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)): __ In this manuscript, the authors describe the discovery of a molecular regulator of the immune transcriptional program, which is activated by intestinal distension upon bacterial colonization of the C. elegans intestine. Taking advantage of the fact that inhibition of aex-5 is known to cause intestinal distension and a C-type lectin gene clec-60 as a marker for the immune response to intestinal distension (clec-60p::gfp), the authors performed a forward genetic screen for suppressors of the immune response activation. Of the two mutants isolated, they focused on the stronger suppressor, which corresponded to a cysteine-type DUB, the Ubiquitin Specific Peptidase-14 (usp-14). Through rescue experiments, phenocopy analyses, and quantitative RT-PCR, they validated usp-14 as the causal gene and initiated characterization of its role in immune response activation. To this end, the authors investigated the tissue of action, identifying the intestine as the tissue in which usp-14 mediates the regulation of the immune response. Through transcriptomic analyses, they found that the signalling pathway likely regulated by usp-14 in response to intestinal distension is the Wnt pathway, as they have observed reduction in the transcriptional level of some of the Wnt pathway components in usp-4(tm1481), in response to infection with S. aureus. Additionally, transcriptomic data indicate that usp-14 plays a role in immunity regulation even in the absence of infection. Based on these findings, the authors propose that usp-14 has a dual role in immune regulation: one in surveillance immunity, preventing overactivation of immune responses, and another as a mediator of pathogen-induced responses, such as those triggered by P. aeruginosa or S. aureus. The experiments are rigorous and the results robust; however, some points would benefit from further investigation or clarification. __Response: We thank the reviewer for an excellent summary of our work and for the valuable feedback.

Comment: The expression domain of usp-14 appears to be quite expanded based on single cell RNAseq data (e.g. PMID: 28818938) therefore it is likely that the transgenes used for expression analysis are lacking key regulatory information. Alternative methods like smFISH would be more appropriate to characterise the spatiotemporal pattern of usp-14 expression in more detail. Response: We thank the reviewer for this valuable suggestion. In the original version of the manuscript, we used a 714 bp region upstream of the usp-14 start codon to generate the transcriptional reporter. In the revised manuscript, we reconstructed the reporter using a longer 1924 bp upstream promoter region together with a portion of exon 1. Using this updated reporter, we observed substantially broader expression of usp-14, particularly during the early larval stages. These results are described on page 6, lines 148-153: “We next examined the spatiotemporal expression pattern of usp-14 in C. elegans. To this end, we generated transgenic worms expressing GFP under the control of the usp-14 promoter (usp-14p::gfp). During early larval development, usp-14 was broadly expressed across multiple tissues (Figure 3A). However, in L4 larvae and adult animals, expression became more restricted and was predominantly observed in the intestine and a subset of neuronal cells. Notably, both intestinal and neuronal expression persisted throughout development (Figure 3A).”

Comment: __The mutation mapped in usp-14(jsn19) is a missense mutation (E122K) that suppresses the immune response to a degree comparable to the usp-14(tm1481) deletion allele. However, the authors do not show the functional domains in Fig. 1E potentially affected by this missense mutation. __Response: We have now updated Figure 1E to include the functional domains of USP-14 and mapped both the usp-14(jsn19) missense allele and the usp-14(tm1481) deletion allele onto the protein schematic.

Comment: __How USP-14 regulates Wnt and how Wnt signalling relates to activation of immune responses is not fully supported. Are the Wnt components mentioned in the study induced specifically in the intestine upon infection and does USP-14 act in the intestine in the context of this regulation? How do the authors interpret that both Wnt ligands and receptors are induced ? Does Wnt signalling appear as a GO term in the transcriptomic analysis? The authors can include Wnt signalling components in the analysis of the transcriptomic results. __Response: We thank the reviewer for these insightful comments. Previous studies have shown that the Wnt pathway components examined in our study are induced in the intestine upon infection and function within the intestine to regulate host defense against bacterial pathogens (PMID: 29768179; PMID: 36323254).

We did not observe significant enrichment of Wnt signaling terms in the GO analysis of our transcriptomic dataset. We believe this is likely due to the stringent thresholds used for differential expression analysis (fold change > 2 and p At present, the precise mechanism by which USP-14 regulates Wnt pathway components remains unclear. One possibility is that USP-14 influences Wnt signaling indirectly through additional substrates or interacting proteins that regulate transcriptional outputs. We have now clarified this point in the Discussion (page 13, lines 344–349): “These observations raise the possibility that additional USP-14 substrates or interacting proteins modulate transcriptional outputs downstream of intestinal distension. Future studies aimed at identifying the direct substrates of USP-14 and defining how USP-14 interfaces with neuronal ACC-4 signaling and other distension-responsive pathways will provide important mechanistic insight into how intestinal distension is coupled to innate immune activation.”

Regarding the simultaneous induction of Wnt ligands and receptors, we interpret this as a potential amplification or reinforcement mechanism that enhances Wnt/β-catenin signaling during infection-induced intestinal distension. However, further studies will be required to determine the mechanistic significance of this coordinated transcriptional regulation.

Comment: __Overall, in most of the figures, the micrographs are in general quite dark and exhibit poor contrast between signal and background, particularly in Fig. 1, panels B and J, and Fig. 2, panels B and F (upper rows). Even though these panels are intended to show absence of response, the outlines of the worms are difficult to discern. __Response: We thank the reviewer for the feedback. We have now improved the image presentation throughout the manuscript by either increasing the intensity or adding dotted outlines to more clearly indicate worm positions.

Comment: __In Figure S3, panels A and B, the pmk-1(km25); usp-14(tm1481) animals subjected to aex-5 RNAi show some level of fluorescence/response induction comparable to pmk-1(km25) alone. This observation is not discussed in the text. __Response: We have now discussed this observation in the text. These results are described on page 9, lines 244-248: “Although pmk-1(km25);usp-14(tm1481) worms displayed relatively higher GFP levels than usp-14(tm1481) single mutants upon aex-5 RNAi treatment, this effect likely reflects the elevated basal GFP expression observed in pmk-1(km25) mutants (Figure S4B). Importantly, pmk-1(km25);usp-14(tm1481) animals still exhibited significantly lower GFP levels than pmk-1(km25) single mutants.”

Reviewer #1 (Significance (Required)): __ __Comment: __The work is interesting because it expands some previous work in the field demonstrating immune response induction as a consequence of intestinal distension even in the absence of bacterial infection. This is known to be mediated by the neuronal acetylcholine receptor ACC-4, which signals to the intestine where it regulates immune genes via the Wnt pathway. However, how USP-14 relates to ACC-4 is currently unclear and whether USP-14 function is really required in the intestine to control Wnt signalling is not demonstrated. The authors should include a model to describe how their findings relate to the previous literature and how USP-14 may link mechanistically to Wnt signalling pathway activation. __Response: We thank the reviewer for this insightful comment. We agree that the relationship between USP-14, ACC-4, and Wnt signaling requires further clarification. As suggested by the reviewer, we have now included a model summarizing the current understanding of intestinal distension-induced immune activation and integrating our findings with previous literature (Figure 6H).

Comment: __It remains also unclear whether usp-14 is the only deubiquitinase involved in intestinal distension-induced signalling via the Wnt pathway, or whether other paralog usp genes might also contribute to regulation of immune-responsive transcription. Notably, several mammalian deubiquitinases have established roles in cancer suppression and inflammatory response and innate immunity in other systems so this would increase the potential significance of the work. __Response: We thank the reviewer for this valuable suggestion. To systematically examine whether additional DUBs contribute to intestinal distension-induced immune activation, we performed an RNAi screen targeting all DUBs available in the Ahringer RNAi library using the aex-5(sa23);clec-60p::gfp reporter strain. Among the DUBs tested, knockdown of usp-14 produced the strongest suppression of clec-60p::gfp expression. Although knockdown of usp-5 also partially suppressed GFP induction, usp-5 RNAi did not affect survival during P. aeruginosa infection, suggesting that usp-5 is not required for host defense under these conditions. Together, these findings identify USP-14 as the major DUB required for intestinal distension-induced immune activation in our experimental system. These results are now included in Figure 1G, H, and Figure S2.

Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ Summary C. elegans are soil-dwelling nematodes that feed on bacteria and fungi and thus must be able to distinguish between innocuous and pathogenic species of microbes to survive. Though they lack adaptive immunity, these animals have an ancient version of an innate immune system that has no circulating sentinel or phagocytic cells yet can still mount a response to pathogen exposure. A consequence of the mode of infection of some ingested bacterial pathogens is intestinal distension which by itself, even in the absence of pathogens, is sufficient to trigger the expression of genes encoding immune effectors, including proteins that are bactericidal. The complete mechanistic scheme connecting intestinal distension to the expression of immunity genes has not been resolved, motivating the authors to perform a forward genetic screen for additional components of this pathway. One mutant that the authors isolated was usp-14, encoding an evolutionarily conserved deubiqutinating enzyme. Functional analysis revealed that usp-14 confers protection from microbial pathogens and that the intestine is its primary site of action for its role in host defense. The authors' data indicate that while USP-14 regulates the expression of innate immunity genes that are induced by intestinal distension, surprisingly it functions independently of several canonical innate immune signaling pathways, including the pmk-1/p38 MAPK pathway. Instead, USP-14 appears to act through Wnt signaling to regulate immune effectors by upregulating the expression of several components of that pathway, including the C. elegans ß-catenin ortholog bar-1. This places usp-14 within a gut-brain axis previously shown to control the C. elegans innate immune response through acetylcholine-mediated activation of Wnt signaling. The authors' findings provide new mechanistic insight to this pathway and add to the understanding of ubiqutination as an immune regulatory module. __Response: We thank the reviewer for providing an excellent summary of our work.

Major comments __1. There are three types of experiments in which the authors use the same set of controls across several different figure panels, as stated in the legend to Figure 2. First, when quantifying GFP levels of clec-60::gfp in RNAi-treated animals, the authors use the same clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls for Fig. 1K, 2C, and 2G. For infection assays with S. aureus NCTC8325, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2E are the same as the ones used in Fig. 1M. Similarly, for infection assays with P. aeruginosa PA14, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2I is the same as was used for Fig 1I. In each case, if the authors in fact collected all of the data for each strain that they studied at the same time but then chose to parse larger datasets into separate figure panels to make it more clear to the reader, then this approach is valid but the authors need to explicitly state that this is what they did. However, if the data pertaining to the control strains were collected at a different time or if it comes from a separate biological replicate, then re-using data from the controls is not appropriate because it would not accurately reflect the specific conditions of the experiment to which the data are being compared. If this is indeed the scenario, then the authors will need to repeat these experiments and include the appropriate control in each iteration. __Response: While preparing the manuscript, these experiments were performed simultaneously. Therefore, all panels that share controls have results from experiments performed simultaneously and represent the same biological replicate. We have added this additional information in the relevant figure legends.

Comment: __2. From the legends describing figure panels that include data pertaining to clec-60p::gfp expression levels as assessed by fluorescence microscopy it seems that, in general, the authors measured GFP fluorescence in about 30 animals to produce quantitative data. How many biological replicates of these types of experiments were carried out? This is not explicitly stated in the section describing fluorescence imaging in the Methods section. Following the description of their methodology regarding statistical analysis of survival curves from microbial infection assays, however, the authors state that, "[a]ll experiments were performed independently at least three times unless otherwise noted." Does this statement apply to microscopy or only to experiments involving infection assays? If the data reporting quantitation of GFP signal is based on only 30 animals, then additional biological replicates are necessary, along with appropriate statistical analyses. __Response: The quantified GFP fluorescence data are derived from three independent biological replicates. In each experiment, we typically imaged and quantified approximately 10 worms per condition, yielding a total of ~30 worms analyzed per genotype or treatment across all replicates (except Figure S1B, where we had two independent replicates). We have added the number of experiments in the figure legends for these data.

Comment: __3. The authors have made all of the RNASeq data publicly available on the Sequence Read Archive, and they include data from several pairwise comparisons for differential gene expression analysis in their supplemental files. One of the most important facts to come out of the authors' Gene Ontology analyses of their RNASeq data is that the genes that are upregulated in a usp-14-dependent manner upon intestinal distension are enriched for those whose products play a role in innate immunity/host defense. The authors should say more about these genes. Are there any commonalities between them with regard to function? Are any of them targets of transcription factors that are known to function in C. elegans innate immunity? If so, this could provide clues as to what the substrates of USP-14 might be. Importantly, the specific identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense should be revealed in a supplemental file, and designated as being dependent on or independent of usp-14 for their expression during intestinal distension. __Response: We thank the reviewer for this insightful suggestion. We have now expanded the Results section to describe the functional categories enriched among the USP-14-dependent intestinal distension-induced immune genes, including C-type lectins, ShK toxin domain-containing proteins, and lysozymes (page 7, lines 194-196).

In addition, we compared our transcriptomic dataset with previously published transcription factor-regulated gene sets using WormExp analysis and identified a substantial overlap with genes regulated by the GATA transcription factor ELT-2. These new analyses are described on page 7, lines 197-210: “To identify transcription factors potentially involved in intestinal distension-induced immune activation, we performed transcription factor enrichment analysis using WormExp on genes upregulated in N2 worms following aex-5 RNAi treatment. This analysis revealed a substantial overlap between aex-5 RNAi-induced genes and genes regulated by the GATA transcription factor ELT-2 (Figure S3D). We next examined whether USP-14-dependent immune genes overlapped with ELT-2-dependent immunity genes induced by intestinal distension. To this end, we identified innate immune genes common to both ELT-2-regulated gene sets and aex-5 RNAi-induced genes. Strikingly, these ELT-2-dependent intestinal distension-induced immune genes showed substantial overlap with USP-14-dependent immune genes (Figure S3E and Table S5), suggesting that USP-14 may regulate distension-induced immunity, at least in part, through ELT-2-dependent transcriptional programs. Consistent with this possibility, RNAi-mediated knockdown of elt-2 did not further increase the susceptibility of usp-14(tm1481) worms to P. aeruginosa infection relative to wild-type worms (Figure S3F), supporting a model in which USP-14-mediated immune responses require ELT-2 activity.”

Finally, we have created a new table (Table S5) that specifies the identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense, for USP-14-dependent and independent genes.

Comment: __4. The authors' data suggest that in response to bacterial infection USP-14 upregulates the expression of bar-1, along with other components of the Wnt signaling pathway, which in turn upregulates innate immunity genes. This could be further substantiated by directly demonstrating that there are USP-14-regulated innate immunity genes whose induced expression in the presence of microbial pathogens also requires bar-1. Along those lines, an initial test would be to assess clec-60p::gfp expression in bar-1 animals versus bar-1;usp-14 double mutants, similar to the experiment whose results are reported in Fig. S4. If generating the bar-1;usp-14 double mutant is not feasible, then RNAi could be used to knockdown bar-1 expression in clec-60p::gfp;usp-14(tm1481) animals. To expand this analysis, the expression of the six innate immunity genes shown to be regulated upon intestinal distension in usp-14-dependent manner could be measured in the presence and absence of intestinal distension or microbial infection in bar-1 and bar-1;usp-14 animals by qRT-PCR. At a minimum, the authors should conduct a bioinformatics analysis to compare the USP-14-regulated innate immunity genes identified in their RNAseq studies to lists of known BAR-1 transcriptional targets to look for potential overlap. __Response: We agree that extending these analyses to qRT-PCR experiments examining additional immune genes would be informative. However, both bar-1 mutants and bar-1 RNAi-treated worms exhibited severe developmental and physiological defects, including sick and dead animals during development, likely reflecting the pleiotropic developmental roles of BAR-1. Although fluorescence imaging and survival assays could be performed by selectively transferring surviving adults, we were concerned that bulk collection of worms for qRT-PCR analyses would introduce confounding effects arising from developmental defects and reduced viability.

To further address the reviewer’s suggestion, we carried out a comparative analysis between USP-14-dependent intestinal distension-induced immune genes and previously identified BAR-1-dependent immune genes. Although transcriptome-wide datasets for BAR-1-dependent pathogen-induced immune genes are not currently available, an earlier study identified seven immune response genes regulated by BAR-1 during infection (PMID: 18981407). We found that six of these genes overlap with the USP-14-dependent intestinal distension-induced immune genes identified in our study. These analyses have now been added to the Results section and included in Table S5.

Comment: __5. While in their Discussion section the authors mention evolutionarily conserved roles for protein ubiquitination as means of immunomodulation, there are few if any comments regarding ubiqutination as a regulatory scheme in C. elegans innate immunity or how their findings enhance our understanding of this phenomenon. Ubiquitination affects C. elegans immunity at multiple levels, from avoidance behavior to gene regulation, and it seems appropriate for the authors to address this in order to more fully contextualize their findings. __Response: We thank the reviewer for the suggestion. We have now added a new paragraph to the Discussion that places our findings in the context of the existing literature on ubiquitination, deubiquitination, and innate immunity in C. elegans. The discussion is added on pages 11-12, lines 299-312: “Although ubiquitin-mediated signaling has emerged as a central regulator of innate immunity across metazoans (Jiang & Chen, 2011; Mello-Vieira & Dikic, 2026), the contribution of DUBs to host defense in C. elegans remains poorly understood. Previous studies in C. elegans have shown that ubiquitin-dependent processes regulate diverse aspects of immunity, including immune surveillance, xenophagy, and pathogen tolerance (Garcia-Sanchez et al, 2021). Perturbations in proteasome function have also been shown to activate surveillance immunity (Ghosh & Singh, 2026; Troemel et al, 2026), highlighting the importance of ubiquitin-associated pathways in sensing pathogen-induced cellular damage. However, most prior studies have focused on ubiquitin ligases, proteasome-associated pathways, or global ubiquitin signaling rather than on specific DUBs directly regulating antibacterial immune responses. To our knowledge, our study provides the first direct evidence that a specific DUB regulates antibacterial innate immunity in C. elegans. Thus, our findings establish USP-14 as a previously unrecognized regulator of host defense and identify deubiquitination as an important regulatory layer in intestinal distension-mediated immunity.”

__Minor comments __1. In the Results section, the authors state that "[k]nockdown of cec-10 led to only a marginal decrease in survival during P. aeruginosa infection" (lines 92 and 93) and that cec-10 "has minimal impact on C. elegans survival during infection" (lines 93 and 94). However, as reported in Supplemental Table 5 the magnitude of the calculated difference in mean survival time between animals treated with RNAi targeting cec-10 and untreated control animals (-20% to -24% and statistically significant in 3/3 replicates) closely approximates the difference in mean survival between usp-14 mutants and controls (-19% to -28% and statistically significant in 3/3 replicates), which the authors clearly find to be significant. If by this metric usp-14 is important for host defense, then so too is cec-10. In light of this, the authors should use different language to describe the impact of cec-10 knockdown on the susceptibility of C. elegans to microbial infection and the potential role of cec-10 in immunity.

Response: We chose not to pursue cec-10 further primarily because it lacks a clear human homolog and because the mutant exhibited reduced expression of the co-injection marker, raising the possibility of broader transgene-related effects. We have modified the text on page 4, lines 93-97: “Knockdown of cec-10 resulted in a significant reduction in survival during P. aeruginosa infection (Figure S1C). However, we did not pursue cec-10 further for two reasons: (i) cec-10(jsn20) mutants exhibited a modest but significant reduction in the myo-2p::mCherry co-injection marker (Figure 1D), raising the possibility of broader transgene-related defects, and (ii) cec-10 lacks a clear human homolog.”

Comment: __2. All of the micrographs in Fig. 1B appear very dark. The GFP expression in the control animals appears dim, making it difficult for the reader to compare the signal in those animals to the GFP expression levels in the mutants. I recommend adjusting the brightness level in an equivalent manner across all of the micrographs to account for this. __Response: We have increased the brightness of all the images, as suggested by the reviewer.

__Comment: __3. Fig. 1E depicts a gene structure diagram for usp-14 with the position of the point mutation in the jsn19 allele isolated in the authors' forward genetic screen indicated by the amino acid substitution symbol drawn over the second exon. Instead of mixing gene- and protein-level information about the jsn19 allele, I recommend replacing the gene structure diagram with a domain structure diagram of the USP-14 protein that depicts the conserved C19 peptidase and ubiquitin-like domains. The relative position of the E122K substitution should still be noted. __Response: __We have now updated Figure 1E to include the functional domains of USP-14 and mapped both the usp-14(jsn19) missense allele and the usp-14(tm1481) deletion allele onto the protein schematic.

Comment: __4. Since all of the information in Fig. 1F appears elsewhere in the text, I recommend eliminating this panel. __Response: We have removed it.

Comment: __5. Regarding the RNAseq analysis, the authors state that 1241 genes are upregulated upon aex-5 knockdown (line 162). The authors then ask which of these genes are regulated by usp-14 in the context of intestinal distension and find that 633 are upregulated a usp-14-dependent manner when aex-5 is targeted by RNAi and that 595 are upregulated even in the absence of usp-14 (Fig. 3D). This accounts for 1228 genes in total, not 1241. Can the authors explain this discrepancy? __Response: We thank the reviewer for carefully noting this discrepancy. The difference arises from the criteria used to classify genes into the categories shown in Figure 5D (previously Figure 3D). Specifically, genes uniquely upregulated in usp-14(tm1481) worms were defined as genes that were either exclusively induced in usp-14(tm1481) worms or expressed at levels more than 2-fold higher in usp-14(tm1481) worms compared to N2 worms. During this classification, 13 genes that were initially identified as upregulated in N2 worms following aex-5 RNAi were found to be expressed at levels more than 2-fold higher in usp-14(tm1481) worms than in N2 worms (Table S4). These genes were therefore reassigned to the “usp-14(tm1481)-specific” category in the Venn diagram. Consequently, the total number of genes represented in the Venn diagram becomes 1228 instead of 1241. To clarify this point, we have now added an explanation to the figure legend.

