focus on dNTPs and RNR1 is based on the following rationale:

1. Gene Ontology Enrichment Analysis: The downregulated genes in our dataset are enriched for the 'nucleotide metabolism' term, which includes pathways critical for dNTP production and directly linked to DNA replication and repair.

2. Role of RNR1: Among the downregulated genes, RNR1 stands out as it encodes the major subunit of ribonucleotide reductase, the rate-limiting enzyme in dNTP synthesis. This enzyme is essential for DNA replication, and cells experiencing constitutive DNA replication stress, as in our system, are particularly sensitive to changes in dNTP levels.

To make this rationale more explicit to the reader, we are adding the following sentence in the discussion:

Line 404: “Nucleotide metabolism, particularly ribonucleotide reductase, is essential for dNTP production. Given the role of dNTPs in regulating DNA replication and repair, the advantage of med14-H919P mutants in the ctf4Δ background may stem from reduced dNTP levels caused by the perturbed TID domain."

In addition, following the reviewers’ suggestions, we are conducting additional experiments to investigate the role of med14-H919P mutants in enhancing fitness under conditions of constitutive DNA replication stress (See response to reviewer #4). We anticipate that the final revised manuscript will offer further insights into the role of dNTPs or present alternative explanations for the observed phenomena.

• The med14-H919P mutation is observed in 22/48 wells. I guess the authors checked already: are some of these wells close to each other in the plate?

Correct. We took significant precautions in our experimental design to prevent cross-contamination, as outlined in the Materials and Methods section. Specifically, rows of ctf4Δ samples were alternated with rows of WT samples. Daily dilutions were then performed row by row using a 12 channels pipette. This approach ensured that any potential carry-over of cells would result in them being placed in wells containing a different genotype, where they would be eliminated by the consistent use of genotype-specific drugs.

As a result of these measures, we do not observe any distinct pattern of core genetic adaptation corresponding to the plate layout (Figure_rev_only 3). The only exception are mutations in IXR1, which appear in all ctf4Δ strains (albeit with different alleles, see supplementary File 3). Moreover, we reasoned that if a highly fit strain had invaded other wells, all the pre-existing mutations from its lineage would have been detected in those wells. However, apart from the recurrent ixr1 and rad9 mutations, which are also strongly adaptive, we find no evidence of shared mutations in wells carrying the med14-H919P allele (Figure_rev_only 4).

• Compensatory evolution of ctf4-null in 2% glucose is the experiment published by Fumasoni and Murray in eLife. In that paper, there is no trace of mutations in MED14. I think the authors should comment on this (different method for detecting putative compensatory mutations?).

We also noticed the absence of MED14 mutations in the eLife study by Fumasoni and Murray and find this discrepancy intriguing. One possible explanation lies in methodological differences. Our current study employed an improved version of the mutational analysis pipeline. However, we have not yet reanalyzed the original data from the previous study to determine whether MED14 mutations were present but undetected.

Interestingly, in the current study, we observed that in 2% glucose, MED14 mutations arose in only 3 out of 12 populations, a frequency lower than in other glucose conditions (Figure S6C). Assuming a similar frequency occurred in the 8 populations evolved in 2% glucose by Fumasoni and Murray (2020), one would expect only 2 populations to carry the mutation. This number falls below the threshold required for our algorithm to detect statistically significant parallelism.

Additionally, two significant experimental differences may also contribute to the observed discrepancy. First, the culture volumes and vessels differed: 10 mL cultures in tubes were used previously, whereas 1.5 mL cultures in 96-well plates were used in the current study.

• I may be mistaken, but Szamecz et al do not actually investigate whether different conditions result in different evolutionary trajectories (i.e., different genetics), and so their results may not be at odds with those presented here.

The reviewer is correct that Szamecz et al. do not explicitly test whether different conditions result in different evolutionary trajectories. However, in the section titled “Compensatory Evolution Generates Diverse Growth Phenotypes across Environments,” they examine how lines evolved in 2% YPD perform across various environments. They report how in roughly 50% of the cases tested, evolved lines showed either no improvement or even some lower fitness than the ancestor (Figure 5A).