Comment: __6. For the sake of clarity, in the legend to Fig. 3D I recommend expanding the description of the categories of genes depicted in the Venn diagram by using the same language as in the first worksheet of Supplemental Table 4. __Response: We thank the reviewer for the suggestion. We have now added these details to the legend of Figure 5D (previously Figure 3D). The legend reads: “(D) Venn diagram showing the overlap between genes upregulated upon aex-5 RNAi in N2 and usp-14(tm1481) worms. The GO analyses for the biological processes of unique and common genes are shown. USP-14-dependent genes were defined as genes that were either exclusively upregulated in N2 worms or expressed at levels greater than 2-fold higher in N2 worms than in usp-14(tm1481) worms. USP-14-independent genes were defined as genes upregulated in both N2 and usp-14(tm1481) worms with expression differences of less than 2-fold between the two strains. Genes uniquely upregulated in usp-14(tm1481) worms were defined as genes that were either exclusively induced in usp-14(tm1481) worms or expressed at levels greater than 2-fold higher in usp-14(tm1481) worms than in N2 worms. Thirteen genes classified as upregulated in N2 worms were more than 2-fold higher in usp-14(tm1481) worms than in N2 worms (Table S4) and were therefore included in the usp-14(tm1481)-specific category.”

Comment: __7. In Fig. 4B, the authors' annotation indicates that there is a statistically significant difference (**, p __Comment: __8. In Fig. S5, the shade of blue used to represent the data from the nhr-49(nr2041);usp-14(tm1481);clec60p::gfp animals in panel E is different from that used to represent data from the same animals in panel B. This breaks the pattern of all of the other panels of this figure in which the data pertaining to a given phenotype are depicted in the same color. Also, in the symbol key in panel E there is an extra semi-colon before clec-60p::gfp that should be eliminated in the second genotype notation. __Response: We thank the reviewer for carefully examining the figure and for bringing these issues to our attention. We have made the changes.

Comment: __9. The authors' data show that USP-14 regulates bar-1 expression, and in the Discussion section they mention that in mammals beta-catenin is a substrate of USP14. Can the authors comment on the possibility of/evidence for BAR-1 autoregulation in C. elegans and the prospect of it being facilitated by USP-14? This could be a minor point to add to the Discussion. __Response: In both contexts, USP-14 appears to stabilize BAR-1 by regulating it at either the transcriptional or post-translational level. However, it is currently unknown whether BAR-1 regulates USP-14 expression and thereby participates in an autoregulatory mechanism. Nevertheless, we have added to the Discussion that USP14 may regulate the Wnt pathway through both transcriptional and post-translational mechanisms, depending on the biological context. __Reviewer #2 (Significance (Required)): __ The study described in this manuscript ties in to the findings from two prior genetic screens carried out in C. elegans that aimed to identify immune regulators (Ren et al., Cell Reports, 2022 and Labed et al., Immunity, 2018). Though their strategies differed, both of these previous studies uncovered a role for acetylcholine receptors in modulating the response to ingested microbial pathogens, especially when infection is associated with intestinal distension, indicating that a neuron-to-gut axis controls innate immunity in C. elegans. Labed and colleagues were the first to show that activation of this pathway results in the upregulation of genes encoding Wnt signaling pathway components, including the worm ortholog of beta-catenin called bar-1, which are necessary for the expression of immune effectors in the intestine. The Labed study also revealed that protein ubiquitination could contribute to regulating host defense gene induction because knockdown of lin-23, the substate binding subunit of a ubiquitin ligase complex that mediates BAR-1 degradation, results in constitutive expression of clec-60p::gfp, the same transcription reporter used by Ghosh and Singh as a readout for the expression of innate immunity genes. In their screen that revisits the Ren et al. approach, Ghosh and Singh find that another protein implicated in regulating protein stability via ubiquitination status, USP-14, also controls the expression of innate immunity genes in response to intestinal distension. Interestingly, their data indicate that it does so by upregulating bar-1. This discovery therefore adds an element of mechanistic detail regarding the regulation of Wnt signaling in immunity. While the Labed data suggest that ubiquitination may regulate BAR-1 at the post-translational level, Ghosh and Singhs' results indicate a second layer of regulation of bar-1 at the transcriptional level that also appears to involve ubiquitination. In this case, USP-14 is predicted to modulate the ubiquitination status of a yet-to-be-identified substrate that directly or indirectly governs bar-1 expression. The authors' findings thus bring the field closer to having a complete picture of the Ach-Wnt pathway in C. elegans. As they point out in the Discussion section of their manuscript, ubiquitination is an evolutionarily conserved yet complex means of tuning the immune system. The work described here helps to shed light on this important immune regulatory mode and could have implications for aspects of epithelial immunity that are in common to both invertebrates and vertebrates.

Response: We thank the reviewer for providing such a thoughtful overview of the field and for placing our findings in the context of previous studies on intestinal distension-induced immunity in C. elegans. We also sincerely appreciate the reviewer’s constructive feedback and insightful comments, which have helped us improve the quality and clarity of the manuscript.

My research interest and specific area of expertise pertains to evolutionarily conserved genetic pathways that control healthspan through affecting cellular resilience later in life. Using C. elegans as a surrogate for aging humans, my group studies age-dependent changes in the activity of regulatory modules that protect older animals from the molecular damage associated with intrinsic and extrinsic sources of cellular stress, with a particular emphasis on microbial infection and oxidative stress.

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

C. elegans are soil-dwelling nematodes that feed on bacteria and fungi and thus must be able to distinguish between innocuous and pathogenic species of microbes to survive. Though they lack adaptive immunity, these animals have an ancient version of an innate immune system that has no circulating sentinel or phagocytic cells yet can still mount a response to pathogen exposure. A consequence of the mode of infection of some ingested bacterial pathogens is intestinal distension which by itself, even in the absence of pathogens, is sufficient to trigger the expression of genes encoding immune effectors, including proteins that are bactericidal. The complete mechanistic scheme connecting intestinal distension to the expression of immunity genes has not been resolved, motivating the authors to perform a forward genetic screen for additional components of this pathway. One mutant that the authors isolated was usp-14, encoding an evolutionarily conserved deubiqutinating enzyme. Functional analysis revealed that usp-14 confers protection from microbial pathogens and that the intestine is its primary site of action for its role in host defense. The authors' data indicate that while USP-14 regulates the expression of innate immunity genes that are induced by intestinal distension, surprisingly it functions independently of several canonical innate immune signaling pathways, including the pmk-1/p38 MAPK pathway. Instead, USP-14 appears to act through Wnt signaling to regulate immune effectors by upregulating the expression of several components of that pathway, including the C. elegans ß-catenin ortholog bar-1. This places usp-14 within a gut-brain axis previously shown to control the C. elegans innate immune response through acetylcholine-mediated activation of Wnt signaling. The authors' findings provide new mechanistic insight to this pathway and add to the understanding of ubiqutination as an immune regulatory module.

Major comments

1. There are three types of experiments in which the authors use the same set of controls across several different figure panels, as stated in the legend to Figure 2. First, when quantifying GFP levels of clec-60::gfp in RNAi-treated animals, the authors use the same clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls for Fig. 1K, 2C, and 2G. For infection assays with S. aureus NCTC8325, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2E are the same as the ones used in Fig. 1M. Similarly, for infection assays with P. aeruginosa PA14, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2I is the same as was used for Fig 1I. In each case, if the authors in fact collected all of the data for each strain that they studied at the same time but then chose to parse larger datasets into separate figure panels to make it more clear to the reader, then this approach is valid but the authors need to explicitly state that this is what they did. However, if the data pertaining to the control strains were collected at a different time or if it comes from a separate biological replicate, then re-using data from the controls is not appropriate because it would not accurately reflect the specific conditions of the experiment to which the data are being compared. If this is indeed the scenario, then the authors will need to repeat these experiments and include the appropriate control in each iteration.
2. From the legends describing figure panels that include data pertaining to clec-60p::gfp expression levels as assessed by fluorescence microscopy it seems that, in general, the authors measured GFP fluorescence in about 30 animals to produce quantitative data. How many biological replicates of these types of experiments were carried out? This is not explicitly stated in the section describing fluorescence imaging in the Methods section. Following the description of their methodology regarding statistical analysis of survival curves from microbial infection assays, however, the authors state that, "[a]ll experiments were performed independently at least three times unless otherwise noted." Does this statement apply to microscopy or only to experiments involving infection assays? If the data reporting quantitation of GFP signal is based on only 30 animals, then additional biological replicates are necessary, along with appropriate statistical analyses.
3. The authors have made all of the RNASeq data publicly available on the Sequence Read Archive, and they include data from several pairwise comparisons for differential gene expression analysis in their supplemental files. One of the most important facts to come out of the authors' Gene Ontology analyses of their RNASeq data is that the genes that are upregulated in a usp-14-dependent manner upon intestinal distension are enriched for those whose products play a role in innate immunity/host defense. The authors should say more about these genes. Are there any commonalities between them with regard to function? Are any of them targets of transcription factors that are known to function in C. elegans innate immunity? If so, this could provide clues as to what the substrates of USP-14 might be. Importantly, the specific identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense should be revealed in a supplemental file, and designated as being dependent on or independent of usp-14 for their expression during intestinal distension.
4. The authors' data suggest that in response to bacterial infection USP-14 upregulates the expression of bar-1, along with other components of the Wnt signaling pathway, which in turn upregulates innate immunity genes. This could be further substantiated by directly demonstrating that there are USP-14-regulated innate immunity genes whose induced expression in the presence of microbial pathogens also requires bar-1. Along those lines, an initial test would be to assess clec-60p::gfp expression in bar-1 animals versus bar-1;usp-14 double mutants, similar to the experiment whose results are reported in Fig. S4. If generating the bar-1;usp-14 double mutant is not feasible, then RNAi could be used to knockdown bar-1 expression in clec-60p::gfp;usp-14(tm1481) animals. To expand this analysis, the expression of the six innate immunity genes shown to be regulated upon intestinal distension in usp-14-dependent manner could be measured in the presence and absence of intestinal distension or microbial infection in bar-1 and bar-1;usp-14 animals by qRT-PCR. At a minimum, the authors should conduct a bioinformatics analysis to compare the USP-14-regulated innate immunity genes identified in their RNAseq studies to lists of known BAR-1 transcriptional targets to look for potential overlap.
5. While in their Discussion section the authors mention evolutionarily conserved roles for protein ubiquitination as means of immunomodulation, there are few if any comments regarding ubiqutination as a regulatory scheme in C. elegans innate immunity or how their findings enhance our understanding of this phenomenon. Ubiquitination affects C. elegans immunity at multiple levels, from avoidance behavior to gene regulation, and it seems appropriate for the authors to address this in order to more fully contextualize their findings.

Minor comments

1. In the Results section, the authors state that "[k]nockdown of cec-10 led to only a marginal decrease in survival during P. aeruginosa infection" (lines 92 and 93) and that cec-10 "has minimal impact on C. elegans survival during infection" (lines 93 and 94). However, as reported in Supplemental Table 5 the magnitude of the calculated difference in mean survival time between animals treated with RNAi targeting cec-10 and untreated control animals (-20% to -24% and statistically significant in 3/3 replicates) closely approximates the difference in mean survival between usp-14 mutants and controls (-19% to -28% and statistically significant in 3/3 replicates), which the authors clearly find to be significant. If by this metric usp-14 is important for host defense, then so too is cec-10. In light of this, the authors should use different language to describe the impact of cec-10 knockdown on the susceptibility of C. elegans to microbial infection and the potential role of cec-10 in immunity.
2. All of the micrographs in Fig. 1B appear very dark. The GFP expression in the control animals appears dim, making it difficult for the reader to compare the signal in those animals to the GFP expression levels in the mutants. I recommend adjusting the brightness level in an equivalent manner across all of the micrographs to account for this.
3. Fig. 1E depicts a gene structure diagram for usp-14 with the position of the point mutation in the jsn19 allele isolated in the authors' forward genetic screen indicated by the amino acid substitution symbol drawn over the second exon. Instead of mixing gene- and protein-level information about the jsn19 allele, I recommend replacing the gene structure diagram with a domain structure diagram of the USP-14 protein that depicts the conserved C19 peptidase and ubiquitin-like domains. The relative position of the E122K substitution should still be noted.
4. Since all of the information in Fig. 1F appears elsewhere in the text, I recommend eliminating this panel.
5. Regarding the RNAseq analysis, the authors state that 1241 genes are upregulated upon aex-5 knockdown (line 162). The authors then ask which of these genes are regulated by usp-14 in the context of intestinal distension and find that 633 are upregulated a usp-14-dependent manner when aex-5 is targeted by RNAi and that 595 are upregulated even in the absence of usp-14 (Fig. 3D). This accounts for 1228 genes in total, not 1241. Can the authors explain this discrepancy?
6. For the sake of clarity, in the legend to Fig. 3D I recommend expanding the description of the categories of genes depicted in the Venn diagram by using the same language as in the first worksheet of Supplemental Table 4.
7. In Fig. 4B, the authors' annotation indicates that there is a statistically significant difference (**, p<0.01) in the fluorescence signal from clec-60p::gfp in usp-14(jsn19);aex-5(sa23);clec-60p::gfp_EV versus usp-14(jsn19);aex-5(sa23);clec-60p::gfp_bar-1 animals. This is likely a typographical error that should be changed to "ns" to indicate no significant difference in the fluorescence signal between these two groups, which is consistent with what the data show and with the authors' description of these data in the text (lines 211-214).
8. In Fig. S5, the shade of blue used to represent the data from the nhr-49(nr2041);usp-14(tm1481);clec60p::gfp animals in panel E is different from that used to represent data from the same animals in panel B. This breaks the pattern of all of the other panels of this figure in which the data pertaining to a given phenotype are depicted in the same color. Also, in the symbol key in panel E there is an extra semi-colon before clec-60p::gfp that should be eliminated in the second genotype notation.
9. The authors' data show that USP-14 regulates bar-1 expression, and in the Discussion section they mention that in mammals beta-catenin is a substrate of USP14. Can the authors comment on the possibility of/evidence for BAR-1 autoregulation in C. elegans and the prospect of it being facilitated by USP-14? This could be a minor point to add to the Discussion.

Significance

The study described in this manuscript ties in to the findings from two prior genetic screens carried out in C. elegans that aimed to identify immune regulators (Ren et al., Cell Reports, 2022 and Labed et al., Immunity, 2018). Though their strategies differed, both of these previous studies uncovered a role for acetylcholine receptors in modulating the response to ingested microbial pathogens, especially when infection is associated with intestinal distension, indicating that a neuron-to-gut axis controls innate immunity in C. elegans. Labed and colleagues were the first to show that activation of this pathway results in the upregulation of genes encoding Wnt signaling pathway components, including the worm ortholog of beta-catenin called bar-1, which are necessary for the expression of immune effectors in the intestine. The Labed study also revealed that protein ubiquitination could contribute to regulating host defense gene induction because knockdown of lin-23, the substate binding subunit of a ubiquitin ligase complex that mediates BAR-1 degradation, results in constitutive expression of clec-60p::gfp, the same transcription reporter used by Ghosh and Singh as a readout for the expression of innate immunity genes. In their screen that revisits the Ren et al. approach, Ghosh and Singh find that another protein implicated in regulating protein stability via ubiquitination status, USP-14, also controls the expression of innate immunity genes in response to intestinal distension. Interestingly, their data indicate that it does so by upregulating bar-1. This discovery therefore adds an element of mechanistic detail regarding the regulation of Wnt signaling in immunity. While the Labed data suggest that ubiquitination may regulate BAR-1 at the post-translational level, Ghosh and Singhs' results indicate a second layer of regulation of bar-1 at the transcriptional level that also appears to involve ubiquitination. In this case, USP-14 is predicted to modulate the ubiquitination status of a yet-to-be-identified substrate that directly or indirectly governs bar-1 expression. The authors' findings thus bring the field closer to having a complete picture of the Ach-Wnt pathway in C. elegans. As they point out in the Discussion section of their manuscript, ubiquitination is an evolutionarily conserved yet complex means of tuning the immune system. The work described here helps to shed light on this important immune regulatory mode and could have implications for aspects of epithelial immunity that are in common to both invertebrates and vertebrates.

My research interest and specific area of expertise pertains to evolutionarily conserved genetic pathways that control healthspan through affecting cellular resilience later in life. Using C. elegans as a surrogate for aging humans, my group studies age-dependent changes in the activity of regulatory modules that protect older animals from the molecular damage associated with intrinsic and extrinsic sources of cellular stress, with a particular emphasis on microbial infection and oxidative stress.

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 manuscript, the authors describe the discovery of a molecular regulator of the immune transcriptional program, which is activated by intestinal distension upon bacterial colonization of the C. elegans intestine. Taking advantage of the fact that inhibition of aex-5 is known to cause intestinal distension and a C-type lectin gene clec-60 as a marker for the immune response to intestinal distension (clec-60p::gfp), the authors performed a forward genetic screen for suppressors of the immune response activation. Of the two mutants isolated, they focused on the stronger suppressor, which corresponded to a cysteine-type DUB, the Ubiquitin Specific Peptidase-14 (usp-14). Through rescue experiments, phenocopy analyses, and quantitative RT-PCR, they validated usp-14 as the causal gene and initiated characterization of its role in immune response activation. To this end, the authors investigated the tissue of action, identifying the intestine as the tissue in which usp-14 mediates the regulation of the immune response. Through transcriptomic analyses, they found that the signalling pathway likely regulated by usp-14 in response to intestinal distension is the Wnt pathway, as they have observed reduction in the transcriptional level of some of the Wnt pathway components in usp-4(tm1481), in response to infection with S. aureus. Additionally, transcriptomic data indicate that usp-14 plays a role in immunity regulation even in the absence of infection. Based on these findings, the authors propose that usp-14 has a dual role in immune regulation: one in surveillance immunity, preventing overactivation of immune responses, and another as a mediator of pathogen-induced responses, such as those triggered by P. aeruginosa or S. aureus. The experiments are rigorous and the results robust; however, some points would benefit from further investigation or clarification.

The expression domain of usp-14 appears to be quite expanded based on single cell RNAseq data (e.g. PMID: 28818938) therefore it is likely that the transgenes used for expression analysis are lacking key regulatory information. Alternative methods like smFISH would be more appropriate to characterise the spatiotemporal pattern of usp-14 expression in more detail.

The mutation mapped in usp-14(jsn19) is a missense mutation (E122K) that suppresses the immune response to a degree comparable to the usp-14(tm1481) deletion allele. However, the authors do not show the functional domains in Fig. 1E potentially affected by this missense mutation.

How USP-14 regulates Wnt and how Wnt signalling relates to activation of immune responses is not fully supported. Are the Wnt components mentioned in the study induced specifically in the intestine upon infection and does USP-14 act in the intestine in the context of this regulation? How do the authors interpret that both Wnt ligands and receptors are induced ? Does Wnt signalling appear as a GO term in the transcriptomic analysis? The authors can include Wnt signalling components in the analysis of the transcriptomic results.

Overall, in most of the figures, the micrographs are in general quite dark and exhibit poor contrast between signal and background, particularly in Fig. 1, panels B and J, and Fig. 2, panels B and F (upper rows). Even though these panels are intended to show absence of response, the outlines of the worms are difficult to discern.

In Figure S3, panels A and B, the pmk-1(km25); usp-14(tm1481) animals subjected to aex-5 RNAi show some level of fluorescence/response induction comparable to pmk-1(km25) alone. This observation is not discussed in the text.

Significance

The work is interesting because it expands some previous work in the field demonstrating immune response induction as a consequence of intestinal distension even in the absence of bacterial infection. This is known to be mediated by the neuronal acetylcholine receptor ACC-4, which signals to the intestine where it regulates immune genes via the Wnt pathway. However, how USP-14 relates to ACC-4 is currently unclear and whether USP-14 function is really required in the intestine to control Wnt signalling is not demonstrated. The authors should include a model to describe how their findings relate to the previous literature and how USP-14 may link mechanistically to Wnt signalling pathway activation.

It remains also unclear whether usp-14 is the only deubiquitinase involved in intestinal distension-induced signalling via the Wnt pathway, or whether other paralog usp genes might also contribute to regulation of immune-responsive transcription. Notably, several mammalian deubiquitinases have established roles in cancer suppression and inflammatory response and innate immunity in other systems so this would increase the potential significance of the work.

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

1. General Statements

We thank the reviewers for their positive and constructive assessment of the manuscript. We are encouraged that all three reviewers recognise the value of coelsch as an open-source framework for haplotyping and crossover detection from single-cell gamete sequencing data, and that they view the study as a useful contribution to the fields of recombination and genetic research. We are particularly grateful that Reviewer 1 described the manuscript as an "interesting and important study" and a "genuinely useful methodological framework that fills a real gap in the recombination biology toolkit", while Reviewer 2 highlighted its "strong innovation, complete technical pipeline, and significant biological implications" and considered it an "important technical breakthrough". We also appreciate Reviewer 3's assessment that the study provides "timely guidance for experimental design", that the results are "important for guiding plant single-cell research" in general, and that the work "has the potential to attract a broad readership".

In our view, the main contribution of the manuscript is the development of a platform-agnostic method for recovering haplotypes and crossover events from single-cell sequencing data. This addresses an important practical gap: single-cell gamete sequencing has strong potential for high-throughput haplotyping and recombination mapping, but its broader use requires tools that can accommodate the very different coverage structures produced by different sequencing modalities and platforms. coelsch was designed to meet this need.

The experimental datasets in the manuscript serve two purposes. First, they demonstrate that coelsch can be applied across multiple single-cell modalities and platforms, including scRNA, scATAC and scWGA sequencing from 10x Genomics, BD, and Takara platforms. Second, they illustrate the kinds of biological and practical questions that can be addressed with single-cell gamete sequencing, including crossover detection in meiotic mutants and large-scale analysis of natural variation in recombination.

While all reviewers strongly supported the publication of the work, they also raised important points about specific aspects, including technical variation and reproducibility, the rationale for using 10x scRNA to generate the diversity panel dataset, and the effects of coverage on crossover localisation, amongst others. We agree that addressing these points will make the manuscript clearer and more useful to readers. Our planned revisions therefore aim to strengthen the experimental and computational support for the framework, clarify the interpretation of the modality comparisons, and provide additional guidance for researchers who may wish to apply coelsch or related single-cell sequencing approaches in future studies.

2. Description of the planned revisions

2.1. Additional technical replicates and clearer treatment of batch/sample-handling effects

Reviewers 1, 2 and 3 all noted that the comparison of different platforms and modalities is based on limited replication, with different nuclei isolation and processing strategies used for different technologies. Reviewer 3 requested a fully controlled benchmark in which the same nuclei preparation is split across all tested platforms. We agree that this would be the ideal design for a dedicated head-to-head benchmarking study. However, the primary aim of the manuscript is to demonstrate the applicability of coelsch across different single-cell sequencing data types, rather than to provide a definitive benchmark of the intrinsic performance of each modality and platform.

In addition, a fully matched and replicated cross-platform experiment for all technologies is not feasible. Isolated nuclei deteriorate rapidly after preparation and must be processed promptly for single-cell library construction; this makes it impractical to distribute the same preparation across multiple time- and labour-intensive workflows. However, this design is feasible for 10x scRNA-seq and 10x scATAC-seq. To address this point directly, we will therefore generate two matched technical replicates each of 10x scRNA-seq and 10x scATAC-seq from nuclei isolated in the same sorting run.

We will also improve our library-level QC summary tables. We will report, where available, the number of nuclei used for loading, recovered barcodes, barcodes retained after QC, inferred high-quality nuclei and artefacts, informative fragments per nucleus, genomic bin coverage, and final nuclei used for crossover calling. This will make the effects of loading, capture efficiency, QC filtering, and modality-specific data loss more transparent.

In the revised text, we will distinguish more clearly between modality-specific effects and possible batch/sample-preparation effects. Where the current manuscript implies that differences are intrinsic properties of sequencing platforms, we will soften the interpretation unless supported by the new replicate data, reproducibility analyses, or well-supported properties that have been reported previously in literature.

2.2. Rationale for using 10x scRNA-seq in the natural variation panel

Reviewers 1 and 3 asked why the natural variation panel was analysed using 10x scRNA-seq, given that Takara scWGA produced higher per-cell crossover localisation accuracy in the modality comparison. We will revise the manuscript to explain this experimental decision more clearly.

The natural variation panel was designed as a high-throughput experiment requiring sufficient numbers of usable nuclei from many pooled F₁ hybrids. In our hands, 10x scRNA-seq has generally produced the largest number of usable nuclei barcodes and the lowest proportion of artefacts. This makes 10x scRNA-seq well suited to experiments where many nuclei are required per genotype. By contrast, applying Takara scWGA to a pooled panel of this scale would be expected to recover only tens of usable nuclei per F₁ hybrid, which would be insufficient for robust recombination-rate or landscape estimation.

We will add this explanation to the relevant Results section and clarify that the choice of 10x scRNA-seq reflects a trade-off between per-cell crossover resolution and the number of informative nuclei recovered per genotype. We will also add genotype-level summaries for the pooled natural variation experiment, including assigned nuclei per genotype and genotype-specific genomic coverage of informative fragments.

2.3. Reproducibility of recombination landscapes across replicates and modalities

Reviewer 1 requested recombination landscape plots for all tested modalities, and several comments raised the need to show within-modality reproducibility. We will add recombination landscape plots for wild-type Col-0 × Ler libraries across the tested modalities, including the newly generated replicate 10x scATAC and scRNA libraries.

We will assess reproducibility using comparisons of unsmoothed, non-overlapping windowed recombination-rate estimates, both within and between modalities. These will be quantified using bootstrapped estimates of spearman rank correlation coefficient, and visualised using scatterplots and/or recombination landscapes.

2.4. Sequencing depth, coverage, and crossover localisation resolution

Reviewers 1 and 3 requested clearer quantitative reporting of crossover resolution and a stronger analysis of depth effects. We will revise the manuscript to report practical crossover localisation resolution for each modality, including median and interquartile localisation error or interval size in genomic units.

We will expand the simulation analyses to compare false-positive and negative rates and localisation accuracy across modalities, including telomere-proximal error profiles for scWGA and scATAC as well as 10x RNA data. We will perform downsampling analyses to assess how crossover detection accuracy changes as a function of informative-fragment depth. Where feasible, we will compare depth-matched subsets across modalities to distinguish effects of sequencing depth from modality-specific coverage structure.

These analyses will be used to clarify the extent to which each modality is suitable for different applications, such as broad landscape estimation, crossover counting, or fine localisation.

2.5. Artefact detection, high doublet rates, and representativeness after filtering

All three reviewers raised concerns about the high proportion of barcodes excluded by the filtering procedure, particularly in the Takara scWGA dataset. In hindsight, we believe part of this concern stems from the poor choice of terminology ("doublets") we used to describe these excluded barcodes.

While true doublets (i.e. two nuclei entering a single droplet or nanowell) are one likely source of such signals, the filtering procedure more broadly identifies artefactual barcodes that do not exhibit a clear single-gamete haplotype structure. These barcodes may arise from a variety of sources, including doublets, multiplets, high levels of ambient DNA or RNA, or empty droplets containing only ambient material. Although visual examination can be used to make predictions about the source of these artefacts, our detection method does not attempt to distinguish between them, and artefacts in different modalities may stem from different sources in varying proportions. We will therefore revise the terminology throughout the manuscript to clarify that these represent a broader class of low-confidence or noise barcodes, rather than confirmed doublets.

For the Takara scWGA data, we will revise the manuscript to discuss the discrepancy between the CellSelect well classifications (which uses proprietary software to label doublets) and the final artefact predictions from coelsch. We can only speculate as to why CellSelect failed to detect many apparent doublet and multiplet artefacts in this experiment, but we agree with the reviewer that the most likely explanation is the small size of Arabidopsis pollen nuclei relative to the expectations of the imaging and classification procedure. To support this interpretation, we will add supplementary analysis comparing the CellSelect images from individual nanowells with the final doublet predictions inferred from scWGA data. This will allow readers to see examples of wells classified as acceptable by CellSelect but subsequently inferred to contain artefacts based on their haplotype structure.

We will also add sensitivity analyses showing how key results change under different artefact-filtering thresholds. These analyses will include crossover count distributions, recombination landscape estimates, and modality-level comparisons. We will examine the extreme upper tail of crossover counts observed in 10x scATAC-seq and assess whether these barcodes are artefacts that have escaped detection.

Finally, we will assess whether retained singlets are representative of the input data with respect to informative-fragment counts, coverage, and inferred crossover patterns. This will address the concern that filtering could preferentially remove nuclei with particular recombination profiles.

2.6. Biases arising from pollen nuclear biology

Reviewer 2 raised an issue concerning the biases arising from the two different nuclei types present in mature trinuclear Arabidopsis pollen, and reviewer 3 endorsed this point. While we do not agree with the reviewer that scRNA and scATAC cannot capture sperm nuclei due to their condensed nature (see Parker et al. 2025 PLoS Biology for evidence against this claim), it is true that technical variation in nuclei isolation and sorting may affect the relative representation of nuclei types - usually, however, resulting in the underrepresentation of vegetative nuclei (Parker et al. 2025). We will add text addressing this point to the manuscript.

It is also true that differences in expressed genes between vegetative and sperm nuclei, which have very different transcriptomic profiles, will affect the distribution of informative reads for crossover analysis in scRNA data, and therefore may also have an impact on the recovered recombination landscapes (despite that the underlying landscapes are biologically identical). We will address this in the manuscript by adding recombination landscape plots and reproducibility scatterplots (as described in point 2.3) comparing sperm and vegetative nuclei from scRNA-seq to the manuscript.

2.7. Robustness of the pipeline and parameter choices

Reviewer 3 raised the concern that quantitative conclusions depend on a single pipeline with fixed parameter choices. We will address this by adding a parameter-sensitivity analysis for the main computational steps. Specifically, we will test the robustness of crossover calling on simulated data to changes in bin size and rHMM parameters, showing how these affect sensitivity to noise and agreement of predictions with ground truth data.

2.8. Natural variation analysis: genotype-specific coverage and terminal crossover enrichment

Reviewers 1, 2 and 3 raised concerns about whether natural variation in crossover rate and terminality could be influenced by genotype-specific coverage, marker density, pooling imbalance, or dropout. We will add a more detailed description of how pollen from different F₁ hybrids was pooled and how genotype assignment was performed. We will report genotype-level recovery statistics, including the six hybrids excluded from downstream analysis, and discuss how imbalances may arise, e.g. through biological variation in pollen count and fertility, biases in nuclei isolation or sequencing, and biases in genotyping and informative fragments.

Reviewer 1 specifically asked whether the lower terminal crossover index observed in Cvi-0 crosses compared with Col-0 crosses could reflect systematic differences in informative-fragment distributions rather than true biological differences in crossover localisation. We will address this by using the genotype-specific informative-fragment distributions observed in the diversity-panel scRNA-seq dataset to simulate crossover datasets with known ground truth. This will allow us to test whether differences in marker variant or expressed-gene distributions causing variation in informative-fragment distribution could systematically bias terminal crossover detection in Cvi-0 crosses relative to Col-0 crosses.

If feasible within the revision timeframe, we will also perform an orthogonal validation experiment for a selected comparison showing a clear difference in crossover terminality, such as Col-0 × Sah-0 and Cvi-0 × Sah-0. This would use progeny sequencing of backcross populations to estimate recombination landscapes independently of single-cell scRNA-seq, providing a direct test of whether the inferred terminality difference is supported by conventional recombination mapping. If this experiment cannot be completed within the revision timeframe, we will clearly state this limitation and base the revised interpretation on the simulation analyses described above.

2.9. Broader applicability and practical guidance for users

Reviewer 1 requested more discussion of applicability beyond Arabidopsis and to outcrossing or polyploid species. We will expand the Discussion to address the requirements and limitations of applying coelsch in other systems.

2.10. Minor figure, reference, and presentation revisions

We will address the remaining minor comments, including adding missing axis labels and checking duplicated references.

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

No revisions have yet been incorporated in the transferred manuscript.

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

4.1. Full new benchmark across all modalities from the same nuclei preparation.

As acknowledged in section 2.1, we agree with Reviewer 3 that a fully controlled benchmark in which the same isolated nuclei preparation is split across all tested platforms would be the ideal experimental design for separating intrinsic modality- or platform-specific effects from sample-handling and batch effects. However, this is not feasible for all technologies within the scope of this revision, because isolated nuclei degrade quickly, the single-cell sequencing methods are time- and labour-intensive, and the relevant platforms are not all available to us in the same location.

We will therefore not perform a complete new cross-platform benchmark across all modalities. Instead, we will address this issue in the parts of the experiment where a matched design is feasible: we will generate two additional matched technical replicates each for 10x scRNA-seq and 10x scATAC-seq from nuclei isolated in the same sorting run. We will also revise the manuscript to more clearly acknowledge the limitations imposed by the lack of a fully matched cross-platform design and to ensure that our conclusions are interpreted in that context.

4.2. Profiling the natural variation panel with a second modality

Reviewer 1 suggested profiling at least a subset of the diversity panel with an additional single-cell modality. We agree that this would be useful, but we do not currently plan to generate a second-modality dataset for the natural variation panel. We would like to point out that this dataset introduces 34 genetic maps in a single sequencing experiment, which is not easily repeated.

The natural variation experiment was designed as a high-throughput survey across many F₁ hybrids, and repeating even a subset with scWGA or scATAC would require substantial additional sample preparation and sequencing. Instead, we will strengthen the justification for the use of 10x scRNA-seq by adding genotype-level coverage summaries and simulations to show which conclusions are well supported at the observed data density.

4.3. Orthogonal progeny sequencing from the exact same F₁ plants

Reviewer 3 suggested that progeny sequencing from the same F₁ plants used for single-cell assays would provide a direct ground truth. This experiment would require additional crosses, progeny generation, and matched single-cell and progeny sequencing, which would not be justified by the insights that this effort delivers: While progeny sequencing can provide an independent validation dataset, we do not agree that it would constitute a substantially better ground truth than the simulations used here. Simulations provide a known ground truth for every individual barcode, whereas progeny sequencing cannot, for the obvious reason that pollen grains are destroyed during single-cell sequencing and therefore cannot be used to generate offspring. In addition, progeny-derived recombination landscapes are not a perfect ground truth at the population level, since segregation distortion and post-meiotic selection can alter the observed distribution of recombination events relative to those present in the original pollen population.

4.4. Formal benchmarking of ____coelsch____ as a structural-variant detection method

Reviewer 2 asked whether large structural variants were identified in other accessions besides Zin-9, and what sensitivity and specificity can be expected from recombination coldspot-based structural-variant detection. We agree that this is an interesting question, given that the Zin-9 inversion was identified through its strong effect on recombination. However, we do not plan to develop or benchmark coelsch as a comprehensive structural-variant detection method as part of this revision.

The Zin-9 event was identified by visual inspection of the recombination maps, where it appeared as an unusually large and conspicuous recombination coldspot. We did not develop a systematic structural-variant calling procedure, as we do not view recombination suppression alone as a sufficiently specific signal for structural-variant detection. Coldspots can arise for many reasons, including centromere proximity or local recombination modifiers. Therefore, although large rearrangements such as inversions or translocations may sometimes be detectable through their effects on recombination, coelsch should not be considered as a general-purpose structural-variant caller.

In the revised manuscript, we will clarify this limitation and avoid implying that recombination coldspot analysis provides comprehensive structural-variant discovery. We will report that we did not observe other genotype-specific coldspots of comparable scale to the Zin-9 event among the other analysed accessions, although smaller coldspots such as one corresponding to the previously reported 2.2Mb inversion on Chromosome 1 of N13 were identifiable. We will not provide formal estimates of sensitivity and specificity for structural-variant detection, as this would require independent benchmark datasets or dedicated simulations that are beyond the scope of the present study.

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 study, Parker et al. benchmark three single-cell sequencing modalities (scRNA-seq, scATAC-seq, and scWGA) in Arabidopsis gametes and deliver an open-source, end-to-end framework for data processing that enables high-throughput crossover mapping across hybrids. By systematically comparing these modalities, the work quantifies trade-offs in throughput, genomic coverage, and crossover detection sensitivity, offering timely guidance for experimental design in plant systems where single-cell genomics is still emerging and platform benchmarks are very limited. The pipelines are further supported by the discovery of a previously unrecognized ~10 Mb pericentric inversion in the Zin-9 accession. The experimental design is technically interesting, and the results are important for guiding plant single-cell research. The work has the potential to attract a broad readership. However, several aspects of the experimental design, validation strategy, and parameter robustness require further clarification and, where possible, additional analyses.

Major comments

1. The modality comparison is based on one scRNA-seq library and two libraries each for scATAC-seq and scWGA. While the limited replication is acknowledged in the Discussion, the authors also report unexpected and run-specific observations (e.g. unusually high doublet rates in the 10x scRNA-seq library; "unexpected" doublet behavior in scWGA), making it difficult to separate platform-intrinsic properties from sample preparation and run-to-run variation. Differences in nuclei isolation buffers, purification strategies (e.g. density gradients, FACS, centrifugation), and potentially loaded nuclei numbers between platforms (which have not been specified in detail) further confound modality-level conclusions. For example, total usable barcodes vary drastically between the samples (e.g. 15k/20k/33k for 10x scRNA-seq, only 3.8k for BD even though it has the same capture capacity as 10X). Do these differences reflect different capture efficiencies between the platforms, or variation in nuclei quality/quantity, or modality-specific limitations in QC thresholds? It would strengthen the study to provide, for each library, the number of nuclei prior to loading and before/after QC, and to add independent biological replicates under modality-appropriate, optimized handling, ideally including a design where the same nuclei pool is split across all three modalities.
2. All quantitative inferences rely on one custom analysis pipeline with multiple interdependent steps and fixed parameter choices (e.g. bin size, HMM transition structure, smoothing settings, background subtraction, doublet filters). The lack of benchmarking against independent crossover callers, or of systematic parameter sweeps, leaves it unclear how robust key patterns are to alternative analytical choices. It would substantially increase confidence to assess sensitivity of the main conclusions to key parameters (for example varying bin size, rigid chain length/transition penalties, enabling/disabling background subtraction and doublet filtering), and/or compare coelsch to other HMM-based crossover callers such as sgcocaller/comapr on at least a subset of the data.
3. Accuracy is evaluated by comparisons to prior backcross/progeny datasets generated in different conditions, and by simulations calibrated to those references. While this is informative, systematic biases shared between the new pipeline and the reference datasets could remain undetected. Internal, orthogonal validation (e.g. progeny sequencing performed on the same F₁ plants used for single-cell assays) would provide a more direct ground truth and avoid potential circularity in bias assessment.
4. The benchmark does not evaluate the impact of sequencing depth across modalities, which could influence the variation in per-barcode fragment counts and genomic bin coverage between scRNA-seq, scATAC-seq, and scWGA. Down-sampling aligned reads or informative fragments to fixed per-barcode targets (e.g. 250, 500, 1000 informative fragments) within each modality would clarify how much of the observed performance gap is attributable to depth rather than modality-specific biology or library structure. Constructing depth-matched subsets between scWGA and scATAC/scRNA datasets would help to test whether the breadth vs. depth trade-offs persist when sequencing resources are equalized.
5. In the pooled 34-hybrid single-nucleus RNA-seq dataset, it would be very informative to present detection sensitivity and resolution across genotypes (e.g. captured nuclei, distributions of informative fragments, covered bins, and expected localization error by genotype). Genotypes will differ in expression patterns, which will alter the number and distribution of informative fragments per nucleus, and thus ultimately influence inferred recombination rates and crossover terminality. Furthermore, the background subtraction filter relies on genotype-level background models. Given that all genotypes were pooled prior to nuclei isolation, can the authors show that estimated ambient/background profiles are comparable across genotypes?

Minor comments

1. The manuscript currently attributes more uneven coverage in scRNA-seq primarily to expression-biased sampling of heterozygous sites. Would the choice of using nuclei, rather than whole cells which would also allow the capture of cytosolic RNA, for the scRNA-seq be an additional reason for lower total number and genomic dispersion of informative fragments?
2. The sentence "This allows informed experimental and analytical choices ..." could be accompanied with a compact infographic or table (for example as an extension of Fig. 1B) summarizing key trade-offs and recommended use-cases for each modality (throughput, per-cell resolution, coverage breadth, susceptibility to doublets/ambient RNA, recommended applications).
3. Related to the point above, the choice to profile the F₁ hybrids using the 10x scRNA-seq modality is understandable from a throughput perspective, but the results presented in Fig. 1 and Table 1 suggest scWGA offers higher crossover accuracy, scATAC superior genomic breadth, compared to 10x scRNA-seq which in addition also showed a high doublet rate. Expanding the rationale for prioritizing scRNA-seq here (e.g. cost, compatibility with downstream expression analyses, or technical constraints for scWGA/scATAC at this scale) would clarify the experimental logic for the reader.

Referee cross-commenting

I strongly agree with the points raised by Reviewers #1 and #2. In particular, including additional replicates (ideally derived from the same pollen pool, processed identically and run across all modalities) would provide robustness to the benchmark. However, repeating these experiments, re-running the benchmark, and updating the interpretation would require substantial additional time, likely exceeding the suggested 1-3 month revision timeframe proposed by the other reviewers. Additional clarification of the analysis and representation of requested details (e.g. the recombination landscape plots (Reviewer #1), clarification of balanced pollen representation from each F₁ during pooling (Reviewers #2 and #3), and evaluation of how varying filtering strategies (e.g. doublet detection thresholds) affect the observed recombination patterns (Reviewers #2 and #3)) would also improve evaluation and transparency of the study. From a technical perspective major point 3 raised by Reviewer #2 (including information on the intrinsic biological characteristics of the material in the modality performance analysis) would provide substantially important context for users and improve interpretation of the benchmark.

Significance

Previous studies have successfully applied single-cell whole-genome amplification and linked-read sequencing to individual gametes to measure recombination rates and distributions, demonstrating the feasibility of this high-throughput alternative to progeny sequencing. This study extends that concept by delivering open-source pipelines for multiple single-cell modalities and by directly comparing the performance of scRNA-seq, scATAC-seq, and scWGA for mapping meiotic recombination in Arabidopsis gametes, offering both a practical resource and a performance evaluation for plant single-cell genomics.

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 presents coelsch, a cross-platform computational framework for single-cell gamete recombination analysis. It systematically benchmarks the performance of four mainstream single-cell sequencing modalities in meiotic crossover detection, successfully applies the method to a natural variation panel of Arabidopsis thaliana, and identifies the largest natural inversion reported in this species to date. This work demonstrates strong innovation, a complete technical pipeline, and significant biological implications. I would like to recommend revision. My concerns are listed below for the authors' consideration and revision.

Major concerns

1. Biological Replicates and Batch Effect Control The number of biological replicates per sequencing modality is limited (2 libraries for 10x scATAC and Takara scWGA, 1 library each for 10x scRNA and BD scRNA), and experiments for different modalities were performed in separate batches. Have the authors evaluated the impact of inter-batch technical variation on recombination rate estimates? In particular, for platforms with drastically different doublet rates (e.g., 49.7% for 10x scRNA vs. 26.3% for BD scRNA), how did the authors distinguish or avoid inherent platform differences from batch effects?

The natural variation analysis used a pooled library strategy for 40 F₁ hybrids without biological replicates. How did the authors ensure balanced pollen representation of each F₁ during pooling? For the 6 F₁ hybrids excluded due to insufficient data, was this due to initial pooling bias or sequencing capture preference? Could this introduce systematic bias into the natural variation analysis results? 2. Consistency of Pollen Nuclei Isolation Methods Different nuclei isolation protocols were used for each sequencing modality: Percoll density gradient centrifugation for 10x scATAC, no Percoll purification for Takara scWGA, and flow cytometry sorting combined with 10x/BD scRNA. Have the authors assessed how these different isolation methods affect nuclei integrity, viability, and capture bias for pollen nuclei? For example, could flow cytometry sorting selectively exclude nuclei of specific sizes or densities, thereby compromising the representativeness of recombination rate estimates? 3.Systematic impact of the inherent structure of pollen on different sequencing modalities Mature Arabidopsis thaliana pollen has a canonical trinucleate structure, consisting of one transcriptionally hyperactive vegetative nucleus and two sperm nuclei with highly condensed chromatin and almost complete transcriptional silencing. While all three nuclei share identical genome sequences, they exhibit fundamental differences in chromatin state and molecular features, which will have profoundly distinct effects on different sequencing modalities-an issue not addressed or controlled for in this study.