While this could be explained by the accumulation of detrimental non-adaptive mutations in specific contexts, it likely implies that the adaptive strategies compensating for the original mutation in one environment do not confer similar benefits in other environments. This observation contrasts with our findings in Figure 6D, where we demonstrate that the main adaptive strategies provide a consistent benefit across diverse environments, including those with glucose, nitrogen, or phosphate abundance or starvation.

We have now modified the introduction, results and discussion to avoid misleading interpretations:

Line 42: “Szamecz and colleagues examined the evolutionary trajectories of 180 haploid yeast gene deletions over 400 generations (Szamecz et al., 2014). They found that, while fitness recovery occurred in the environment where evolution took place, the evolved lines often showed no improvement over their ancestors in other environments. This suggests that compensatory mutations beneficial in one environment often fail to restore fitness in others.”

Line 327: “A previous study in yeast showed how evolved lines which compensate for detrimental defects of gene deletions in standard laboratory conditions often failed to show fitness benefits compared to their ancestor when tested in other environments (Szamecz et al., 2014). We thus investigated the extent to which the core genetic adaptation to DNA replication stress was beneficial under alternative nutrient conditions.”

Line 422: “What could explain the discrepancies between our results, and previous studies on evolutionary repair highlighting the role of the environment in shaping evolutionary trajectories (Filteau et al., 2015), and the heterogeneous behavior of evolved lines in various environments (Szamecz et al., 2014)?”

typos

p.18, line 564 preformed -> performed

1. 6 line 189 with a strongly skew -> with a strong skew ?

Typos are now corrected in the main text

Reviewer #2 (Significance):

This is a well-done paper that could be of interest for the community of evolutionary biologists, scientists working on metabolism and cell division. It addresses an interesting problem, how metabolism affects compensatory evolution. Among the strengths: experiments are well done, the results are novel, the cross-talk between metabolism and evolutionary repair is intriguing. Among the weaknesses, the fact that the molecular explanations for the observations are only hypothesized and not tested experimentally. This is where the authors could improve the manuscript.

Reviewer #3 (Evidence, reproducibility and clarity):

This paper combines phenotypic and genomic data from an experimental evolution study in yeast to assess how repeatable evolution is in response to DNA replication stress. Importantly, the authors ask whether genotype by environment interactions influence repeatability of their evolved lines. To this end, the authors have constructed an elegant highly-replicated experiment in which two yeast genotypes (WT and CTF4 KO) were evolved under a variety of glucose levels for 1,000 generations. Recurrent mutations are found across many replicates, suggesting that repeatability is robust to GxE interactions. Of course, the authors correctly identify that these results are dependent on many particulars, as is always the case in biology, but provide a comprehensive discussion to accompany their results. I do not have any major comments to give, but simply some suggestions and points of clarification.

Major comments: N/A

Minor comments:

L19: I found the definition for compensatory evolution/mutations to be somewhat vague in the introduction (and subsequently throughout the text). It's clear that this was written for a more medical/physiological audience, but without a more explicit explanation of compensatory evolution/mutations, it became difficult to properly weigh some claims/discussions made by the authors later on. Do you define compensatory mutations as those which completely recover WT function/fitness, or are simply of opposite effect to the altered genotype? Others define "compensatory evolution" as simply any epistastically interacting amino acid substitutions (Ivankov et al, 2014). It would be nice to see more explicitly defined.

We thank the reviewer for highlighting the need for a precise definition of compensatory evolution and compensatory mutations. We recognize that the literature encompasses multiple definitions, including the one cited by the reviewer, which emphasizes compensatory mutations within the context of structural biology. This particular definition, prevalent in molecular evolution, was introduced by Kimura (Kimura, 1985) and is frequently used to explain the co-occurrence of amino acid mutations within a protein. These mutations offset each other’s defects, restoring or maintaining protein function. Here, however, we are using an older and broader definition of compensatory mutation, first introduced by Wright (Wright, 1964, 1977, 1982) and frequently used in evolutionary genomics (e.g., Moore et al., 2000; Szamecz et al., 2014; Rajon and Mazel, 2013; Eckartt et al., 2024). This definition includes any mutation in the rest of the genome that compensates (fully or partially) for another mutation's detrimental effects on fitness.