Differential technical capture bias: scRNA-seq and scATAC-seq rely on mRNA and accessible chromatin signals, respectively, and thus theoretically can only capture valid data from vegetative nuclei; sperm nuclei will be filtered out during quality control due to insufficient signal. In contrast, scWGA is based on whole-genome DNA amplification, independent of transcriptional activity or chromatin state, and can capture both vegetative and sperm nuclei. Have the authors validated the actual nuclear type composition in datasets from each modality through experiments (e.g., nuclear size sorting, DAPI staining quantification, immunofluorescence labeling)? Could this systematic difference in nuclear type composition compromise the fairness of performance comparisons between modalities? The uneven coverage of scRNA/scATAC is primarily determined by gene expression levels and chromatin accessibility (e.g., high coverage at highly expressed genes, extremely low coverage at heterochromatic regions such as centromeres), whereas coverage bias in scWGA mainly stems from technical preferences of whole-genome amplification. When comparing the resolution and accuracy of recombination detection across modalities, did the authors clarify the contributions of "intrinsic biological characteristics of nuclear types" from "technical characteristics of the sequencing technologies themselves"? 4. Accuracy and Validation of Doublet Detection Method This study reports exceptionally high doublet rates (~49% for 10x scATAC, ~70% for Takara scWGA), and there is a significant discrepancy with the results from Takara's official CellSelect software (80% of wells labeled "Good" by CellSelect were classified as doublets by coelsch). Have the authors validated the false positive and false negative rates of coelsch's doublet detection method through independent experiments (e.g., mixing pollen of known genotypes, manual microscopic validation of selected wells)? Such a high doublet filtering rate leads to a drastic reduction in the number of effective cells (e.g., only 628 singlets remained from a total of 2081 barcodes in the two Takara scWGA libraries). Have the authors assessed the representativeness of the remaining cells after filtering? In particular, for low-coverage scRNA data, could filtering result in the loss of cells with specific recombination patterns? 5. Depth and Breadth of Natural Variation Analysis This study finds significant differences in recombination rate and terminal crossover enrichment among different natural accessions, with Cvi-0 hybrids exhibiting higher overall recombination rates but lower terminal recombination rates. Have the authors further explored the genetic basis underlying these differences? Besides the 10 Mb inversion in Zin-9, did the authors identify similar large structural variations in other natural accessions? What is the sensitivity and specificity of the recombination coldspot-based method for detecting structural variation? For example, what is the minimum size of inversions or translocations that can be reliably detected?

Minor concerns

• The mutants used in this study (zyp1, figl1, recq4ab, etc.) were generated by crossing mutant lines in the Col-0 background with corresponding mutant lines in the Ler background, resulting in heterozygous F₁ backgrounds. For example, the zyp1 mutant used Col-0 background zyp1-1 and Ler background zyp1-6. Could this heterozygous mutant background affect the accurate measurement of meiotic processes and recombination rates? Have the authors considered validation using F₁ populations from homozygous mutant lines?
• The Takara scWGA dataset for wild-type Col-0 × Ler contains only 224 high-quality nuclei, while mutant sample sizes range from tens to hundreds. Is this sample size sufficient for fine-scale analysis of recombination rate distributions, especially for the detection of low-frequency recombination events? There are also a few minor issues regarding the references-some appear to be duplicates, such as references 11 and 31, which seem to be the same in both the published version and the bioRxiv preprint. Please double check. Additionally, have the authors considered the cost implications of these single-cell-based technologies, as well as their previously published linked-read sequencing approach?

Overall, this manuscript represents an important technical breakthrough in the field of meiotic recombination research, providing a unified computational framework for large-scale, cross-platform single-cell gamete recombination analysis. The above questions mainly focus on the rigor of experimental design (especially the omission of the unique biological issue of pollen trinucleate structure), the depth of computational method validation, and the expansion of biological findings, and do not affect the core conclusions of the manuscript. I suggest that the authors address these questions and provide clear responses in the revised manuscript. If these issues are properly resolved, this work will provide a powerful tool for investigating the genetic and molecular mechanisms of plant meiotic recombination.

Referee cross-commenting

I agree with Reviewers 1 and 3. Addressing most of the points we raised would bring this manuscript to publication standard.

Significance

This study develops a unified computational framework for meiotic crossover (CO) mapping using single‑cell sequencing of Arabidopsis pollen, benchmarks four single‑cell modalities, and identifies natural recombination variation and a large novel pericentric inversion. Overall, the work is technically sound, biologically meaningful, and fills a key gap in scalable gamete‑based recombination profiling.

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:

Parker et al. present coelsch and coelsch_mapping_pipeline, two open-source tools for platform-agnostic haplotyping and crossover detection from single-cell sequencing data, benchmarked across four modalities: 10x scATAC, 10x scRNA, BD scRNA, and Takara scWGA. The study applies these tools to Arabidopsis thaliana F₁ pollen to recover known recombination frequencies, characterise the effects of coverage sparsity via simulation, and profile natural variation in crossover rate and distribution across 34 F₁ hybrids from 22 diverse accessions. As a by-product of the recombination maps, the authors identify a previously unrecognised ~10 Mb pericentric inversion in the accession Zin-9 - the largest natural inversion described to date in A. thaliana.

This is an interesting and important study and is suitable in scope and rigour for publication in a Review Commons affiliate journal. By combining computational and experimental framework, the authors address a genuine methodological gap: while single-cell gamete sequencing is a powerful approach for recombination mapping, the consequences of choosing among available sequencing modalities have not been systematically evaluated. The tools are open-source, data are deposited, and the biological conclusions are well-grounded. Importantly, the limitations of the tools are also mentioned, which is appreciated. Therefore, this manuscript presents a genuinely useful methodological framework that fills a real gap in the recombination biology toolkit. The biological discovery (Zin-9 inversion) adds independent value. However, several analytical choices require better justification, some results sections are under interpreted, and a number of presentation issues should be addressed before acceptance.

Major comments:

1. Mismatch between best-performing modality and diversity panel application

The most critical concern is a logical inconsistency in the experimental design. The authors demonstrate convincingly that Takara scWGA achieves higher per-cell resolution and more accurate crossover detection than the droplet-based RNA methods. Yet the diversity panel - the study's key biological application - is analysed exclusively using 10x scRNA. No comparison with other modalities is provided for the panel, and no external recombination data for these accessions are included for validation. The authors should either: (i) include at least a subset of accessions profiled by an additional modality; or (ii) provide a more thorough quantitative justification for why 10x scRNA throughput outweighs the loss of resolution in this specific context, showing that cross-accession comparisons remain interpretable at scRNA coverage levels. 2. Could variation in crossover terminality result from analysis artefacts?

The authors demonstrate consistently higher rates of terminal crossovers in Col hybrids than in Cvi hybrids, 'implying genetic background modulation of crossover localisation'. However, their simulation analysis also demonstrates that telomere proximal crossovers are disproportionally missed in 10x RNA data. Therefore, could the Col vs. Cvi terminality differences result from a greater/lower occurrence of false negatives in different genotypes using this approach, rather than bona fide differences in CO number (caused by e.g. differences in telomere proximal marker density in Col vs. Cvi)? If so, this should be explicitly mentioned.<br /> 3. Doublet rates in Takara scWGA are unexplained

The Takara iCELL8 platform implements microscopy-based automated well selection to prevent doublets, yet coelsch identifies a ~70% doublet rate in these libraries. This is mentioned briefly but not adequately explained in the main text. The authors should provide a more thorough explanation for why the CellSelect imaging software fails to exclude pollen nuclei doublets (likely due to small nuclear size), and they should discuss what this implies for the utility of this platform for future experiments. This is important practical information for readers considering the Takara workflow. 4. Recombination landscape figures are incomplete

Figure 2C shows recombination landscapes only for mutant genotypes profiled by Takara scWGA. Equivalent per-chromosome landscape plots should be provided for all modalities tested on wild-type Col-0 × Ler material. This is essential to visually communicate the coverage-driven differences in landscape resolution that the authors describe, and to verify that 10x scATAC and scRNA recover similar gross distributions despite lower per-cell depth. 5. Extreme crossovers number in 10x scATAC are not discussed

The violin plots in Figure 2A show that 10x scATAC produces a wider upper tail of estimated crossover numbers than other modalities, with some barcodes exceeding 20 crossovers per nucleus - values far above the biological expectation for Arabidopsis. This is not acknowledged or explained. Is this an artefact of the high doublet contamination in this dataset (even after filtering), or a property of the HMM applied to fragmented ATAC data? An explicit discussion or supplementary analysis is required. 6. Resolution of crossover detection is undereported

Figure 3C shows boxplots of crossover localisation error across modalities, but this analysis is not discussed quantitatively in the main text. Readers need to understand the practical resolution (in kb) achievable by each modality in terms of crossover interval size. This is particularly important because the paper claims applicability for genetic mapping experiments, where localisation precision directly determines utility. 7. Telomeric false-negative rate in scWGA is not reported

The simulation analysis of false negatives near telomeres (Figure 3B) is presented only for 10x RNA data. Given that the authors use Takara scWGA for mutant genotyping and claim higher sensitivity, it is critical to also show the telomeric false-negative profile for scWGA. The current text implies that scWGA should avoid this problem, but this is not demonstrated. 8. Comparison between libraries from the same modality is absent

Two independent 10x scATAC and two Takara scWGA libraries were generated, but no within-modality reproducibility analysis of crossover rates or landscapes is presented. Crossover rates and landscape correlations between technical replicates should be shown to establish that the observed modality-level differences are not driven by library-preparation variability. 9. Applicability to non-Arabidopsis and heterozygous species

The Discussion notes that the approach relies on isogenic founder crosses and high-quality parental assemblies but does not explore the practical barriers to applying coelsch in outcrossing or polyploid species. Given the broad framing of the title ('platform-agnostic'), the authors should discuss what adaptations would be needed for crop species or other organisms where chromosome-scale haplotype-resolved assemblies are not available.

Minor comments:

1. Figure 5B - Please add axis labels in Mb.
2. Figure 2A - library replicates: The two 10x scATAC libraries are not differentiated in Figure 2A. Showing them separately (or indicating per-library medians) would improve transparency.
3. Droplet vs. plate combination: The Discussion does not address whether complementary modalities could be combined (e.g., using droplet-based data for landscape estimation and scWGA for localisation refinement within the same experiment). A brief discussion of this possibility would strengthen the practical utility of the framework.

Referee cross-commenting

All points raised by reviewers 2 & 3 seem reasonable and would substantially improve the quality of the manuscript

Significance

General assessment: The paper from Parker et al., provides the first systematic evaluation of single-cell sequencing modalities for recombination mapping in Arabidopsis and presents new bioinformatic tools for analysing recombination in single-cell data. The novel utility of the approach is demonstrated for assessing recombination rate across a wide variety of Arabidopsis hybrids. Different platforms provide different benefits/limitations and these are well presented. However, the manuscript would benefit from a more thorough presentation of all the different analyses that were performed.

Advance: Most recombination mapping studies in Arabidopsis utilise progeny sequencing. Here, the authors present an alternative approach, using single-cell gamete sequencing which will more easily facilitate recombination mapping in large populations, which will be particularly useful for future studies investigating the influence of natural variation on recombination rate and location. The advance is mostly technical, but the study also generates novel biological observations about chromosome structural rearrangements in Arabidopsis.

Audience: The study is likely to be of main interest to individuals studying recombination in plants (particularly using bioinformatic approaches and analysing the influence of natural variation). However, researchers with an interest in single-cell sequencing and broader genomics will also be an audience for this paper.

Describe your expertise:

I am a researcher in plant meiotic recombination and I am well placed to assess the general importance and impact of the study within the context of the field. However, I would not consider myself a specific expert in 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

RESPONSE TO REVIEWERS

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

Summary:

This is an interesting and ambitious study by Tabilo-Agurto and co-workers. It combines deep learning structure prediction (AlphaFold2), targeted molecular dynamics simulations, and in vivo functional assays to probe structural, functional, and evolutionary aspects of the metamorphic protein RfaH. More broadly, the work addresses an important question: whether intermediate structural states may exist along evolutionary trajectories of metamorphic proteins. A particular strength of the study is the integration of computational and experimental approaches. The manuscript is generally well written and clearly organized.

Major comments:

A key aspect of the study is the classification of predicted structures into three classes based on the conformation of the C-terminal domain (CTD): the autoinhibited alpha-helical fold, a beta-barrel fold, and a mixed alpha/beta fold. These classes are further described as corresponding to metamorphic (alpha fold), mixed alpha/beta, and monomorphic (beta fold) proteins.

While I can see how this organizational scheme is helpful in some respects, it may also overstate what can be concluded from the data. As the authors are well aware, AlphaFold2 tends to predict a single conformation even for genuine metamorphic proteins, and therefore does not, on its own, distinguish between monomorphic and fold-switching proteins. I note in particular that the functional data indicates that the "monomorphic" variants studied in the in vivo assays behave similarly to the RfaH E48A mutant. However, E48A is known to remain metamorphic, populating both alpha and beta folds with roughly equal probability. This suggests that the sequences in this class may retain some degree of fold-switching capability, even if the underlying regulatory mechanism differs from that of wild-type RfaH. In other words, the presented data does not fully support these sequences as monomorphic. I am not suggesting that the authors must revise their classification scheme. However, it may strengthen the manuscript if the authors explicitly acknowledge this alternative interpretation and moderate the corresponding claims.

We appreciate the comment from the reviewer, which can be seen from two different perspectives.

On the one hand, it might be reasonable to think that the ‘monomorphic’ RfaH orthologs have lower transcription elongation activity than E. coli RfaH. Other highly divergent orthologs of E. coli RfaH (Salmonella enterica serovar Typhimurium, Klebsiella pneumoniae, Yersinia enterocolitica and Vibrio cholerae) have similar in vitro recruitment and pausing at the C45 nucleotide from the ops element, as well as restoring the RfaH-dependent hemolytic activity of E. coli in a strain that lacks chromosomal RfaH to levels similar to the wild-type strain (doi: 10.1128/jb.186.9.2829-2840.2004). However, V. cholerae RfaH (43% sequence identity to E. coli RfaH) exhibits diminished antitermination effects in in vitro transcription assays, better resembling the antitermination levels in the absence of RfaH (doi: 10.1128/jb.186.9.2829-2840.2004), despite this protein also being predicted in the alpha-folded state when using AF2 (10.1016/j.csbj.2022.10.024). A particular observation from the RfaH complementation work is that increasing the concentration of V. cholerae in in vitro transcription assays lessens the transcription elongation effects observed when using concentrations similar to E. coli RfaH. These transcription elongation defects can be extrapolated to potentially similar issues with transcription in vivo and, therefore, luciferase translation in our in vivo translation assays for our ‘monomorphic’ proteins.

On the other hand, it is possible that these so-called ‘monomorphic proteins’ still populate the alpha-folded state, but that their predominant fold in solution is the one corresponding to the active beta-fold. This can be biophysically tested using circular dichroism to distinguish their alpha or beta propensity, as proposed in a remarkable work from Porter et al (10.1038/s41467-022-31532-9).

In both cases, quantification of the protein titers obtained after attempts of protein purification of the ‘monomorphic’ RfaH orthologs would be required. In this way, we can ascertain whether the differences in activity are due to differences in expression levels and determine if sufficient amounts of stable and well-folded protein can be obtained for these RfaH orthologs, followed by measuring their circular dichroism spectra to ascertain their secondary structure propensity.

Our current attempts are to recombinantly express these proteins for determining their protein titers and solubility in the supernatant, which will enable us to indirectly ascertain their expression levels, and test those solubly expressed proteins biophysically using circular dichroism experiments. If the circular dichroism experiments prove to be unsuccessful due to problems with the solubility of the purified proteins, we strongly believe that the aforementioned discussion should be included in the manuscript to take into account the limitations of the methods utilized in our work.

Therefore, we will add the following paragraph in the discussion, while we work on ascertaining the feasibility of the circular dichroism assays:

“It is worth noting that, in the absence of RBS (Figure 5C-F), the putative monomorphic RfaH orthologs have similar or lower in vivo activity than the E. coli RfaH E48A mutant; a similar mutant (E48S) exhibits a 1:1 equilibrium between the autoinhibited and active states (Burmann et al, 2012). This observation can be partly explained by two factors. First, sequence divergence and expression levels may limit functional compatibility with the host machinery. Highly divergent V. cholerae RfaH ortholog, which shares only 43% sequence identity with E. coli RfaH but is predicted to fold into the autoinhibited state (Artsimovitch & Ramírez-Sarmiento, 2022), maintains both ops-dependent recruitment and hemolysin secretion in the ∆rfaH E. coli strain, yet exhibits transcription elongation defects in vitro, requiring a 5-fold higher concentration than E. coli RfaH to match increased elongation rates of E. coli RNAP (Carter et al, 2004). Low in vivo protein titers or structural mismatches between the monomorphic orthologs and E. coli RNAP may prevent higher luciferase expression relative to the E48A mutant. This limitation is supported by the fact that IPTG-induced overexpression rescues activity when an RBS is present (Figure 5B). Second, these proteins may be predominantly folded in the active state while still transiently populating the autoinhibited state. Confirming this conformational equilibrium would require overexpression and purification of these proteins followed by biophysical assays, such as circular dichroism (Porter et al, 2022).”

Reviewer #1 (Significance (Required)):

An intriguing, but speculative, aspect of the study is the finding that some sequences are predicted to adopt a CTD with mixed alpha/beta secondary structure, and that such structures also appear in targeted molecular dynamics simulations. If this idea holds up, it could represent an intermediate along the evolutionary pathway between the alpha-helical and beta-barrel folds of RfaH. Although the evidence is only computational, it is a compelling idea and it would benefit from further investigation.

It is indeed very compelling, and this is something that we should immediately address in a revised version of our manuscript. We somehow missed an article published in 2025, regarding the study of the structural interconversion of the isolated CTD using NMR, finding at least three intermediate states along the fold-switching pathway of RfaH (doi: 10.1073/pnas.2506441122). One of such intermediate states observed, which is also one of the highest populated ones (~23%), corresponds to an ensemble of largely unfolded structures that include the formation of transient alpha-helix a5 (corresponding to helix a2 in our article) and beta-hairpin (b1/b2) secondary structure elements, which fold to form a compact ensemble of structures in which the beta-hairpin lies on top of the alpha-helix. This is fully consistent with our predictions of a mixed alpha/beta state in full-length.

We will add this external experimental validation of the mixed alpha/beta secondary structure of the CTD of RfaH in the discussion of our final manuscript:

“Interestingly, a recent nuclear magnetic resonance spectroscopy study of the E. coli RfaH CTD, aimed to uncover transient states potentially en route of the αCTD interconversion (Cai et al, 2025), described an intermediate state (populated in ~23% of the captured ensembles) in which a β-hairpin formed by β-strands β1-β2 lies on top of a transient α-helix α5 that corresponds to helix α2 in our article. This finding is fully consistent with the mixed α/β CTD structures found both in our TMD simulations and our AF2 predictions of divergent RfaH orthologs.”

In summary, the work is a valuable contribution to the field of protein fold switching. The combination of computational tools with experimental validation makes it interesting and the results should be of broad interest. The manuscript should be well positioned for publication in a high-impact journal.

We are very thankful for the reviewer’s comments on our manuscript.

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

Summary:

In their paper "Exploration of the structural and functional diversity in the metamorphic RfaH subfamily," Tabilo-Agurto et al. use AlphaFold2 to predict the structures of ~3,900 RfaH homologs, sort the predicted C-terminal domains into α-helical (autoinhibited), β-barrel (NusG-like), and mixed α/β topologies, and find that about 14% of homologs come out predominantly in the β-barrel state. They then take nine representative homologs and run them through a heterologous *E. coli* DH5α Δ*rfaH* reporter assay. The putative monomorphic candidates behave a lot like the constitutively active E48A variant - active across every ops context and even without an RBS - while the mixed α/β candidates barely show activity. Targeted MD simulations of *E. coli* RfaH, run through AF2Rank, also pulls out the mixed α/β state as its own distinct cluster, hinting that it sits somewhere along the fold-switching transition path.

This is a genuinely interesting piece of work that pulls together structure prediction, in vivo activity, and genomic context to make a concrete case for extant monomorphic βRfaH proteins - a long-hypothesized but until now unseen intermediate in the proposed stepwise evolution of RfaH from NusG. The experimental design is thoughtful, especially the five-construct ops/RBS matrix, and comparing the monomorphic candidates against the E48A benchmark is a nice touch as a positive control. Overall, I think the paper deserves to be published, but a few things would need shoring up before acceptance.

Major comments:

1. The paper would be a lot stronger with at least one biophysical measurement (a CD spectrum, say) on a purified monomorphic candidate. I get that this might be outside the planned scope, but even a single CD trace showing β-rich content for an isolated full-length protein would move the claim from "putative" to "demonstrated."

We agree with the comment from the reviewer, and as such we are currently attempting to recombinantly express these proteins for determining first their solubility after purification (which will largely determine our ability to characterize them by circular dichroism) and then follow up with circular dichroism experiments if the solubility and protein concentration of these ‘monomorphic’ homologs is sufficient to pursue these experiments. In case this is unfeasible, we will include the solubility analysis in our revised version of the article, as well as a discussion on this topic – and also on the topic of why the activity of the ‘monomorphic’ proteins resembles the E48A mutant of E. coli RfaH that co-exists between two folds – as indicated in our response to the major comment from reviewer #1.

1. Only nine homologs were tested - three per category. The conclusions about monomorphic behavior generalizing across the whole βRfaH clade are basically resting on three proteins. Bringing in even one or two phylogenetically distant βRfaH candidates would help guard against the possibility that what they're seeing is just a genus-specific quirk. If new experiments aren't on the table, the limitation should at least be called out explicitly in the Discussion.

We agree with the reviewer that drawing conclusions from a single clade of RfaH could raise concerns about bias, although we must note that the tested putative monomorphic candidates were selected before a phylogenetic tree was constructed. What we propose is to perform a phylogenetic analysis for the InterPro sequences and look at their genomic neighborhood as well, replicating what was done in the manuscript for the Genomic Cluster group. We hope this would provide more compelling evidence that the predictions, phylogeny and gene organization of these extant monomorphic RfaH is distinct from those metamorphic.

1. The classification thresholds (α > 32.5% / β 30.0% / α

Thanks to the reviewer for raising this concern. We will perform a sensitivity analysis by slightly nudging the cutoffs by ±5% as recommended by the reviewer and indeed we see minimal changes in the number of structures in each class. We have added a small paragraph indicating this sensitivity test:

“To determine that these values were adequate for our analysis, we performed a sensitivity test by changing the thresholds by ±5% over the data for all structures predicted from all databases, showing that the predictions of RfaH orthologs with monomorphic CTD and mixed secondary structure in their CTD is robust, and only metamorphic RfaH orthologs were reduced with an increase in uncategorized structures (Supplementary Figure S13)”

1. The Discussion notes that uncontrolled, ops-independent RfaH recruitment could be lethal, since RfaH outcompetes the much more abundant NusG. But if monomorphic RfaH proteins really are extant and stably maintained in these genomes, there has to be something keeping them from interfering with NusG's essential functions - maybe very low expression, restricted induction, or compensating differences in NusG affinity. The paper would benefit from tackling this directly, even speculatively.

We agree with the reviewer in this point, and after careful consideration we believe that we did not emphasize this point appropriately in the manuscript. In fact, we included Figure 7 to state our perspective on how RfaH may have evolved but we did not emphasize how this perspective stems from a previous work that we thoroughly discussed in the introduction (doi: 10.1038/emboj.2008.268) and that explicitly states that low solubility of the dissociated NTD and CTD could be a factor imposing this restricted action in cis operons. We have included this in our revised version of the manuscript as follows:

“Our findings are in line with the previous hypothesis regarding the emergence of RfaH within the universally conserved family of NusG transcription factors (Belogurov et al, 2009). Under that model, a gene duplication event produced an intermediate variant (NusG2 in Figure 7) that lost its Rho-binding capability and acquired a deletion in the NTD that reduced the protein's overall size and remodeled its hydrophobic profile. Crucially, this intermediate retained an exposed, hydrophobic RNAP-binding region, a feature shared by monomorphic RfaH and the ancestor of all RfaH orthologs (NusGSP in Figure 7). This increased hydrophobicity would have reduced solubility, restricting its regulatory activity to the site of synthesis, i.e. in cis. Indeed, when structural alignment is used to identify conserved NTD residues that bind to RNAP, orthologs contain more than 70% hydrophobic residues (Supplementary Figure S11) at those positions. This percentage is much closer to that of RfaH (80%) than NusG (57.14%). The protein only regains solubility and the ability to operate in trans when its CTD refolds into a helical conformation. Ultimately, our results strengthen this evolutionary model by demonstrating that several extant RfaH orthologs appear to resemble this insoluble, cis-acting ancestral state.”

Minor comments:

Table 1 should show percentages alongside the raw counts. 7/7 LPS-in-operon for monomorphic candidates is striking, but with n=10, the small denominator really deserves to be flagged.

We agree with the reviewer in that the higher raw count of metamorphics may undersell the message the article conveys. We added the percentages next to raw counts in Table 1, regarding “Total” and “Next to operon” categories. We also modified the legend as follows:

“A summary of genomic contexts of RfaH orthologs classified according to the AF2 predictions. The numbers indicate how many rfaH genes are next to an operon and whether the operon contains lipopolysaccharide biosynthesis genes, and the percentages next to them display the relation to the previous category, i.e, “Next to operon”/”Total” and “LPS in operon”/”Next to operon”.”

In Figure 3, the sequence logo on top is informative - consider adding the number of sequences per dataset to the axis labels so readers can interpret the boxplot widths.

We believe that this would be rather confusing for the readers, because it is counting all 5 structures predicted by AlphaFold2 for each sequence in each dataset that fit each classification, and thus the same sequence can lead to structures that are monomorphic, metamorphic of have mixed secondary structure in their CTD. Thus, the number of sequences per box plot will be higher than the number of sequences per dataset. For example, one sequence from InterPro can be present in more than one box plot, because different AlphaFold2 models can lead to the prediction of different states from the same sequence. We believe it is less confusing if it is presented as it is.

There's some redundancy between the Results (pp. 14-17) and the Discussion that could probably be trimmed, particularly the recap of the ops/RBS construct logic.

Thanks for the recommendation. We reduced this redundancy in the new version of the manuscript, mainly on page 16:

“The orthologs classified as monomorphic, and thus expected to be constitutively active, exhibited activity across all tested ops contexts, including in the absence of RBS (Figure 5B-F). Notably, their activity levels were comparable to the ops-independent E. coli RfaH E48A mutant, in which the key salt bridge at the NTD:CTD interface is disrupted. All monomorphic orthologs were found to lack a few key residues that make contacts to ops DNA in RfaH, as well as the conserved residues in loop 2 that mediate contacts with Rho in NusG (Supplementary Figure 11). This mosaic architecture enables these orthologs to promote the expression of the long lux operon even when the RBS is absent. Our study provides the first indirect evidence of putative, constitutively active RfaH proteins, which are predicted to have monomorphic NusG-like fold, in other bacteria.”

Reviewer #2 (Significance (Required)):

If the central claim holds up, this is a meaningful contribution to the metamorphic-protein and bacterial-transcription literatures: it identifies what appear to be extant evolutionary "way-stations" in the NusG→RfaH transition, and it does so using a tractable computational pipeline that could be applied to other suspected fold-switch families. The work is timely given the ongoing discussion about how AF2 and its descendants handle conformational heterogeneity. With the strengthening suggested above - particularly any direct biophysical confirmation of a monomorphic candidate - I would expect this to be a well-cited paper in its niche.

We are very thankful for the reviewer’s comments on our manuscript.

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

Summary:

The manuscript by Tabilo-Agurto et al. uses in silico and experimental methods to elucidate the diversity of the metamorphic RfaH protein family. Of particular note is the sophisticated usage of AlphaFold2 to reconstruct the evolutionary tree of RfaH as well as the in vivo luminescence assays to substantiate the different structural states of the RfaH-CTD. Overall this is a well-written manuscript providing deeper insight into the structural and functional diversity of RfaH proteins, potentially relevant for other metamorphic proteins as well.

Minor comments:

1. 3rd paragraph of the introduction: The sentence starting with "To date, ...and nuclear magnetic resonance of these ancestors.." seems incomplete as this reviewer believes the author´s wanted to say "..and structural characterization by nuclear magnetic resonance spectroscopy of these ancestors..."

Thanks for the attention to these details, we will amend this paragraph appropriately.

1. 4th paragraph of the introduction: "..., that binds Rho or the ribosome (Mooney et al. 2009b). Whereas this citation is correct for NusG-Rho interactions it does not indicate ribosome binding. The direct interaction of NusG with the ribosome was shown in Burmann et al. Science 2010 and this reference should be added here.

Thanks for the recommendation, we will include both citations in this section of the manuscript.

1. More a curiosity question, did the author also test for a subset of the RfaH variants the AlphaFold3 predictions and obtain similar or different results?

Thanks for the comment. The reason we did not use AlphaFold3 predictions to check on the variability of the results is that there is much more known about the use of AlphaFold2 – and its limitations – regarding their use in the study of metamorphic proteins, whereas a deep understanding of the advantages and limitations of AlphaFold3 for studying metamorphic proteins is still under development.

Referees cross-commenting:

Overall there is an agreement among all reviewers that the present MS is an interesting and timely study. The point raised by reviewer 2 to add simple biophysical characterization, if feasible, would be clearly an excellent addition and likely make the MS stronger. In general all three reviewers mainly point to minor changes and additions to improve the MS in a rather short timeframe.

We indeed agree with this comment, which is why we will commit to attempt the recombinant expression and protein purification of the RfaH orthologs and to perform circular dichroism assays if the solubility of the obtained proteins allows for such experiments to be done.

Reviewer #3 (Significance (Required)):

The present MS is an interesting large-scale usage of the AlphaFold2 algorithm to reconstruct the evolutionary tree of the specialized transcription elongation factor RfaH. Revealing a different degree of this evolution in a diverse set of bacterial strains indicating its evolutionary distance from the cognate NusG transcription elongation factor. Of particular note is the experimental verification of the obtained in silico finding by in vivo luminescence approaches.

We are very thankful for the reviewer’s comments on our 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 #3

Evidence, reproducibility and clarity

The manuscript by Tabilo-Agurto et al. uses a in silico and experimental methods to elucidate the diversity of the metamorphic RfaH protein family. Of particular note is the sophisticated usage of AlphaFold2 to reconstruct the evolutionary tree of RfaH as well as the in vivo luminescence assays to substantiate the different structural states of the RfaH-CTD. Overall this is a well-written manuscript providing deeper insight into the structural and functional diversity of RfaH proteins, potentially relevant for other metamorphic proteins as well.

Minor Points:

• 3rd paragraph of the introduction: The sentence starting with "To date, ...and nuclear magnetic resonance of these ancestors.." seems incomplete as this reviewer believes the author´s wanted to say "..and structural characterization by nuclear magnetic resonance spectroscopy of these ancestors..."
• 4th paragraph of the introduction: "..., that binds Rho or the ribosome (Mooney et al. 2009b). Whereas this citation is correct for NusG-Rho interactions it does not indicate ribosome binding. The direct interaction of NusG with the ribosome was shown in Burmann et al. Science 2010 and this reference should be added here.
• More a curiosity question, did the author also tested for a subset of the RfaH variants the AlphaFold3 predictions and obtained similar of different results?

Referees cross commenting

Overall there is an agreement among all reviewers that the present MS is an interesting and timely study. The point raised by reviewer 2 to add simple biophysical characterization, if feasible, would be clearly an excellent addition and likely make the MS stronger. In general all three reviewers mainly point to minor changes and additions to improve the MS in a rahter short timeframe.

Significance

The present MS is an intereting large scale uage of the AlphaFold2 algorithm to reconstruct the evolutionary tree of the specialized transcription elongation factir RfaH. Revealing a different degree of this evolution in a diverse set of bacterial strains indicating its evolutionary distance from the cognate NusG transcription elongation factor. Of particular note is the experimental verification of the obtained in silico finding by in vio luminescence approaches.

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 their paper "Exploration of the structural and functional diversity in the metamorphic RfaH subfamily," Tabilo-Agurto et al. use AlphaFold2 to predict the structures of ~3,900 RfaH homologs, sort the predicted C-terminal domains into α-helical (autoinhibited), β-barrel (NusG-like), and mixed α/β topologies, and find that about 14% of homologs come out predominantly in the β-barrel state. They then take nine representative homologs and run them through a heterologous E. coli DH5α ΔrfaH reporter assay. The putative monomorphic candidates behave a lot like the constitutively active E48A variant - active across every ops context and even without an RBS - while the mixed α/β candidates barely show activity. Targeted MD simulations of E. coli RfaH, run through AF2Rank, also pulls out the mixed α/β state as its own distinct cluster, hinting that it sits somewhere along the fold-switching transition path.

This is a genuinely interesting piece of work that pulls together structure prediction, in vivo activity, and genomic context to make a concrete case for extant monomorphic βRfaH proteins - a long-hypothesized but until now unseen intermediate in the proposed stepwise evolution of RfaH from NusG. The experimental design is thoughtful, especially the five-construct ops/RBS matrix, and comparing the monomorphic candidates against the E48A benchmark is a nice touch as a positive control. Overall, I think the paper deserves to be published, but a few things would need shoring up before acceptance.

Major comments

1. The paper would be a lot stronger with at least one biophysical measurement (a CD spectrum, say) on a purified monomorphic candidate. I get that this might be outside the planned scope, but even a single CD trace showing β-rich content for an isolated full-length protein would move the claim from "putative" to "demonstrated."
2. Only nine homologs were tested - three per category. The conclusions about monomorphic behavior generalizing across the whole βRfaH clade are basically resting on three proteins. Bringing in even one or two phylogenetically distant βRfaH candidates would help guard against the possibility that what they're seeing is just a genus-specific quirk. If new experiments aren't on the table, the limitation should at least be called out explicitly in the Discussion.
3. The classification thresholds (α > 32.5% / β < 2.5% for αRfaH; β > 30.0% / α < 2.5% for βRfaH) are described as coming from histogram inspection, but they feel a bit arbitrary as stated. A quick sensitivity analysis - how do the population fractions shift if you nudge the cutoffs {plus minus}5%? - would help reassure the reader.
4. The Discussion notes that uncontrolled, ops-independent RfaH recruitment could be lethal, since RfaH outcompetes the much more abundant NusG. But if monomorphic RfaH proteins really are extant and stably maintained in these genomes, there has to be something keeping them from interfering with NusG's essential functions - maybe very low expression, restricted induction, or compensating differences in NusG affinity. The paper would benefit from tackling this directly, even speculatively.

Minor comments

Table 1 should show percentages alongside the raw counts. 7/7 LPS-in-operon for monomorphic candidates is striking, but with n=10, the small denominator really deserves to be flagged.

In Figure 3, the sequence logo on top is informative - consider adding the number of sequences per dataset to the axis labels so readers can interpret the boxplot widths.

There's some redundancy between the Results (pp. 14-17) and the Discussion that could probably be trimmed, particularly the recap of the ops/RBS construct logic.

Significance

If the central claim holds up, this is a meaningful contribution to the metamorphic-protein and bacterial-transcription literatures: it identifies what appear to be extant evolutionary "way-stations" in the NusG→RfaH transition, and it does so using a tractable computational pipeline that could be applied to other suspected fold-switch families. The work is timely given the ongoing discussion about how AF2 and its descendants handle conformational heterogeneity. With the strengthening suggested above - particularly any direct biophysical confirmation of a monomorphic candidate - I would expect this to be a well-cited paper in its niche.

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:

This is an interesting and ambitious study by Tabilo-Agurto and co-workers. It combines deep learning structure prediction (AlphaFold2), targeted molecular dynamics simulations, and in vivo functional assays to probe structural, functional, and evolutionary aspects of the metamorphic protein RfaH. More broadly, the work addresses an important question: whether intermediate structural states may exist along evolutionary trajectories of metamorphic proteins. A particular strength of the study is the integration of computational and experimental approaches. The manuscript is generally well written and clearly organized.

Major comment:

A key aspect of the study is the classification of predicted structures into three classes based on the conformation of the C-terminal domain (CTD): the autoinhibited alpha-helical fold, a beta-barrel fold, and a mixed alpha/beta fold. These classes are further described as corresponding to metamorphic (alpha fold), mixed alpha/beta, and monomorphic (beta fold) proteins.

While I can see how this organizational scheme is helpful in some respects, it may also overstate what can be concluded from the data. As the authors are well aware, AlphaFold2 tends to predict a single conformation even for genuine metamorphic proteins, and therefore does not, on its own, distinguish between monomorphic and fold-switching proteins. I note in particular that the functional data indicates that the "monomorphic" variants studied in the in vivo assays behave similarly to the RfaH E48A mutant. However, E48A is known to remain metamorphic, populating both alpha and beta folds with roughly equal probability. This suggests that the sequences in this class may retain some degree of fold-switching capability, even if the underlying regulatory mechanism differ from that of wild-type RfaH. In other words, the presented data does not fully support these sequences as monomorphic. I am not suggesting that the authors must revise their classification scheme. However, it may strengthen the manuscript if the authors explicitly acknowledge this alternative interpretation and moderate the corresponding claims.

Significance

An intriguing, but speculative, aspect of the study is the finding that some sequences are predicted to adopt a CTD with mixed alpha/beta secondary structure, and that such structures also appear in targeted molecular dynamics simulations. If this idea holds up, it could represent an intermediate along the evolutionary pathway between the alpha-helical and beta-barrel folds of RfaH. Although the evidence is only computational, it is a compelling idea and it would benefit from further investigation.

In summary, the work is a valuable contribution to the field of protein fold switching. The combination of computational tools with experimental validation makes it interesting and the results should be of broad interest. The manuscript should be well positioned for publication in a high-impact journal.

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-2026-03474

Corresponding author(s): Priyanka, Verma

[Please use this template only if the submitted manuscript should be considered by the affiliate journal as a full revision in response to the points raised by the reviewers.

• *

If you wish to submit a preliminary revision with a revision plan, please use our "Revision Plan" template. It is important to use the appropriate template to clearly inform the editors of your intentions.]

1. General Statements [optional]

Point-by-point rebuttal is presented below. Reviewer’s comments are in BLACK; author’s response is in BLUE and figure numbers corresponding to the manuscript are in RED.

2. Point-by-point description of the revisions

• *

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

ALC1 suppression has been shown to potentiate PARP inhibitor lethality in HR-deficient cells. Rather than revisiting the underlying mechanism, which has been characterized and remains an active area of investigation, this study aims to define the clinical contexts in which combined ALC1 and PARP inhibition may be beneficial. The clinical efficacy of PARP inhibitors, and their FDA approval, is largely restricted to HR-deficient tumors. This study dissects the combined effects of ALC1 and PARP suppression across a panel of HRD ovarian cancer cell lines, multiple classes of PARP inhibitor, and cells harboring distinct PARPi resistance mechanisms. In doing so, the authors delineate both the potential utility and the limitations of combined ALC1 and PARP inhibitor treatment in HRD ovarian cancers. The most impactful finding of the study, however, is likely the demonstration that ALC1 suppression sensitizes HR-proficient, CCNE1-amplified high-grade serous ovarian cancers to PARP inhibitors. These tumors are associated with particularly poor outcomes owing to the current absence of effective targeted therapies, making this observation of considerable clinical relevance.

We thank the reviewer for appreciating the significance of our work in “HR-proficient, CCNE1-amplified high-grade serous ovarian cancers to PARP inhibitors” which is a critical unmet need.

Of note, the study relies on genetic rather than pharmacological depletion of ALC1, a choice likely reflecting the current lack of a commercially available ALC1 inhibitor. While genetic suppression may not fully recapitulate the effects of combined drug treatment, it offers the advantage of not being tied to any specific compound, allowing the authors to establish more general principles. I have only a few comments.

We are grateful to the reviewer for providing the unique perspective on our genetic study that “it offers the advantage of not being tied to any specific compound, allowing the authors to establish more general principles.”

We have included this in our discussion to strengthen the study.

The effect of ALC1 KO on PARPi sensitivity is less pronounced in OVSAHO cells (BRCA2-mutated) than in BRCA1-mutated cells. In these cells, it looks like there is an additive effect rather than synergy. 1- The authors should calculate, if possible, whether there is synergy or additive effect of ALC1-KO lethality (BLISS).

We thank the reviewer for recognizing our limitations to perform BLISS score analysis, as our experiments were conducted at a single level of total protein depletion. Ideally, synergy assessments require a range of depletion levels to generate a full response matrix. Regardless, to address the reviewer’s concern regarding the impact of ALC1 on olaparib response in BRCA1- and BRCA2-mutant cells, we performed a BLISS score calculation under the conservative assumption that total ALC1 depletion alone has no effect on cell viability. We then employed the following formula for BLISS score calculation:

Bliss Score =Eobs- (EA+EB-EAX EB)

Where Eobs is viability of ALC1-depleted cells at a given drug concentration. This is observed impact upon combined loss of ALC1 and olaparib treatment.

EA is impact on viability upon ALC1 depletion only. This was considered to be zero.

EB is impact on viability on ALC1 WT in the presence of drug. This assesses the impact of drug alone.

BLISS score was calculated at all non-saturating drugs concentration and then averaged to obtain a final BLISS value. We used the following cut off:

> 10: Synergistic (the interaction is considered significant);

-10 to 10: Additive (no significant interaction);

__

Olaparib

Rucaparib

Niraparib

Veliparib

Cisplatin

UWB1.289

22.34

25.21

13.24

14.95334

0.26

JHOS-4

37.27

47.14

26.3

27.94

-0.37

OVSAHO

19.34

27.6

23.2

19.15

7.04

Kuramochi

11.38

11.98

-3.56

6.79

-0.39

We observe that ALC1 loss synergistically enhances olaparib and rucaparib response in both BRCA1- and 2-mutant cells. However, as correctly noted by the reviewer, we notice that the BLISS score is higher in BRCA1-mutant cells compared to BRCA-2 mutant, OVSAHO.

In the revised manuscript, we have also included data for another BRCA-2-mutant cell line: KURAMOCHI (Fig.1d; Supp. Fig1b). We chose this cell line because, despite having a BRCA2-mutation, it is highly resistant to PARP inhibitors and cisplatin, owing to KRAS amplification. Notably, we observe that ALC1 loss can synergistically enhance the response of Kuramochi to olaparib and rucaparib.

We have included a statement in the manuscript that the impact of ALC1 loss was more profound in BRCA1- versus BRCA2-settings. However, if acceptable to the reviewer, we would prefer not to include the BLISS values in the manuscript, as these calculations were not performed using the standard approach of titrating multiple levels of protein depletion.

2- Another BRCA2-mutated cell line should be included.

As discussed above, we have now included data from another BRCA2-mutant cell line, Kuramochi. Consistent with data in other BRCA-mutant cell lines, loss of ALC1 enhances olaparib and rucaparib sensitivity in these cells (Fig. 1d; Supp. Fig.1b).

Minor comments: • Figure key is missing for S2C (I assume it's grey DMSO, blue olaparib)

We apologize for this oversight. Figure key has now been included.

• Page 8: "BRCA1-mutant ovarian cancer cells eventually develop chemoresistance when exposed to PARPi for a prolonged period. Mechanistically, this is due to rewiring of ATR signaling, which enables RAD51 loading at DNA breaks and reversed forks independent of BRCA1 protein(25)." This sentence suggest this is the only existing resistance mechanism, which should be correct. Modify to "mechanistically, this CAN be due to", or "this is OFTEN due to".

We thank for the reviewer for suggesting this important correction. This has now been fixed.

Reviewer #1 (Significance (Required)):

ALC1 inhibitors have been developed and clinical trials are starting. The significance of this manuscript lies in establishing the clinical potential for combined ALC1-PARP inhibition in high grade serous ovarian cancer. Especially, the authors demonstrate that combined ALC1 suppression with PARP inhibition efficiently kills HR-proficient CCNE1-amplified ovarian cancers, which represent 20% of ovarian cancers and are resistant to current therapies.

__Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ The manuscript by Lindsey et al. explores the role of ALCN1 (Amplified in Liver Cancer 1) loss in enhancing the sensitivity of PARPi in ovariar carcinomas, including BRCA1/2 mutated tumors (both sensitive and resistant to platinum) as well as cyclin E amplified settings. The data are interesting but the in some cases there is an overinterpretation of the results. I have listed below my major concerns.

We appreciate that the reviewer finds our data interesting. We also appreciate the reviewer insightful comments and have addressed them below.

Figure 1. Could the authors demonstrate that OVASAHO cells are BRC2 muted? Indeed, I have always though they were BRCA wt type (10.1016/j.ygyno.2015.08.017).

OVSAHO cells have a homozygous deletion in the BRCA2 gene (PMID:23839242), which could be the reason why a mutation was not detected in the study referred to by the reviewer (PMID: 26321251). We have now included the Domcke et al; 2013 reference in manuscript. The loss of BRCA2 expression in OVSAHO is also evident in our blots (Fig. 1a), as well as in data from protein atlas analysis.