We have now included this definition in the introduction:

Line 19: “Compensatory evolution is a process by which cells mitigate the negative fitness effects of persistent perturbations in cellular processes across generations. This adaptation occurs through spontaneously arising compensatory mutations anywhere in the genome (Wright, 1964, 1977, 1982) that partially or fully alleviate the negative fitness effects of perturbations (Moore et al., 2000). The successive accumulation of compensatory mutations over evolutionary timescales progressively repair the cellular defects, ultimately restoring fitness.”

Line 361: “Our findings demonstrate that while glucose availability significantly affects the physiology and adaptation speed of cells under replication stress, it does not alter the fundamental genome-wide compensatory mutations that drive fitness recovery and evolutionary repair.”

Along these lines, I would have liked to see a more direct comparison/discussion of the degree to which deletion lines recovered. I can see from Fig 2E and Fig S2B that fitness increased quite a bit; would it not be possible to include a figure on the degree of compensation (basically relative fitness of evolved deletion lines - relative fitness of ancestral deletion lines)?

If the reviewer is suggesting calculating the difference between the evolved and ancestor fitness, the data is already in Figure S2B and S2D, defined as ‘Absolute fitness gains Δ’ and calculated as Δ = evo% - anc%.

If instead is suggesting to plot the fitness of evolved deletion lines (Y axis) against the relative fitness of ancestral deletion lines (X axis), we have now produced the plot is Figure S2F.

To better understand the extent of the fitness recovery in Ctf4 strains, we have also calculated and plotted the ‘relative fitness gain’ calculated as |evo%| / |anc%| *100 (Figure S2C)

We are now commenting on these comparisons in the following paragraph:

Line 171: “By generation 1000, both WT and ctf4Δ evolved lines achieved, on average, slightly higher fitness in low glucose compared to high glucose conditions (Fig S2B). However, due to the varying initial fitness of ctf4Δ cells across different glucose environments, they recovered the same extenct of the original defect (Fig S2C), displaying an opposite trend to WT, with increasing absolute fitness throughout the experiment as glucose concentration rose (Fig S2B vs S2D). The differint absolute fitness gains over the same number of generations highlight distinct mean adaptation rates (Fig 2B). These differences are evident when examining the evolutionary dynamics of the evolved lines over time (Fig 2C). Additionally, we approximated the fitness trajectories using the power law function (Fig 2C, dashed purple lines), previously proposed to describe long-term evolutionary dynamics in constant environments (Wiser et al., 2013). The parameter b in this formula determines the curve's steepness, and can be used to quantify the global fitness change over generations (Fig S2E). Collectively, these analyses demonstrate that, unlike WT cells, ctf4Δ lines adapt faster in the presence of high glucose. This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

L57: Another minor nitpick that just comes down to semantics. When discussing "96 parallel populations", it invokes a higher sense of replication than is actually present in the study. I would rephrase this to something along the lines of "12 replicate populations across 8 treatments under conditions of [...]".

We changed the sentence as follows:

Line 66: “We evolved 96 parallel populations of budding yeast, organized into 12 replicate lines, across four conditions of glucose availability (from starvation to abundance) with or without replication stress.”

L185-187: The wording here needs to be clarified. Be explicit in that are examine the ratio (or count) of synonymous to non-synonymous mutations here, otherwise the interpretations appears to be direct contradiction to the (as written) results. Only after viewing the supplemental figure was I able to figure out what exactly was meant here.

We changed the sentence as follows:

Line 212: “We found no significant differences in the numbers of synonymous mutations detected in evolved populations in WT and ctf4∆ populations (Fig. S3A). These results support the hypothesis that replication stress in ctf4∆ lines favors the retention of beneficial mutations, rather than simply increasing the overall mutation rate.”

L349-350: The authors observe higher rates of adaptation in deletion lines than WT lines, and discuss this in adequate detail. Although not explicitly mentioned, this is consistent with a diminishing returns epistasis model (that could be beneficial to discuss, but is not necessary), which has been implicated in modulating the degree of repeatability observed along evolutionary trajectories (Wünsche et al. 2017). Although definitely not required for this already very nice manuscript, I think it would be very rewarding if the authors were to eventually analyze fine-scale dynamics of phenotypic and genomic adaptation to mine for these putative interactions and their influence on repeatability.