While the data on cisplatin suggest that indeed ALC1 loss do not impact its sensitivity, I disagree with the statant that "the correlation between dispensability of ALC1 in platinum response suggests that this chromatin remodeler likely does not contribute to MMEJ (page 6)" or " is dispensable for HR (page 7). Indeed, it is has to be stressed that cisplatin induced DNA damage (interstrand crosslinks) are substrates also for nucleotide excision repair, that has a key role in repairing these lesions.

We agree with the reviewer that transcription-coupled NER is the key pathway for the resolution of cisplatin-induced damage. We therefore have revised this statement in the manuscript as “Our data showing the dispensability of ALC1 in cisplatin response, both in BRCA1 and 2-mutant settings, is consistent with previous reports demonstrating the dispensability of this remodeler for MMEJ or transcription-coupled nucleotide excision repair.” We have cited previous work where ALC1 has been shown to be dispensable for MMEJ or TC-NER. Similarly, we have modified the text on page 7 as “Furthermore, ALC1 loss did not impact sensitivity to cisplatin in HRP cyclin E1-high cells. This observation is consistent with previous studies showing its dispensability for HR repair.”

Figure 2. Please explain better why niraparib is not active in cyclinE1-high cells.

Our comprehensive studies examining the impact of ALC1 depletion on PARPi response uncover the generalized theme that targeting is most effective in enhancing sensitivity of olaparib and rucaparib, which have moderate PARP1/2 trapping ability, as compared to niraparib and talazoparib, which are strong trappers. One possible explanation could be that moderate PARP1/2 trappers are more amenable for combination strategies because their effects do not reach full saturation, preserving a dynamic range that allows for additive or synergistic enhancement. This was included in the discussion section of the manuscript.

It is not clear to me if the authors consider a cyclin E "gain" an overexpressing tumor (i.e. OVCAR8). The authors need to show the response to PARPi in one (possibly two) cell lines with very low expression of cyclin E and knock-down of ALC1.

We have present data in multiple BRCA1-WT cell lines with very low expression of cyclin E compared to OVCAR8. These include: FT282 cell line (Fig. 4), two FT282 clones of BRCA1-/+ FT cells (Fig. 5), and full length BRCA1 addback UWB1.289 (Fig. 3c). Additionally, we have added immunoblotting data showing that in OVCAR8, the level of cyclin E1 protein and activity as assessed by pCdk2 is comparable to OVCAR3 and OVCAR4, two CCNE1-amplified lines (Fig. S2d). In contrast, FT 282 and UWB1.289 BRCA1 add back cells have low levels of cyclin E and thus low pCdk2.

The deletion of ALC1 do interfere with tumor take and tumor growth? No clear is the in vivo experiments.

Tumor uptake: We injected OVCAR8 cells in mice three days post-transduction of sgALC1. Depletion of ALC1 is only achieved at 14 days post transduction. This explains why tumor uptake is not impacted. We do not observe a significant impact of ALC1 loss on tumors derived from OVCAR8 cells. This is consistent with the dispensability of ALC1 in the proliferation of HR-proficient cells (PMID: 33333017; PMID: 33462394). We have added text in the manuscript to clarify this point.

Injecting OVCAR8 cells in the peritoneum is not associated with the formation of ascites?

We thank the reviewer to bring up this important point. The objective of this study is to examine how ALC1 loss can enhance PARPi responses and therefore we chose an earlier time point (~50 days) to assess the impact on tumor growth. Ascites formation upon intraperitoneal injection of OVCAR8 cells has primarily been reported at late stages of disease development. For example, Anirban Mitra et al. (2015) (PMID: 26050922) reported consistent ascites formation, but only at extended timepoints (up to ~90 days post-injection). Similarly, Yong-Tae Shen et al. (2019) (PMID: 31117198) injected 5-10 x106 cells and observed ascites emergence beginning around day 49, with progressive accumulation toward the endpoint, indicating that fluid buildup coincides with advanced peritoneal dissemination. In contrast, studies using comparable inoculation doses (e.g., 1×10⁶ cells) and shorter observation periods (~6 weeks) such as Luis Hernandez et al. (2016) (PMID: 27235858) did not report detectable ascites. Taken together, these findings suggest that, while OVCAR8 cells can generate ascites, this phenotype typically manifests at later stages of disease progression and is not expected within shorter experimental windows. Therefore, the absence of ascites in our model is consistent with the study design and timeframe, rather than indicative of a failure of tumor establishment.

We have added relevant discussion in the results section to clarify this point.

How was tumor weight calculated?

Tumor burden was quantified by direct collection and measurement of peritoneal tumor nodules. For the sacrificed mice, all visible tumor nodules within the peritoneal cavity were carefully excised, counted, and pooled per animal. The total tumor weight was then determined by weighing the combined mass of all collected nodules using an analytical balance. Thus, “tumor weight” represents the cumulative mass of macroscopic peritoneal implants per mouse. No estimations or indirect calculations were used. This has now been elaborated on in the methods section.

It seems that tumors grow as solid mass, but how were nodulesAll mice at endpoint exhibited disseminated peritoneal disease, characterized by multiple tumor nodules and invasion into the peritoneal wall. Tumor nodules were quantified by direct visual inspection during necropsy. Small nodules ( Why survival curves were not shown?

Survival analysis was not included because the study was designed with a predefined experimental endpoint to enable controlled comparison of tumor burden across groups. Animals were therefore euthanized at the same timepoint rather than followed longitudinally to survival. As a result, Kaplan–Meier analysis was not applicable to this experimental design. We agree that survival is an important outcome and would be valuable in future studies specifically powered and designed for that purpose.

The dose of 50mgr/kg every third day is a very low olaparib dose. Generally the in vivo dosing is 100mgr/kg , 5 days a week for 4 weeks (doi: 10.1158/1535-7163.MCT-21-0420; 10.1158/2767-9764.CRC-22-0423).

We agree that higher doses of olaparib (e.g., 100 mg/kg, 5 days/week) are commonly used and have demonstrated single-agent efficacy in vivo. In this study, however, our objective was to specifically evaluate the combinatorial effect of olaparib with genetic knock-out of ALC1. To enable this, we intentionally employed a reduced dosing regimen (50 mg/kg every third day) to minimize single-agent activity. This approach allowed us to establish a condition in which olaparib in sgAAVS1 control tumors had limited impact on tumor burden, thereby providing a dynamic range in which to detect potential sensitization effects mediated by sgALC1. Using a fully efficacious dose would likely mask such interactions by producing a near-maximal response in the control group. Thus, the selected dosing strategy reflects a deliberate experimental design to assess potentiation effects rather than to model maximal therapeutic efficacy of olaparib as a monotherapy.

Figure 4. I could not find the data of the minimal impact of ALC1 in UWB1.289 cells. What the author refer to? They refer to the fact that ALC1 deletion di not cause any cell growth alteration or to something else? But were there the data?

The minimal impact being referred to was PARPi responses in BRCA1-proficient UWB1.289. We have now fixed the statement to read: “The minimal impact of ALC1 in BRCA1-proficient UWB1.289 cells on PARPi responses suggested that targeting this remodeler may have minimal impact on normal healthy cells.” and included the relevant figure number (Fig.3c) for clarity.

The modest increment in pRPA in hTER-FT282 is statistically significant and not very different from what observed in UWB.289, suggesting that ACL1 deletion could indeed impact normal cells. These data should be interpreted more conservatively.

The increase in pRPA levels upon ALC1 loss in hTERT FT282 BRCA1 het cells and UWB1.289 cells is 1.2 and 1.4 respectively. This is consistent with the literature that BRCA1-/+ het cells have compromised replication stress response. Unresolved replication stress gets processed into double-strand breaks (DSBs). Consistent with the proficiency of hTERT FT282 BRCA1-/+ het cells in DSBs repair, ALC1 deficiency does not increase yh2ax in these cells. Hence, despite an increase in pRPAS33 signal in hTERT FT282 BRCA1 het cells, these cells can resolve downstream breaks. In contrast, a profound, 1.7-fold increase in yh2ax signal was observed upon ALC1 loss in BRCA-mutant UWB1.289 cells, reinforcing that ALC1 loss has a more profound response in BRCA-mutant cancer cells.

To align with the reviewer’s suggestion, we have removed the word “modest’ and have retained the fold differences in the median values.

Figure 6. Questionable is the OS as endpoint in this heterogeneous patient population (treated in front line and recurrent) and in my opionion OS, much more than PFS, is influences by the many different treatment these patients underwent and that could influence the OS. Why not considering PFS after/or on PARPi treatment? The authors should clarify the patient population, Indeed, 48 patients were treated with PARPI and were platinum sensitive and possibly HRD. What patients are the HPR patients? How many were they? It is not clear the HRP and high replication stress cohort were treated with PARPi? How many of these were Cyclin E amplified or with high levels? Figure 6F should also include, beside UVB+BRCA1, other tumor cells with no Cyclin E overexpression and non BRCA mutation or HRD. The discussion of limitations should be addressed to strengthen the manuscript.

We thank the reviewer and agree that PFS is often preferred for evaluating treatment-specific effects. However, in this cohort, PFS was not a reliable endpoint for several reasons. Tumor samples were obtained at diagnosis, whereas PARPi was administered later, in either the frontline maintenance or recurrent setting, introducing temporal and prognostic heterogeneity that limits the interpretability of PFS. These factors confound attribution of PFS specifically to PARPi response. We therefore selected OS from the time of PARPi exposure as a more consistently defined endpoint across this heterogeneous cohort, while acknowledging its limitations.

Reviewer #2 (Significance (Required)):

The manuscript by Lindsey et al. explores the role of ALCN1 (Amplified in Liver Cancer 1) loss in enhancing the sensitivity of PARPi in ovarian carcinomas, including BRCA1/2 mutated tumors (both sensitive and resistant to platinum) as well as cyclin E amplified settings. The data are interesting but the in some cases there is an overinterpretation of the results.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)): __ The manuscript by Aubuchon, Wong et al. presents strong insights into the value of ALC1 as novel target for sensitization strategies against PARPi. The authors show that a PARPi resistance is reversible when ALC1 is knocked down and convincingly highlight the genetic circumstances for these approaches. Also, the authors point out that especially the weak PARP-trappers olaparib and rucaparib could benefit from concomitant ALC1 inhibiton and high levels of replication stress by elevated p-T21 RPA2 could serve as biomarker in clinical settings. Furthermore, the authors show that benign fallopian tube cells are not affected by ALC1-kd, which is an important finding for in vivo approaches.

We thank the reviewer for acknowledging that our work provides “strong insights” and makes “important finding for in vivo approaches”.

As the manuscript covers a broad experimental field, I would only suggest a few additional experiments to further strengthen the overall story:

1. How does an ALC1 knock-down affect the expression of PARP1 and if so, how does this contribute to the effects seen by ALC1-kd? The authors could add Western Blot experiments for cell lines belonging to the respective groups that are distinguished in the manuscript: BRCA wt, BRCA mutated and Cyclin E1-high cancer cells and also a benign fallopian tube cell line.

This was an interesting point brought up by the reviewer. To address this, we examined and compared total PARP1 protein levels in BRCA1 add-back UWB1.289, BRCA1-mutant UWB1.289, cyclin E1-high OVCAR8, and FT282, between ALC1 WT and depleted cells. However, we do not observe any consistent alteration in PARP1 level upon ALC1 depletion (Fig. Supp. Fig. 6a, b).

In some of the Western Blot data, it also looks like BRCA1 expression is affected by ALC1 kd. The authors could provide some quantified protein expression or qPCR data if there is a correlation between both expressions.

To address the reviewer’s question, we quantified changes in BRCA1 levels upon ALC1 loss across all cell lines used in this study. As expected, BRCA1 levels were higher in UWB del 11q and Cyclin E1-overexpressing cell lines. In contrast, cell lines harboring heterozygous BRCA1 mutations or BRCA1 promoter methylation were among those with the lowest BRCA1 expression. This trend provides us confidence in reliably quantifying our immunoblotting data. Although minor fluctuations in BRCA1 protein levels were observed following ALC1 depletion, no consistent trend towards either an increase or decrease was evident (Fig. Supp. Fig. 6c). Likewise, when cell lines were grouped according to their sensitivity to PARP inhibition upon ALC1 loss, no clear pattern emerged (Fig. Supp. Fig. 6d). Together, these data suggest that ALC1 depletion does not substantially affect BRCA1 protein levels, consistent with our previous RNA-seq and functional studies indicating that this chromatin remodeler is dispensable for transcriptional regulation or homologous recombination (PMID: 33462394).

To further strengthen the hypothesis that the effects of strong PARP-trappers are not improved by ALC1 kd, the authors should add data regarding the viability of the cells presented in Figure 3b upon treatment with niraparib and talazoparib in sgALC1 cells (versus vector control). Also, the authors should add cell viability data using talazoparib for the sgALC1 OVCAR cell lines (versus vector control) in Figure 2 and Supplement Figure 3.

Sensitivity to niraparib and talazoparib upon ALC1 depletion have now been added in Figure 3b, and for OVCAR lines in Supplement Figure 3. As correctly pointed by the reviewer, we consistently observe that impact of ALC1 loss is more profound on olaparib and rucaparib compared to niraparib and talazoparib.

Some minor points I noticed while reading the manuscript:

We apologize for the oversight and thank the review for pointing this out.

• in Figure 3b, both graphs have the same title. I think the right one should be "SYr14" instead of "SYr12" again

Fixed. - In the heading of Figure 2 an "in" is missing

Fixed.

• There are some citations, that seem to be made with another citation style (superscript numbers) than numbers in brackets across the manuscript.

Fixed.

Reviewer #3 (Significance (Required)):

The most important aspect resulting from this manuscript is that ALC1 inhbitors could improve the response to some PARPi without damaging healthy cells. Thereby, the authors also mention the limitation of the use of ALC1 as a target and offer a potential biomarker for combinatory approaches. This study offers a very detailed insight into the potential role of ALC1 as a target for sensitization approaches under the different genetic conditions that can occur in HGSOC. These novel insights contribute to further broaden the therapeutic options by PARPi in clinical settings if the results can be approved by in vivo trials.

• *

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 Aubuchon, Wong et al. presents strong insights into the value of ALC1 as novel target for sensitization strategies against PARPi. The authors show that a PARPi resistance is reversible when ALC1 is knocked down and convincingly highlight the genetic circumstances for these approaches. Also, the authors point out that especially the weak PARP-trappers olaparib and rucaparib could benefit from concomitant ALC1 inhibiton and high levels of replication stress by elevated p-T21 RPA2 could serve as biomarker in clinical settings. Furthermore, the authors show that benign fallopian tube cells are not affected by ALC1-kd, which is an important finding for in vivo approaches.

As the manuscript covers a broad experimental field, I would only suggest a few additional experiments to further strengthen the overall story:

1. How does an ALC1 knock-down affect the expression of PARP1 and if so, how does this contribute to the effects seen by ALC1-kd? The authors could add Western Blot experiments for cell lines belonging to the respective groups that are distinguihed in the manuscript: BRCA wt, BRCA mutated and Cyclin E1-high cancer cells and also a benign fallopian tube cell line.
2. In some of the Western Blot data, it also looks like BRCA1 expression is affected by ALC1 kd. The authors could provide some quantified protein expression or qPCR data if there is a correlation between both expressions.
3. To further strengthen the hypothesis that the effects of strong PARP-trappers are not improved by ALC1 kd, the authors should add data regarding the viability of the cells presented in Figure 3b upon treatment with niraparib and talazoparib in sgALC1 cells (versus vector control). Also, the authors should add cell viability data using talazoparib for the sgALC1 OVCAR cell lines (versus vector control) in Figure 2 and Supplement Figure 3.

Some minor points I noticed while reading the manuscript:

• in Figure 3b, both graphs have the same title. I think the right one should be "SYr14" instead of "SYr12" again
• In the heading of Figure 2 an "in" is missing
• There are some citations, that seem to be made with another citation style (superscript numbers) than numbers in brackets across the manuscript.

Significance

The most important aspect resulting from this manuscript is that ALC1 inhbitors could improve the response to some PARPi without damaging healthy cells. Thereby, the authors also mention the limitation of the use of ALC1 as a target and offer a potential biomarker for combinatory approaches. This study offers a very detailed insight into the potential role of ALC1 as a target for sensitization approaches under the different genetic conditions that can occur in HGSOC.

These novel insights contribute to further broaden the therapeutic options by PARPi in clinical settings if the results can be approved by in vivo trials.

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 manuscript by Lindsey et al. explores the role of ALCN1 (Amplified in Liver Cancer 1) loss in enhancing the sensitivity of PARPi in ovariar carcinomas, including BRCA1/2 mutated tumors (both sensitive and resistant to platinum) as well as cyclin E amplified settings.

The data are interesting but the in some cases there is an overinterpretation of the results. I have listed below my major concerns

Figure 1. Could the authors demonstrate that OVASAHO cells are BRC2 muted? Indeed, I have always though they were BRCA wt type (10.1016/j.ygyno.2015.08.017). While the data on cisplatin suggest that indeed ALC1 loss do not impact its sensitivity, I disagree with the statant that "the correlation between dispensability of ALC1 in platinum response suggests that this chromatin remodeler likely does not contribute to MMEJ (page 6)" or " is dispensable for HR (page 7). Indeed, it is has to be stressed that cisplatin induced DNA damage (interstrand crosslinks) are substrates also for nucleotide excision repair, that has a key role in repairing these lesions. Figure 2. Please explain better why niraparib is not active in cyclinE1-high cells. It is not clear to me if the authors consider a cyclin E "gain" an overexpressing tumor (i.e. OVCAR8). The authors need to show the response to PARPi in one (possibly two) cell lines with very low expression of cyclin E and knock-down of ALC1.<br /> The deletion of ALC1 do interfere with tumor take and tumor growth? No clear is the in vivo experiments. Injecting OVCAR8 cells in the peritoneum is not associated with the formation of ascites? How was tumor weight calculated? It seems that tumors grow as solid mass, but how were nodules<1mm quantified? Please clarify. Why survival curves were not shown? The dose of 50mgr/kg every third day is a very low olaparib dose. Generally the in vivo dosing is 100mgr/kg , 5 days a week for 4 weeks (doi: 10.1158/1535-7163.MCT-21-0420; 10.1158/2767-9764.CRC-22-0423).

Figure 4. I could not find the data of the minimal impact of ALC1 in UWB1.289 cells. What the author refer to? They refer to the fact that ALC1 deletion di not cause any cell growth alteration or to something else? But were there the data? The modest increment in pRPA in hTER-FT282 is statistically significant and not very different from what observed in UWB.289, suggesting that ACL1 deletion could indeed impact normal cells. These data should be interpreted more conservatively.

Figure 6. Questionable is the OS as endpoint in this heterogeneous patient population (treated in front line and recurrent) and in my opionion OS, much more than PFS, is influences by the many different treatment these patients underwent and that could influence the OS. Why not considering PFS after/or on PARPi treatment? The authors should clarify the patient population, Indeed, 48 patients were treated with PARPI and were platinum sensitive and possibly HRD. What patients are the HPR patients? How many were they? It is not clear the HRP and high replication stress cohort were treated with PARPi? How many of these were Cyclin E amplified or with high levels? Figure 6F should also include, beside UVB+BRCA1, other tumor cells with no Cyclin E overexpression and non BRCA mutation or HRD.

The discussion of limitations should be addressed to strengthen the manuscript.

Significance

The manuscript by Lindsey et al. explores the role of ALCN1 (Amplified in Liver Cancer 1) loss in enhancing the sensitivity of PARPi in ovarian carcinomas, including BRCA1/2 mutated tumors (both sensitive and resistant to platinum) as well as cyclin E amplified settings. The data are interesting but the in some cases there is an overinterpretation of the results.

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

ALC1 suppression has been shown to potentiate PARP inhibitor lethality in HR-deficient cells. Rather than revisiting the underlying mechanism, which has been characterized and remains an active area of investigation, this study aims to define the clinical contexts in which combined ALC1 and PARP inhibition may be beneficial. The clinical efficacy of PARP inhibitors, and their FDA approval, is largely restricted to HR-deficient tumors. This study dissects the combined effects of ALC1 and PARP suppression across a panel of HRD ovarian cancer cell lines, multiple classes of PARP inhibitor, and cells harboring distinct PARPi resistance mechanisms. In doing so, the authors delineate both the potential utility and the limitations of combined ALC1 and PARP inhibitor treatment in HRD ovarian cancers. The most impactful finding of the study, however, is likely the demonstration that ALC1 suppression sensitizes HR-proficient, CCNE1-amplified high-grade serous ovarian cancers to PARP inhibitors. These tumors are associated with particularly poor outcomes owing to the current absence of effective targeted therapies, making this observation of considerable clinical relevance. Of note, the study relies on genetic rather than pharmacological depletion of ALC1, a choice likely reflecting the current lack of a commercially available ALC1 inhibitor. While genetic suppression may not fully recapitulate the effects of combined drug treatment, it t offers the advantage of not being tied to any specific compound, allowing the authors to establish more general principles. I have only a few comments.

The effect of ALC1 KO on PARPi sensitivity is less pronounced in OVSAHO cells (BRCA2-mutated) than in BRCA1-mutated cells. In these cells, it looks like there is an additive effect rather than synergy.

1. The authors should calculate, if possible, whether there is synergy or additive effect of ALC1-KO lethality (BLISS).
2. Another BRCA2-mutated cell line should be included.

Minor comments:

• Figure key is missing for S2C (I assume it's grey DMSO, blue olaparib)
• Page 8: "BRCA1-mutant ovarian cancer cells eventually develop chemoresistance when exposed to PARPi for a prolonged period. Mechanistically, this is due to rewiring of ATR signaling, which enables RAD51 loading at DNA breaks and reversed forks independent of BRCA1 protein(25)." This sentence suggest this is the only existing resistance mechanism, which should be correct. Modify to "mechanistically, this CAN be due to", or "this is OFTEN due to".

Significance

ALC1 inhibitors have been developed and clinical trials are starting. The significance of this manuscript lies in establishing the clinical potential for combined ALC1-PARP inhibition in high grade serous ovarian cancer. Especially, the authors demonstrate that combined ALC1 suppression with PARP inhibition efficiently kills HR-proficient CCNE1-amplified ovarian cancers, which represent 20% of ovarian cancers and are resistant to current therapies.

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:

In this paper, Tomasek and colleagues describe a series of experiments illuminating the effects of OM-89, a bacterial lysate taken orally for prevention of recurrent UTI, on intracellular dynamics of UPEC, using cell culture and organoid models. Suggestions for improvement and for clarification of the authors' conclusions and relevance to human UTI (and OM-89 use) are offered below.

Major points:

1. The data indicate that OM-89 exposure in the organoids enhances lysosomal degradation pathways and (in mBOs) autophagic flux, and the authors conclude this is a mechanism by which UPEC regrowth after antibiotic treatment (modeling rUTI) is inhibited by OM-89. They also show enhanced cellular uptake of fluorescently labeled antibiotics (ampicillin) in organoids - this leads them to conclude (and state in the paper's title) that increased intracellular antibiotic concentration effects increased killing of UPEC and decreased regrowth. These are two separate proposed mechanisms, and especially with regard to the antibiotics, they have not shown that increased intracellular antibiotic concentration actually kills intracellular UPEC in their model - only that regrowth as measured microscopically is less. In total, a mechanistic connection between the observed lysosomal effect and the intracellular antibiotic uptake, and which one is more important for UPEC control in this model, is incomplete. The precise wording of the paper's title should be reconsidered accordingly.

We agree with the point raised by the reviewer that we did not show a mechanistic connection between the observed lysosomal effect and the intracellular antibiotic uptake. Further experiments dissecting the exact involved mechanistic pathways driving both - either in conjunction or separately - would improve our understanding on how OM-89 leads to its positive effects. In future studies we will focus on dissecting the underlying pathways and determining whether a mechanistic connection exists to explain the observed positive effects of OM-89 between lysosomal degradation and enhanced intracellular antibiotic accumulation.

Accordingly, we changed the title to "Targeted lysosomal activation in bladder epithelium enhances clearance of intracellular uropathogenic ____Escherichia coli". This revised title avoids implying a direct causal link between increased intracellular antibiotic accumulation and bacterial clearance, while still reflecting the central biological process identified in our study.

Additionally, we incorporated changes in the introduction, as highlighted in our reply to point number one raised by reviewer number two.

OM-89 is taken orally for rUTI prevention, and some "components" reach the urinary tract (line 81). But it isn't explained how applying OM-89 directly to organoids models how its components may reach the bladder epithelium (from the basolateral side, if the OM-89 is applied outside the organoids) in the whole animal or human. At the least, this limitation should be stated in the Discussion.

We thank the reviewer for pointing out this limitation. Although advanced in vitro models help to better mimic the in vivo situation, they still do not fully recapitulate all aspects of drug exposure and delivery observed in vivo. We included the following statement of limitation now in the discussion in line 449-459: "One limitation of our study is that OM-89 was applied directly to epithelial cultures and organoids, whereas in clinical use it is administered orally. Although pharmacokinetic studies have demonstrated systemic distribution and urinary accumulation of OM-89-derived components following oral administration (van Dijk, 1982), our experimental setup does not recapitulate the exact route, kinetics or concentration profiles encountered ____in vivo. Rather, our models were designed to determine whether bladder epithelial cells are capable of responding directly to OM-89-mediated signals and to identify the intracellular pathways involved. Given the documented systemic exposure following oral administration, direct effects on the urothelium are biologically plausible. However, future studies will be required to determine how the epithelial responses identified here integrate with the complex systemic and immune-mediated effects of OM-89 under physiological administration conditions."

In the lysosome studies starting on line 319, the cultured cells are all infected (and either treated with OM-89 or not). What observations regarding number and size of vesicles, etc (all the measures in Fig 6) are evident when cells are treated with OM-89 only? These data should be presented (at least as a supplemental figure) to enable optimal interpretation of the OM-89+UPEC data in Fig 6. As the authors themselves indicate, OM-89 may be having a generalized effect on endocytic and/or autophagic flux by bladder epithelial cells, independent of infection.

We thank the reviewer for this suggestion and agree that evaluating OM-89 treatment in the absence of infection provides important context for interpreting the infection-associated phenotypes shown in Figure 6. Our original intention was to focus the main manuscript on the effects of OM-89 during UPEC infection, and we therefore did not include the corresponding uninfected conditions.

As part of the planned revision, we will include additional supplementary data examining the effects of OM-89 alone in both murine and human bladder epithelial cells. Specifically, we will present analyses of Lamp1-positive lysosomal vesicles, lysosomal acidification (LysoSensor), and Cathepsin L activity under uninfected conditions. These experiments will allow readers to assess the extent to which OM-89 activates epithelial lysosomal pathways independently of infection and will provide important context for interpreting the infection-associated responses presented in the main figures.

We agree with the reviewer that OM-89 may exert broader effects on epithelial lysosomal pathways beyond the setting of infection, and inclusion of these data will strengthen the interpretation of OM-89 as a direct modulator of epithelial antimicrobial function.

With the organoids, beyond the microscopic quantification of UPEC, can CFUs be measured?

We understand the wish of the reviewer to see CFU measurements performed on organoids. However, this imposes strong technical limitations, mainly due to the tedious and technically challenging microinjections, e.g. the exact same amount of organoids would need to be infected by microinjections in both conditions (OM-89 and control) and injections would need to be performed extremely precise with no bacteria spreading into the surrounding extracellular matrix (frequently, organoids would get penetrated with the microneedle all the way, leading to bacteria being not injected into the lumen but rather into the wall of the organoid or even be released on the other side of the organoid) as otherwise also bacteria escaping into the extracellular matrix would be collected upon recovering the organoids from the extracellular matrix domes, strongly affecting the CFU measurements.

However, using differentiated monolayers of mouse bladder epithelial cells and performing a classic gentamicin protection assay would add an additional layer of information on the purely intracellular bacterial population, whilst overcoming the previously mentioned technical challenges. Therefore, we aim to perform CFU measurements on monolayers with and without OM-89 treatment to support our microscopic quantification and specifically be able to make a statement on reduced intracellular bacterial burden with OM-89 treatment. The CFUs will therefore provide an orthogonal measure of intracellular bacterial burden and complement the microscopy-based quantification during the infection and antibiotic-treatment phases.

Adding to this point of the reviewer, we wanted to clarify that with the higher-throughput microscopic quantification used in our approach (Thunder widefield microscope at 25x magnification), we cannot distinguish between strictly intracellular or tissue-associated bacteria, hence we used the wording "intra-organoid" in our methods section. We now added this information also into the results section for clarification (line 116): "Hence, the microscopy data represent the total "intra-organoid" bacterial burden at each experimental stage, without distinguishing the exact localization of the bacteria - which can be luminal, intracellular or tissue-associated.". To further reflect this, we stepped back from referring to antibiotic-mediated "killing", but changed the wording to antibiotic-mediated "clearance" or referred to reduced bacterial burden throughout the manuscript.

__Minor points:____ __

1. In Fig 1A, the "co-application" horizontal line is under the 7-10 hour window, but the text suggests that the application of antibiotics and OM-89 in this experiment is between 4-7 hours.

We thank the reviewer for pointing this out. Indeed, in the co-application regime, OM-89 is added at the same timepoint as the antibiotic - meaning straight after monitoring the growth phase at 4h post-infection (pi). We now adapted the horizontal line for the "co-application" treatment in Figure 1A accordingly to represent the time-point of OM-89 addition better. Additionally, we added a line for the antibiotic-treatment in order to further facilitate readability.

How are antibiotics and OM-89 "removed" at the 7-hour mark? This was not detailed in the Methods.

Although we had specified this in the methods section at line 603 "For every media exchange (e.g. antibiotic treatment or withdrawal), each well was washed with 9 ml of the respective media before leaving 1 ml in the well.", we realized the positioning was not optimal as we had mentioned this part under the point "Bacterial injection" in "Injection experiments". We therefore now separated this part, together with the lid preparation, from the "Bacterial injection" part and created the new subsection "Lid preparation for media changes" (line 613 onwards).

What time point was used for the transcriptomic profiling of organoids? This is not clear from the relevant Methods or Results sections.

As stated in the methods section, RNA for transcriptomic profiling from mBOs was extracted at 4h post-infection (pi) (line 842).

In showing that OM-89 "attenuated" the magnitude of inflammatory responses (Fig 2C and S3B), it would be helpful to add a panel showing the comparison of OM89+UPEC to PBS alone - this would be expected to convey activity (red) in the infection-related pathways, but to a lower magnitude than seen in UPEC vs PBS.

We thank the reviewer for this suggestion, as well as comment number 5 below. We comment more on both suggestions below.

Similarly, in the results outlined starting on line 196, it would be helpful to add a panel showing OM89+UPEC vs OM89 alone.

We performed the requested, combined GOBP analyses and they confirm that infection-associated pathways remain strongly activated in OM89-treated infected organoids relative to baseline (PBS) controls and relative to OM89-treated uninfected organoids. These results confirm the reviewer's hypotheses and further confirm the results presented in Figure 2C. In fact, induction of genes involved in detrimental effects of UPEC infections are induced at a lower extent when organoids are exposed to OM-89 only.

However, because the direct comparison between OM89+UPEC and PBS+UPEC already highlights the effect of OM-89 while controlling for the infection status, we believe our original analysis presented in Figure 2C remains the most informative representation of attenuation. Therefore, we will include the new comparison in the supplementary section of the manuscript.

In line 236, what is meant by lysosomal "activation"? A more specific term should be chosen here.

We thank the reviewer for this question and aim to increase readability of this section. With lysosomal activation in the first sentence of the mentioned paragraph, we referred to the observed effect of upregulated lysosomal pathways and altered lysosomal vesicles in the previous paragraph. However, to make the connection to the previous paragraph better, and given the comment number two of reviewer number two, we changed the whole first paragraph of this section. Therefore, the first sentence of this paragraph (line 235 onwards) reads now: "To test whether the observed effects on lysosomal pathways could mechanistically, at least in parts, explain OM-89-mediated protection, we first used Genebridge analysis (Li et al, 2019) to examine how the lysosomal gene signature identified in our RNA-seq data relates to host defense programs in the human bladder."

In the Abstract (line 25), the phrase "Using bladder organoids..." is a dangling modifier.

We thank the reviewer for pointing this out and changed the sentence accordingly to "In bladder organoids and differentiated epithelial monolayers, OM-89 promotes lysosomal acidification and increases lysosomal protease activity, driving intracellular UPEC toward degradative compartments."

Typographical and copyediting:

We thank the reviewer for pointing out the typographical errors below and we corrected them all.

1. Line 74 should read "For instance..."

2. Line 76 should read "when combined with antibiotic therapy..."

As this sentence is to emphasize the already observed protective effects of OM-89, and the two studies mentioned were either performed without or in combination with antibiotics, we changed the sentence to "For instance, rodent infection studies have demonstrated protective effects of OM-89 alone (Bosch et al, 1988; Lee et al, 2006) and in combination with antibiotic therapy (Canton et al, 2025; Bessler et al, 2010), although this observed in vivo protection could not be linked to any major quantitative changes in bladder immune cell infiltration (Canton et al, 2025), leaving the underlying molecular mechanism not fully resolved." for better readability.

Line 122 should read "...regrowth following antibiotic treatment" or "regrowth post-antibiotic treatment"

Line 138 should use "regimen" not "regime"

Line 196 delete comma after "Although"

Line 244 fully hyphenate "OM-89-mediated"

Line 374 should read "...significantly enhance antibiotic-mediated killing"

• *

__Significance:____ __

The paper is very well written and though a lot of data are included, the presentation is excellent and helps the reader to follow the story. The paper makes a strong contribution to the UTI pathogenesis field, and the use of mouse and human bladder organoids is innovative in studying intracellular UPEC. My scientific expertise as a reviewer is in UPEC pathogenesis, directly relevant to the content of this paper.

Reviewer #2

---

Evidence, reproducibility and clarity:

This study examined the effect of OM-89 on UPEC infection, antibiotic clearance, and resurgence in mouse and human organoid models. The goal of the study was to understand the molecular mechanisms by which OM-89 is effective at preventing rUTI in patients.

Major comments:

The manuscript is well-written and the figures are well presented. Adequate background information is provided to give the study context and sufficient experimental details are provided to allow replication by other groups. Experiments contain appropriate controls and sufficient replicates to allow appropriate statistical analyses. The authors are careful to acknowledge the differences they observed between the mouse and human system and provide satisfactory potential explanations for these differences. The conclusions they draw are well supported by their data and none of their claims from their data are overstatements. Below are some, which I believe if addressed could improve the paper.

1. I think the authors overstate the novelty of the concept that the urothelium is an active targetable determinant of infection and treatment outcomes. This is not an entirely new concept since previous studies have examined antimicrobial peptides and other factors from the urothelium.

We thank the reviewer for this important point and agree that the urothelium has long been recognized as an active participant in host defense through mechanisms such as antimicrobial peptide production, pathogen sensing and regulation of inflammatory responses. We have therefore revised the manuscript to avoid implying that urothelial involvement in infection outcome is itself a novel concept. Instead, we now emphasize the specific advance of our study: the identification of lysosome-centered epithelial activation as a therapeutically targetable mechanism that enhances intracellular bacterial clearance and potentiates antibiotic efficacy.

In the abstract we changed: "Our findings position the bladder epithelium from a passive barrier to an active, targetable determinant of treatment outcome and suggest host-directed modulation of epithelial antimicrobial pathways as a promising strategy to enhance intracellular bacterial clearance." to "Our findings demonstrate that bladder epithelial antimicrobial pathways can be pharmacologically reinforced to influence treatment outcomes by enhancing intracellular bacterial clearance." in line 30.

In the introduction we changed: "Together with increased intracellular accumulation of antibiotics across different classes, this leads to improved intracellular killing and reduced bacterial regrowth across diverse UPEC strains." to "Together with increased intracellular accumulation of antibiotics across different classes, this leads to improved intracellular clearance and reduced bacterial regrowth across diverse UPEC strains." in line 91 and "Together, these findings reveal a previously unrecognized epithelial lysosome-centered mechanism by which OM-89 enhances intracellular antibiotic performance and repositions the bladder epithelium from a passive reservoir of infection reactivation to an actively transformable antimicrobial compartment influencing treatment outcomes." to "Together, these findings reveal a previously unrecognized epithelial-centered mechanism by which OM-89 enhances intracellular antibiotic performance and establishes lysosomal activation as a therapeutically targetable component of epithelial host defense against intracellular UPEC." in line 96.

In the discussion we changed: "Together, these findings provide a mechanistic framework for the long-observed clinical efficacy of OM-89. Our findings reveal that the urothelium itself can be therapeutically targeted to reduce pathogen regrowth by transforming the epithelial barrier from a passive refuge for UPEC into an active defense site." to "Together, these findings provide a mechanistic framework for the long-observed clinical efficacy of OM-89 and identify epithelial lysosomal pathways as a therapeutically targetable component of host defense that can be used to improve intracellular bacterial clearance." in line 398 and "In the face of rising antimicrobial resistance (2024), strengthening epithelial antimicrobial function offers a complementary route to shift the bladder mucosa from a passive niche of bacterial survival and infection reactivation toward an active site of accelerated pathogen clearance." to "In the face of rising antimicrobial resistance (2024), our findings provide a mechanistic rationale for the clinical use of OM-89 and support epithelial lysosomal pathways as a promising target for host-directed therapeutic strategies that enhance intracellular bacterial clearance and improve the efficacy of existing antibiotics." in line 469.

Depending on the target audience, the Module-Module association analysis could need more introduction. I am not a computational biologist and it was not obviously apparent how Figure 4A is generated and what it actually showing. How specifically does this analysis demonstrate a functional link between lysosomal activity and immune defense pathways? Without further explanation, it is my opinion that this figure panel is an unnecessary distraction that is not required for any of the conclusions that the group can already draw from the rest of their data.

We thank the reviewer for this constructive critique. We agree that the rationale and interpretation of this analysis were not sufficiently explained in the original manuscript. We have therefore expanded the description of the MMAS approach and clarified how these data support the translational relevance of the lysosomal pathways identified in our experimental models. We also agree that for a broader biological audience, the computational framework and the strategic necessity of Figure 4A required a clearer introduction and stronger justification.

To address the reviewer's concerns, we have thoroughly revised the text (lines 235-250) to clarify the methodology and emphasize the essential translational value this analysis adds to our study:

• How the figure is generated and what it shows: We have added explicit language clarifying that we used a computational Module-Module Association Score (MMAS) to evaluate the transcriptional correlation between the lysosomal gene network and functional biological pathways. Rather than relying on a single experimental dataset, this analysis compiles data across eight independent human bladder transcriptomic datasets encompassing over 1,400 clinical samples.
• Demonstrating the link to immune pathways: We have explicitly named the specific host defense modules highlighted in Figure 4A, namely "Response to molecule of bacterial origin", "cell activation involved in immune response", and "innate immune response" to guide the reader directly to the strong positive correlations shown in the panel.
• Justifying its inclusion (mouse-to-human translational bridge): While the rest of our data characterizes the cellular mechanics of OM-89 in murine organoids and cell culture, Figure 4A demonstrates that the link between lysosomal activity and bacterial defense is a conserved feature of bladder tissue biology across species. This cross-species alignment (our mouse-data at this stage of the manuscript compared to human-derived data) provides critical clinical justification for targeting epithelial lysosomal pathways as a therapeutic strategy in human patients. The new paragraph reads as follows: "To test whether the observed effects on lysosomal pathways could mechanistically, at least in parts, explain OM-89-mediated protection, we first used Genebridge analysis (Li et al., 2019) to examine how the lysosomal gene signature identified in our RNA-seq data relates to host defense programs in the human bladder. To evaluate the translational relevance of our experimental findings, we used a computational Module-Module Association Score (MMAS) analysis across eight independent human bladder transcriptomic datasets comprising over 1,400 clinical samples. This network-based approach evaluates the transcriptional correlation between the lysosomal gene network and functional biological pathways across diverse human cohorts. Module-Module association analysis performed on these human bladder datasets indicated that the lysosome module has strong positive associations with specific host defense modules, including "response to molecule of bacterial origin", "cell activation involved in immune response", and "innate immune response" (Figure 4A), highlighting a conserved functional link between lysosomal activity and immune defense pathways in the bladder epithelium. Altogether, these positive correlations suggest that lysosomal activation represents a conserved pathway integrated within mucosal immunity across species, rather than an isolated cellular response unique to our experimental models."

__Significance:____ __

General assessment: Solid experimental design with appropriate controls. Appropriate statistical rigor. Conclusions justified by the data. Limitations acknowledged. Differences in results between mice and humans acknowledged.

Advance: Moderate technical advance building on prior organoid models. Significant mechanistic advance because OM-89 has been widely used for a long time without detailed understanding of why it works. Moderate conceptual advance that urothelial cells are a targetable determinant of treatment outcomes.

Audience: I am a basic science researcher in the field of female urogenital tract microbiome and infections. Other researchers studying UTI will certainly be interested in this study. It also may be of interest to people studying other bladder conditions that involve the urothelium (bladder cancer).

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 study examined the effect of OM-89 on UPEC infection, antibiotic clearance, and resurgence in mouse and human organoid models. The goal of the study was to understand the molecular mechanisms by which OM-89 is effective at preventing rUTI in patients.

Major comments:

The manuscript is well-written and the figures are well presented. Adequate background information is provided to give the study context and sufficient experimental details are provided to allow replication by other groups. Experiments contain appropriate controls and sufficient replicates to allow appropriate statistical analyses. The authors are careful to acknowledge the differences they observed between the mouse and human system and provide satisfactory potential explanations for these differences. The conclusions they draw are well supported by their data and none of their claims from their data are overstatements. Below are some, which I believe if addressed could improve the paper.

1. I think the authors overstate the novelty of the concept that the urothelium is an active targetable determinant of infection and treatment outcomes. This is not an entirely new concept since previous studies have examined antimicrobial peptides and other factors from the urothelium.
2. Depending on the target audience, the Module-Module association analysis could need more introduction. I am not a computational biologist and it was not obviously apparent how Figure 4A is generated and what it actually showing. How specifically does this analysis demonstrate a functional link between lysosomal activity adn immune defense pathways? Without further explanation, it is my opinion that this figure panel is an unnecessary distraction that is not required for any of the conclusions that the group can already draw from the rest of their data.

Significance

General assessment: The manuscript has several methodological strengths. These include the use of both mouse and human urothelial models, inclusion of appropriate controls, and sufficient replicates to ensure reproducibility. The statistical methods employed were appropriate. No major methodological weaknesses were identified. The descriptions of methods provide sufficient experimental details to allow the experiments to be reproduced by other labs. The authors did a nice job interpreting their data in light of previous literature. They did not overstate the magnitude or significance of their findings and were careful to acknowledge the limitations in their study design.

Advance: Moderate technical advance building on prior organoid models. Significant mechanistic advance because OM-89 has been widely used for a long time without detailed understanding of why it works. Moderate conceptual advance that urothelial cells are a targetable determinant of treatment outcomes.

Audience: I am a basic science researcher in the field of female urogenital tract microbiome and infections. Other researchers studying UTI will certainly be interested in this study. It also may be of interest to people studying other bladder conditions that involve the urothelium (bladder cancer).

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 paper, Tomasek and colleagues describe a series of experiments illuminating the effects of OM-89, a bacterial lysate taken orally for prevention of recurrent UTI, on intracellular dynamics of UPEC, using cell culture and organoid models. Suggestions for improvement and for clarification of the authors' conclusions and relevance to human UTI (and OM-89 use) are offered below.