We agree with the reviewer on how our results align with a model of diminishing returns epistasis. This pattern is apparent not only between ctf4Δ and WT lines but also among ctf4Δ lines evolved in different glucose conditions. This phenomenon likely arises from the interaction of various adaptive mutations, which we aim to explore further in a dedicated manuscript. However, until we do so, we prefer to refer generally to a pattern of declining adaptability. To explicit this trend we have now included Fig S2F and commented on it in the manuscript:

Line 181: “This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

Line 388: "Our results are consistent with declining adaptability, as evidenced by the reduced rates of adaptation observed both between ctf4Δ and WT lines and among ctf4Δ lines evolved in different glucose conditions (Fig S2F)"

Reviewer #3 (Significance):

It is clear to me that a great deal of time and care has been put into this study and the preparation of this manuscript. The science and analyses are appropriate to answer the questions at hand, and it bodes well that whenever I had a question pop up while reading, they were typically answered immediately after. I think that this manuscript will be broadly relevant to both biologists both evolutionary and clinical, and was written in a way to be accessible to both.

As someone with an expertise in repeatable evolution, I felt most excited by the observation of so many parallel substitutions at a single amino acid across deletion lines. As the authors rightfully point out in the results and discussion, it's likely that this degree of robustness is highly dependent on the particular mechanism of disruption that cells experience. The authors then go above and beyond to functionally validate the putative molecular mechanisms of (repeatable) adaptation in this system. While it may not always be possible to accomplish in non-model organisms, such multi-modal approaches will be crucial to advance the field of repeatable evolution.

Reviewer #4 (Evidence, reproducibility and clarity):

The authors investigated the effects of DNA replication stress on adaptation in different nutrient availabilities by passaging wild-type and ctf4Δ Saccharomyces cerevisiae in media with varying levels of glucose over ~1000 generations. The ctf4Δ strain experiences increased DNA replication stress due to the deletion of a non-essential replication fork protein. The authors found differences in evolution between wild-type and ctf4Δ yeast, which held across different growth media. This study identified a compensatory single amino acid variant in Med14, a protein in the mediator complex of RNA polymerase II, that was specifically selected in ctf4Δ strains. The authors conclude that while environmental nutrient availability has implications for cell fitness and physiology, adaptation is largely independent and instead dependent on genetic background. The data provide excellent support for the key aspects of the models, although some details are (to me) overstated.

Major comments:

• A ctf4Δ mutant strain was used to investigate the effects of replication stress. Why was this mutant chosen instead of other deletions that cause different types of replication stress?

We appreciate the opportunity to clarify our rationale for choosing the ctf4Δ mutant. The following are the main reasons why we believe ctf4Δ strains represent an ideal tool to study a global perturbation of the DNA replication program over evolutionary timescales:

1. General replication stress: The absence of Ctf4 perturbs replication fork progression, leading to a spectrum of replication stress-related phenotypes, including DNA damage sensitivity, single-stranded DNA gaps, reversed forks (Abe et al., 2018; Fumasoni et al., 2015), checkpoint activation (Poli et al., 2012), cell cycle delays (Miles and Formosa, 1992), increased recombination (Alvaro et al., 2007), and chromosome instability (Kouprina et al., 1992). This broad disruption makes it an excellent model for observing global perturbations in replication processes. In contrast, other mutants typically affect specific enzymatic (e.g., POL32 and RRM3) or signaling (e.g., MRC1) functions, making them better suited to address specific questions.
2. Constitutive stress: Unlike drug-induced stress (e.g., Hydroxyurea; Krakoff et al., 1968) or conditional depletion systems (e.g., GAL1-POLε; Zhang et al., 2022), which cells can easily circumvent through single mutations, ctf4Δ enforces persistent replication stress. Its deletion cannot be complemented by a single mutation, ensuring a robust and consistent stress environment for evolutionary studies.

We have now modified the main text to convey these advantages in a concise form:

Line 91: “In the absence of Ctf4, cells exhibit multiple defects commonly associated with DNA replication stress, such as single-stranded DNA gaps and altered replication forks (Fumasoni et al., 2015), leading to basal cell cycle checkpoint activation (Poli et al., 2012). These defects result in severe and persistent growth impairments, cell cycle delays, elevated nucleotides pools and chromosome instability (Miles and Formosa, 1992; Kouprina et al., 1992; Poli at al., 2012), making ctf4Δ mutants an ideal model for studying the cellular consequences of general and constitutive replication stress over evolutionary time.”

It's not clear from the study that the effects are generalizable to other forms of replication stress.

As with any method to induce DNA replication stress (including commonly used drugs like HU) each approach inevitably affects replication in a specific manner. Testing the broader applicability of our conclusions would require evolving additional strains with different replisome perturbations. For instance, mutations in ELG1 and CTF18 (affecting the alternative Replication Factor C), POL30 (affecting the sliding clamp PCNA), POL32 (affecting Polε), RRM3 (protective helicase) and (MRC1 (coordinating leading strand activities and signalling to the checkpoint) would have to be taken into account. Furthermore, specific mutant alleles of Ctf4 that disrupt interactions with particular binding partners (Such as ctf4–4E and ctf4–3E, perturbing the interaction with the CMG helicase and accessory factors respectively) will be highly informative on which specific aspects of the replication stress generated by the lack of Ctf4 each adaptive mutation alleviate.

However, accommodating such extensive variability would inflate the sample size to an extent that will become unfeasible within the experimental design focused on capturing parallel evolution over a nutrient gradient (the primary focus of this study). We agree that this is an important question and intend to address it comprehensively in a dedicated future study.

• The authors could be clearer that a (the?) cause of the ctf4∆ fitness defect is spurious upregulation of RNR1. I don't think it is mentioned until the Discussion, but it is highly relevant to Fig 4, and to the adaptations one would expect from ctf4∆.

We thank the reviewer for the opportunity to clarify this aspect. We do not think that the fitness defects of ctf4∆ cells stem solely from the spurious upregulation of RNR1. However, we believe that a major aspect of the evolutionary adaptation is aimed at decreasing dNTP levels, potentially through different mechanisms. We are now mentionig increased dNTPs as major phenotype of ctf4∆ and commenting on the hypothesis more clearly in the discussion.

Line 93: “These defects result in severe and persistent growth impairments, cell cycle delays, elevated nucleotides pools and chromosome instability (Miles and Formosa, 1992; Kouprina et al., 1992; Poli at al., 2012)”

Line 409: “This condition will, in turn, be detrimental when proliferation rates are high (as in WT in high glucose) but beneficial under constitutive DNA replication stress (ctf4Δ), where cells experience spurious upregulation of dNTP production (Poli et al., 2012; Davidson et al., 2012).

• In Figure 1E, there is a very large spread in the relative fitness at 2% and 8% glucose, but this was not commented on. Is this heteroscedasticity expected?

The observed heteroscedasticity is expected. Our competition assays tend to exhibit increased variability when a strain approaches very low fitness levels. Specifically, as one strain nears extinction by the third day of competition, its abundance is estimated based on a much smaller number of events in the flow cytometer. Furthermore, we noticed a small number of reference cells carrying pACT1-yCerulean not showing strong fluorescence in 8% glucose. The nature of this effect is uncertain, and possibly linked to metabolism-linked changes in the cytoplasm. The combination of these two phenomena amplifies the impact of noise inherent to the methodology, leading to increased variability across replicates.

Nontheless, the overall decreasing fitness trend across glucose conditions, combined with the statistical significance observed between high and low glucose levels, collectively convey a roboust phenotype

• The med14-H919P mutant was highly selected in ctf4Δ strains, independent of glucose availability. Is this variant found in any natural yeast strains (i.e., are there environments that select for this variant)? Also, if this variant is found in natural strains, does it co-occur with other mutations that could affect DNA replication?

We agree that this is an intriguing question. To address it, we plan to explore existing databases of variants identified in S. cerevisiae natural isolates. Specifically, we will investigate whether the med14-H919P mutation is present in these strains, identify any potential environmental factors that may select for it, and assess whether it co-occurs with other mutations that could influence DNA replication processes.

• The statement on lines 271-273 is not particularly well-supported. The analysis of the Warfield data suggest that reduced expression of RNR1 could be causal, but the data don't go as far as showing how the med14 mutation is advantageous in ctf4∆. Further experimentation would be necessary to support the possibilities that the authors discuss.

The sentence the reviewer refers to is: “Overall, these results show how an amino acid substitution in the Med14 subunit of the mediator complex, putatively affecting transcription, is strongly selected, and advantageous, in the presence of constitutive DNA replication stress.” We are unsure which aspect of the statement is seen as unsupported. The mutation's strong selection in ctf4∆ is demonstrated in Figures 5A, 6A, and S4C, while its advantageous nature is supported by Figures 5B and S4B. Regarding the mechanism, we have been cautious with our phrasing, describing its effect on transcription as "putative" (Line 272) and suggesting that our observations “are compatible with” reduced dNTP availability in med14-H919P cells due to RNR1 downregulation (Line 361).

The main focus of this study is to explore how nutrient availability influences evolutionary dynamics and compensatory adaptation in cells lacking Ctf4. We believe the identification of a novel selected allele (Fig. 5A) and confirmation of its benefit across glucose conditions (Fig. 5B) serves as an excellent complement to the primary conclusions (present in the title). We invite the reviewer to consider that the molecular basis of such a phenotype is not mentioned in our abstract, as we believe that its precise characterization would require a dedicated study on Med14.

Nonetheless, we are encouraged by the reviewer’s interest in this newly identified compensatory mutant (also noted by Reviewer #2), and we are eager to perform further experiments to better understand the biological processes affected by this mutation. We plan to extend our work as follows:

Based on known phenotypes associated with perturbations of Med14, we propose the following novel hypotheses regarding the mechanism by which med14-H919P alleviates ctf4Δ defects:

1. Decreased replication-transcription conflicts: Conflicts between the transcription machinery and replication forks are known to cause fragile sites, leading to increased chromosome breaks and genomic instability (Garcia-Muse and Aguilera, 2016). A general reduction in PolII transcription during replication, resulting from perturbations of the mediator complex, could reduce these conflicts and mitigate the fitness defects observed in ctf4Δ cells.
2. Increased cohesin loading: We have demonstrated that amplification of the cohesin loader SCC2 is beneficial in the absence of Ctf4. Recent findings (Mattingly et al., 2022) indicate that the mediator complex recruits SCC2 to PolII-transcribed genes. The med14-H919P mutation may enhance the fitness of ctf4Δ cells by facilitating cohesin loading during DNA replication.
3. Decreased dNTP levels: As discussed in the manuscript, perturbations of Med14 subunits in the mediator complex reduce the expression of genes, including those associated with nucleotide metabolism. Notably, these include RNR1, the major subunit of ribonucleotide reductase. The med14-H919P mutation could benefit the ctf4Δ background by counteracting the reported spurious increase in dNTPs, which affects replication fork speed (Poli et al., 2012).

We plan to distinguish between these hypotheses using the following approaches. First, the proposed mechanisms underlying Hypotheses 1 and 3 suggest that med14-H919P is a loss-of-function mutation, while Hypothesis 2 implies a gain-of-function effect. Testing the impact of a heterozygous med14-H919P allele in a homozygous ctf4Δ strain will allow us to differentiate between these two categories of mechanisms. Additionally, we aim to investigate the molecular process affected by the med14-H919P allele by analyzing its genetic interactions with genes involved in replication-transcription conflicts, cohesin loading, and dNTP production (See also response to reviewer #2).

We believe that the results of these experiments will provide further insights on the mechanism of suppression exerted by med14-H919P in the presence of constitutive DNA replication stress, without diverting the reader from the main message of the paper.

• The authors comment that the med14-H919P mutant could have implications for the stability of Med14, based on computational modelling. Verifying the stability of the med14-H919P in vivo would strengthen this discussion.

We believe that in vivo and in vitro structural studies investigating the effect of this mutation on the stability and function of the Mediator complex are beyond the scope of this manuscript. These investigations would be more appropriately addressed in future, dedicated studies focused on these specific aspects.

• In the discussion, the authors propose that the context of the perturbation may influence the robustness of adaptation. A more detailed explanation of this point (including a discussion of the findings of other similar studies investigating different conditions) would be helpful to further bolster this section.

We are now supporting this concept more explicitly by commenting on other studies as follows:

Line 429: “Third, the environment’s influence on compensatory evolution may depend on the specific cellular module perturbed and its genetic interactions with other modules that are significantly influenced by environmental conditions. For example, the actin cytoskeleton, which must rapidly respond to extracellular stimuli, is likely to be more directly influenced by environmental factors (Filateau et al., 2015) compared to the DNA replication machinery, which operates within the nucleus and is relatively insulated from such changes. Supporting this idea, a study examining mutants’ fitness across diverse environments found that conditions such as different carbon sources or TOR inhibition, similar to those used in this study, primarily affected genes involved in vesicle trafficking, transcription, protein metabolism, and cell polarity. In contrast, genes associated with genome maintenance, as well as their epistatic interactions, were largely unaffected (Costanzo et al., 2021)”.

In addition, to further substantiate this hypothesis, we plan to re-analyze published datasets on fitness and epistatic interactions among genes in various environments, testing whether specific cellular modules are more prone to changes following shifts in nutrient conditions.

Minor comments: - Competitions were performed between ctf4Δ strains and a constructed strain with yCerulean integrated at ACT1. Is the fitness of the fluorescent strain comparable to the ancestral wild-type strain (i.e., in a competition between the ancestral WT and the fluorescent strain, does either have an advantage)?

We noticed a slight disadvantage of the reference strain compare to WT, likely due to the costs of the extra fluorescence reporter. However, the disadvantage is minimal, ranging from -0.5 to -2.5 depending on the glucose environment (raw measurments are reported supplementary file 1, sheet 5). To take this into account, all fitness reported in figures are normalized for the WT value measured in the same environment line 613: “Relative fitness of the ancestral WT strain was used to normalize fitness across conditions.​​”

• In Figure 3, the legends for panels B and C appear to be swapped. Discussion of Figure 3 on pages 6 and 7 appear to reference the wrong panels.

We are unsure about this typo. Main text and figure legend seem to refer to the appropriate panels, 3B for mutation fractions and 3C for mutation counts. Perhaps the organization of the panels with B being under A instead of on its right confounds the reader?

• In Figure 4A and B, having the same colour scale between both heatmaps is misleading, as the scales are different. Consider having the same scale across both heatmaps so that enrichments are visually comparable.

Following the reviewer’s suggestion we have have chosen a uniform heatmap to visually represent GO terms enrichment in WT and ctf4∆ genetic backgrounds.

• In Figure 4C, having a legend in the figure for node size would be helpful to understand the actual number of populations with mutations in each gene.

A legend for node size has now being added next to Figure 4C.

Reviewer #4 (Significance):

In this study, a high-throughput evolution experiment uncovered the effects of genetic background on the development of adaptive mutations. The authors were able to identify a single amino acid variant of Med14 (med14-H919P) that was positively selected in ctf4Δ. Furthermore, they demonstrated the causality of med14-H919P in conferring a fitness advantage in ctf4Δ. The novelty of this mechanistic finding opens future avenues of investigation regarding the interaction network of the mediator complex in conditions of DNA replication stress. A limitation of the study is that only one mechanism of replication stress was assessed (ctf4Δ). Other gene mutations that cause replication stress would be interesting to assess and would provide a more thorough investigation of the effects of DNA replication factors on evolvability. This work will be of interest to researchers in the population genetics and genotype-by-environment fields, as it suggests the robustness of evolvability to environmental factors in the specific condition of DNA replication stress. As discussed by the authors, this finding differs from other works that have linked environmental conditions to adaptive evolution to different conditions, and is concordant with work that indicates the robustness of genetic interactions to environmental stresses. Furthermore, the identification of the highly-selected med14-H919P variant will be of interest to the DNA replication field. There is the potential for future work investigating the role of Med14 in mediating the response to DNA replication stress in both yeast and mammalian cell contexts, since the authors note that there are links between altered mediator complex regulation and cancers. Although I suspect that the very different regulation of RNR in mammalian cells makes it unlikely that the kind of upregulation of dNTP pools seen in ctf4∆ would be induced by replication stress in mammalian cells.