Major points:

1. The data indicate that OM-89 exposure in the organoids enhances lysosomal degradation pathways and (in mBOs) autophagic flux, and the authors conclude this is a mechanism by which UPEC regrowth after antibiotic treatment (modeling rUTI) is inhibited by OM-89. They also show enhanced cellular uptake of fluorescently labeled antibiotics (ampicillin) in organoids - this leads them to conclude (and state in the paper's title) that increased intracellular antibiotic concentration effects increased killing of UPEC and decreased regrowth. These are two separate proposed mechanisms, and especially with regard to the antibiotics, they have not shown that increased intracellular antibiotic concentration actually kills intracellular UPEC in their model - only that regrowth as measured microscopically is less. In total, a mechanistic connection between the observed lysosomal effect and the intracellular antibiotic uptake, and which one is more important for UPEC control in this model, is incomplete. The precise wording of the paper's title should be reconsidered accordingly.
2. OM-89 is taken orally for rUTI prevention, and some "components" reach the urinary tract (line 81). But it isn't explained how applying OM-89 directly to organoids models how its components may reach the bladder epithelium (from the basolateral side, if the OM-89 is applied outside the organoids) in the whole animal or human. At the least, this limitation should be stated in the Discussion.
3. In the lysosome studies starting on line 319, the cultured cells are all infected (and either treated with OM-89 or not). What observations regarding number and size of vesicles, etc (all the measures in Fig 6) are evident when cells are treated with OM-89 only? These data should be presented (at least as a supplemental figure) to enable optimal interpretation of the OM-89+UPEC data in Fig 6. As the authors themselves indicate, OM-89 may be having a generalized effect on endocytic and/or autophagic flux by bladder epithelial cells, independent of infection.
4. With the organoids, beyond the microscopic quantification of UPEC, can CFUs be measured?

Minor points:

1. In Fig 1A, the "co-application" horizontal line is under the 7-10 hour window, but the text suggests that the application of antibiotics and OM-89 in this experiment is between 4-7 hours.
2. How are antibiotics and OM-89 "removed" at the 7-hour mark? This was not detailed in the Methods.
3. What time point was used for the transcriptomic profiling of organoids? This is not clear from the relevant Methods or Results sections.
4. In showing that OM-89 "attenuated" the magnitude of inflammatory responses (Fig 2C and S3B), it would be helpful to add a panel showing the comparison of OM89+UPEC to PBS alone - this would be expected to convey activity (red) in the infection-related pathways, but to a lower magnitude than seen in UPEC vs PBS.
5. Similarly, in the results outlined starting on line 196, it would be helpful to add a panel showing OM89+UPEC vs OM89 alone.
6. In line 236, what is meant by lysosomal "activation"? A more specific term should be chosen here.
7. In the Abstract (line 25), the phrase "Using bladder organoids..." is a dangling modifier.

Typographical and copyediting:

1. Line 74 should read "For instance..."
2. Line 76 should read "when combined with antibiotic therapy..."
3. Line 122 should read "...regrowth following antibiotic treatment" or "regrowth post-antibiotic treatment"
4. Line 138 should use "regimen" not "regime"
5. Line 196 delete comma after "Although"
6. Line 244 fully hyphenate "OM-89-mediated"
7. Line 374 should read "...significantly enhance antibiotic-mediated killing"

Significance

The paper is very well written and though a lot of data are included, the presentation is excellent and helps the reader to follow the story. The paper makes a strong contribution to the UTI pathogenesis field, and the use of mouse and human bladder organoids is innovative in studying intracellular UPEC. My scientific expertise as a reviewer is in UPEC pathogenesis, directly relevant to the content of this paper.

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

Summary: Edvalson and colleagues use transcriptomics, cell biology and genetics to study variation between segregation distorter (meiotic drive) strains and find several important results. These include apparent suppression of small RNAs mapping to responder (the drive target) in one of the lines, a general pattern of differential expression consistent with the drive mechanism being upstream of sperm individualization (where defects have been seen previously), and genetic confirmation that perturbing Rsp expression can influence the strength of drive.

Major comments: I found the total RNA sequencing experiment a bit oddly presented. This is partly because it was in the middle of the results (might fit better first), partly because few specific genes were discussed (this might be appropriate given then question, but maybe the question should be more clearly stated), and the complexity of the approach (WCGNA + PANGEA) and how it all fits together. I suggest working to clarify the main points of this section (which are a bit different than the main focus of the Rsp work).

We thank the reviewer for these important points. We liked the suggestion to swap the order of our results. We attempted the change, but we found that we weren't able to make the flow of the results much better. Instead, we primed the transition from smRNA to totRNA in the last paragraph of the smRNA results (lines 190-196). This paragraph now reads:

The dearth of Rsp smRNAs in SD-Mad heterozygotes could be due to a disruption in transcription of the locus or subsequent processing steps. Many factors can influence piRNA production. For example, the piRNA pathway can amplify piRNAs independently of transcription, such as the ping pong cycle, (Czech and Hannon 2016). Notably, Rsp piRNAs do not have a strong ping pong signature in testes (Wei et al. 2021; Chen et al. 2021a). To distinguish between a disruption in transcription or some downstream process, we examined total RNA.

The main reason we elected to describe patterns rather than specific genes is that the 2nd chromosomes we tested (R-16, SD-Mad, SD-5) have all diverged from each other and any single differentially expressed gene could be due to differences in genetic background. Therefore, we elected to point out more broad systematic changes in pathways and correlated gene networks rather than specific genes. We have made it more obvious throughout the total RNA section in the text what our question is regarding the transcriptome and the reasoning for using WGCNA and gene set analysis.

We also appreciate the reviewers point that the complex approach we used to extract changes in pathways and networks is difficult to follow. We have modified our wording to better describe the flow of analyses.

We also note that we have extended our analysis for the comparison of SD-Mad and SD-MadRev, which only differ by the Sd-RanGAP locus. Here we do discuss individual genes that are differentially expressed. See below for details about this new analysis.

Minor comments:

Abstract - Probably worth mentioning Sd-RanGAP here, even if you are using it as a straw man. I agree that the specific mechanism is not known, but some of the genetics are established.

This is a good point. While our study doesn't address RanGAP, it is important to point out that, although its role in drive is unclear, Sd-RanGAP is a necessary component of the system. We added the following language to the abstract:

SD is a multigene complex, frequently associated with chromosomal inversions, where the main driver locus, a truncated duplication of the gene RanGAP kills wild-type sperm containing a satellite DNA called Responder (Rsp).

Line 80 and elsewhere - it would be helpful to be specific here - you are looking at both small and total RNA

We've modified our wording throughout the manuscript to specify when we are referring to total RNA and small RNA.

Fig 1B - is there a reason not to show the values of the replicates here? It would be more transparent.

We thank the reviewer for this comment. We replaced Fig 1B with a chart that is computed from the DESeq2 normalized counts for each comparison and added replicates to all related graphs.

Line 139 - does the experimental design control for 1.688 genomic copy number? Where is it located?

We indeed control for the 1.688 copy number here. Most 1.688 repeats are found on the X chromosome and all flies in our experiments have identical X chromosomes. We changed the text to specify that copy number for 1.688 are the same between conditions.

144-146 - this could be written clearer, and I think it should only refer to 1C, not 1B. Part of the issue is that there are several repeats not discussed, and it isn't clear what is happening with them. I suggest expanding this description so it is more clear.

Thank you for this feedback. We have expanded the description to make this section clearer.

Line 161 - what do you mean (specifically) by "repetitive loci"?

Repetitive loci in this case refers to transposons, satellite DNAs (except simple satellites), and piRNA clusters. We have added text explaining what is included the grouping of "repetitive loci". We have added the following sentence to the text:

Our results demonstrate that SD-Mad and SD-5 haplotypes, despite sharing the same main drive locus, have different effects on smRNAs derived from repetitive loci such as complex satellites (including Rsp), transposable elements, and piRNA clusters.

193-203 - This is an important finding that is somewhat lost in trying to keep track of WCGNA and PANGEA and the different Modules. I suggest clarifying to drive home the point that differential expression appears to start prior to individualization, which suggests and earlier mechanism of drive.

We thank the reviewer for this feedback. We have added wording to out discussion that points out this finding in lines 501-505 which reads:

We suspect that the timing of the proximal cause of SD-mediated drive may align with early spermatogenetic processes; perhaps where cell cycle-related genes are active and appear to be broadly differentially expressed (Figure 2B, Module H). This earlier timing is consistent with temperature shift experiments that place the critical period for SD at or before meiosis (Mange 1968).

Fig 3B & 3C, Fig 4 - same as 1B, is there a reason not to show the actual data points?

A similar issue was brought up earlier, in response we modified all our figures to show replicate points where applicable.

Line ~245 - was the same experiment done with SD-5? (as you do below for Rsp overexpression)

We originally did not include SD5 in this experiment, but we have since measured drive strength of SD5 in a kipfKO background. We found a small but statistically significant difference in drive strength. We added the new SD5 results to the figure and moved the kipfKD data to the supplement along with some added data on a Rsp deletion line generated from Iso1 that bolsters our confidence in the SDMad results.

Significance

This is a strong paper that moves the field forward, even if it leaves questions still to be answered (why the difference between drivers? what is the mechanism? how is rsp interacting with drive?

Several findings move the field forward: the Rsp small RNA results, the differential expression hinting at a molecular mechanism that is upstream of sperm individualization.

The audience is moderately broad. Genetic conflict is gaining in general interest, but aspects of this will be mostly interesting to the hardcore drive crowd.

Reviewer #2

Evidence, reproducibility and clarity

I have only one request: I found it unclear whether the authors were referring to small RNAs or their precursor (long RNA). By reading the text carefully, I could deduce that Fig1A/Table S2 represent the small RNA sequencing, while FigS3A represents total RNA seq (detecting precursor). However, the labeling in the Fig1A and Table S2 only says 'piRNA cluster' or 'Rsp' (without clarifying 'piRNA from piRNA cluster' or 'piRNA from Rsp'), and it took quite some time for me to understand which Fig/data is smallRNA vs. longRNA.

This is helpful feedback. We have added more clarity to which type of RNA is being represented in our figures throughout.

Significance

This manuscript by Edvalson et al. describes their study on SD (segregation distorter) meiotic drive system, examining the role of piRNA derived from Rsp satellite. Although the exact mechanism of drive is still unknown, this study represents a significant step forward in understanding SD-mediated drive.

By using two SD alleles (SD-5 and SD-Mad), they show that Rsp-derived piRNA is depleted in SD-Mad. The authors used total RNA sequencing/small RNA sequencing mutants and carefully designed controls (such as deletion of Sd-RanGAP) to reach the model that Rsp-derived piRNA is involved in SD-Mad-mediated drive. The result that kipferl depletion (that lead to sat DNA expression) rescues SD-Mad's drive phenotype is very interesting. This supports that the decreased Rsp piRNA indeed corresponds to SD-Mad-mediated drive. They further back up this idea by overexpressing Rsp.

Interestingly, SD-5 was not impacted by changes in Rsp expression. Based on this result, the authors state that there are mechanistic variations in the same (SD) drive system. This statement is certainly justified by the data, but I cannot help wondering there might be a unifying mechanism that explains both SD-5 and SD-Mad. I am not suggesting to edit the manuscript or add the discussion: but do they have any speculations on this? For example, SD-5 is simply epistatic to Rsp piRNA production? For example, SD-RanGAP > SD-Mad (some gene on SD-Mad inversion) > Rsp piRNA production > SD-5 > sperm killing?

We thank the reviewer for this insight. We indeed think that the proximal cause of sperm dysfunction could be the same, but there are components of SD5 that act downstream of Rsp piRNAs. The small difference in drive strength in the SD5 KipfKO experiments might support this hypothesis, although it is also possible instead that drive is influenced by changes in some other piRNAs (from the piRNA clusters or satellites).

We modified our wording in the first paragraph of the discussion to point out this possibility. Lines 367-370 now reads

These results suggest that, while SD chromosomes share a target and main drive locus (Sd-RanGAP), the modifiers accumulated on each haplotype may influence the drive mechanisms, either by creating new pathways to drive or acting as tuning knobs on drive strength.

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

Summary

In the presented manuscript Edvalson and Wei et al use Drosophila genetics and NGS experiments to investigate the mechanism of meiotic drive through the Segregation Distorter (SD) system. They reveal that two driving haplotypes seem to function via different mechanisms, with drive through SD-Mad but not SD-5 involving small RNAs produced from the Responder (Rsp) satellite, the target of SD drive. SD-Mad testes displaying drive are characterized by lower levels of Rsp sRNAs compared to non-drive controls as well as SD-5, and the ectopic overexpression of Rsp sRNAs through two distinct mechanisms decrease drive in SD-Mad genetic background, specifically. With this work, the authors are adding an important piece of information to the highly complex SD system, indicating that sperm killing is likely achieved by different mechanisms in different SD haplotypes, despite sharing a common driver.

Major comments

Fig1C: It might be interesting to show the fold change between SD-Mad and SD-MadRev in addition to what is displayed. Moreover, can the authors comment on what might be causing the increased smRNA counts for 38C2? Is this because R16 has particularly low 38C2 values?

We appreciate the reviewer's comment concerning the fold change between SD-Mad and SD-MadRev. We have made a figure showing the difference between and put it in Figure S1.

We suspect that the expression difference in 38C2 between the R16 heterozygotes and SD heterozygotes may be due to genetic divergence, since these are different 2nd chromosomes. We have added language pointing this out to the manuscript in line 182. The paper now reads:

*There is no evidence that either 38C2 or Flamenco are involved in SD-mediated drive. *

Fig1/S1: Could the authors also display the Rsp smRNA counts for all Gla crosses similar to panel 1B? What is the interpretation for the increase in Rsp smRNAs in SD-5/Gla relative to R16/Gla but the lack of such an increase in the SD-5/iso1 vs R16/iso1 comparison? Do SD-Mad and SD-5 induce the same strength of drive against each of the two wildtype chromosomes? Experiments: smRNAseq for SD-MadRev/Gla.

We have added a plot to Fig S1 to show the abundance of Rsp small RNAs in the Gla background, similar to Figure 1.

It is difficult to interpret the apparent overabundance of Rsp small RNAs in the SD-5/Gla background. Because differences in Rsp smRNA abundance for SD-5 are inconsistent between the Iso1 and Gla background, our interpretation is that SD-5 is not manipulating Rsp levels. The apparent overabundance of Rsp in the Gla background could be due to an epistatic interaction between Rsp and other components of that particular background. Consistent with this interpretation, the SD-Mad induced reduction of Rsp smRNAs in the Gla background is less dramatic than in the Iso1 suggesting that something about that background is increasing Rsp expression slightly when paired with an SD chromosome.

Fig1: The authors note changes in smRNA levels for other satellites as well as piRNA clusters but do not give any interpretation to this observation. Are they meaningful? Should they be attributed to genetic background?

Our interpretation of the observation that some satellites or piRNA clusters are differentially expressed is that these differences are likely due to epistatic effects from the different 2nd chromosomes used in the study or are incidental to mechanism of SD.

FigS2: Same question also for the deregulated TEs: do they share sequence features with Rsp or are they overrepresented in the clusters that change? Are these explained by differences in insertions between genotypes? Do their total RNAseq values change in any way? What do the percentages in line 162 correspond to? Number of TEs that are deregulated? At which cutoff? It might be informative to compare the data to a cross between driver and R16, or even better the SD-MadRev control. Experiments: totRNAseq for SD-MadRev crosses and optionally crosses to R16.

The Rsp repeat unit does not share significant homology to portions of the genome outside of the pericentromere of 2R with the exception of ~6-12 copies in the intron of Ago3.

As far as TEs are concerned, we surprisingly don't see a strong correlation between piRNA cluster content, dysregulation, and TE transcript abundance. For example, in the SD/Gla backgrounds the total RNA for R1, R2, IGS, and Tc1-Mariner family TEs is down regulated. However, the only major piRNA cluster that is upregulated in both SD/Gla backgrounds (80F) is not enriched for TE fragments matching any of those 4 families. One thing we can note is that the definition of the major piRNA clusters are given in relation to the Iso1 genome which may differ from that of our experimental backgrounds. Without long read resolved genomes for our specific experimental lines generated at the same time as the RNA samples it is difficult to determine how expression at the major piRNA clusters and the corresponding TEs are related. We have described this lack of a correlation in lines 210-217 in the text along with our interpretation for why this could be. The paper now reads:

On the other hand, we did find some differences in repetitive elements related to rDNA (R1, R2, and IGS) and Tc1-Mariner family TEs (all backgrounds; Figure S6). Interestingly, there was no correlation between the expression of TEs and the expression of piRNA clusters that contain fragments of these TEs in the total RNA, nor was there any correlation between the small RNAs from piRNA clusters and the total RNAs for those TEs. PiRNA clusters are usually defined in one isolate of Iso1: rapid turnover of TEs and piRNA sources could explain why we do not see a correlation between piRNA cluster expression and TE expression in our backgrounds.

We investigated differences in TE and piRNA cluster expression in our SD-Mad/Iso1 vs SD-MadRev/Iso1 comparison, but a lack of power due to inter-sample variation prevents us from confidently making any assessments on any TEs or piRNA clusters in that comparison. We did however generate additional gene level transcriptomic data using 3' Digital Gene Expression to bolster our confidence in the totRNA data and found some interesting genes that were in the top most differentially expressed. We have noted those genes in lines 276-287 which read:

To identify genes that might interact to cause drive, we compared the gene expression of SD-Mad/Iso1 to SD-MadRev/Iso1. These genotypes only differ by the presence of the main drive locus, Sd-RanGAP. We performed both totRNA and 3' Digital Gene Expression (DGE) RNA sequencing and examined the overlap in differential expression between the totRNA and DGE sequencing. There are 69 differentially expressed genes where the DGE comparison is significant (PDGE {less than or equal to} 0.01), and the sign of the Log2FC of the totRNA matches that of the DGE. Among this set of differentially expressed genes, 57 show at least a 50% difference in gene expression (absolute Log2FC value of at least 0.58 in DGE). These genes are not enriched in any Reactome gene sets. The top 20 most differentially expressed genes consists of 9 lncRNAs (3 anti- sense RNAs) and 11 protein coding genes: 8 of which are uncharacterized. The 3 characterized genes are Artemis (Arts), Gr61a, and Tono (Figure S98, Supplemental File 1).

We discuss two of these genes in further detail in the discussion in lines 476-486 which read:

First, Tono, a BTB zinc finger-containing transcription factor is upregulated (Log2FCDGE = 1.7) in all SD-Mad comparisons. Tono plays a role in regulating transcription in muscle cells in response to mechanical pressure (Zhang et al. 2024) but also shows enrichment in male germ cells (Li et al. 2022). The putative DNA-binding capacity and ability to form nuclear condensates (Zhang et al. 2024) makes this an interesting candidate gene for interacting with the Rsp satellite. Second, the importin-4 ortholog, Artemis (Arts), which facilitates Ran-mediated import of H3 and H4 is overexpressed in SD-Mad (Log2FCDGE = 2.5). Interestingly, Arts expression is antagonistic to male fertility (VanKuren and Long 2018). Also of note, Apollo, a duplicate of Arts which supports male fertility (VanKuren and Long 2018) is downregulated (Log2FCDGE = -0.6) though it is not in the top-most differentially expressed genes.

Figure S3: Am I reading the PCA plots right in that there are very few gene expression changes when the drivers are in iso1 background but much more in the Gla background? Comment on possible explanations for that. Please indicate the number of significantly changed genes in each comparison. Again, are these changes correlated between the two drivers or can they be attributed to genetic background of Gla vs R16? Would it be interesting to see how SD-Mad/Gla and SD-5/Gla gene expression profiles compare? Experiment: totRNAseq for SD-MadRev crosses.

There did tend to be more differences in the Gla background compared to Iso1. This difference can best be explained by inter-sample variation in the SD-Mad/Iso1 background which we see in the PCA plot in Fig S4A. Another reason for the difference could be that the Gla and Iso1 chromosomes are very different from each other which prevents us from making any 1-to-1 comparisons between the SD/Iso1 and SD/Gla backgrounds. We generally avoid comparing between genetic backgrounds for this reason unless they share differences as these are more likely related to drive.

In Figure S5A it seems that totalRNA levels of Rsp are strongly increased in SD-Mad/Gla but not in SD-Mad/iso1. The iso comparison (less piRNAs but same transcript) could indicate that it is actually transcription of the Rsp that is affected here. This is even pointed out in line 205 without discussion of the fact that the Gla comparison (less piRNAs but more transcript) would rather indicate that transcription is intact, but processing into piRNAs is defective. Could this be clarified using FISH as in Figure S8? If true, SD-Mad/Gla should have much more FISH signal than SD-Mad/iso1. Either way, this discrepancy should be further discussed. Experiments: comprehensive smFISH panel for all crosses (including SD-MadRev).

The reviewer makes an excellent point. Why would Rsp long RNAs be overexpressed in the SD-Mad/Gla background? Earlier we noted that in the Gla background specifically the genotypes that contain an SD chromosome seem to have a higher level of Rsp small RNAs than we might expect given our Iso1 results. We conclude that this is likely due to an epistatic interaction between the 2nd chromosomes used in the study and the rest of the chromosomes. This interpretation could extend to the long noncoding precursors as well.

Further, although the difference between SD-Mad/Gla is significant and SD-5/Gla is not, they do move in the same direction. This is also true in the Iso1 backgrounds but in the opposite direction. Given an interpretation that Rsp expression is higher than expected in the SD/Gla background due to epistatic effects, it becomes clearer that changes in long RNA abundance are related to changes in small RNA abundance though not perfectly indicative. However, due to lower count levels for Rsp in the totRNA, we do not have the power to confidently draw that conclusion.

In general, the totRNA profiles of repeats don't seem to correlate well between the genotypes (iso vs Gla crosses, neither for SD-5 nor for SD-Mad). Is this because values are in general small and/or replicates don't correlate? Should these data even be considered? Also panels 2A and S5C are very different from each other. The additional comparison with the SD-MadRev allele crossed into both Iso1 and Gla should give additional insight. Experiment: totRNAseq for SD-MadRev crosses.

The reviewer brings up a good point. While some repetitive elements had relatively small counts in the totRNA (like Rsp) most had adequately high counts. But these differences are to some degree expected. Although the other chromosomes are controlled for, the second chromosomes are different by design including the two SD haplotypes. In this context, similarities between the two haplotypes may be helpful in determining some unifying aspects of the SD mechanism but differences could be incidental to the genotype and not necessarily related to SD.

It may be generally informative to set the sRNA and RNA comparisons into perspective, for example by including the comparison of SD-Mad crosses versus SD-MadRev crosses to exclude unrelated genetic background components as much as possible.

The reviewer is correct here. Differences in the transcriptomes of SD-Mad and our revertant are much more likely due to the drive phenotype. Due to variation between SD-Mad total RNAseq replicates, we have substantially less power when comparing SD-Mad/Iso1 to SD-MadRev/Iso1. We therefore generated new data to address this point: we did digital gene expression for three biological replicates of SD-Mad/Iso-1 and SD-MadRev/Iso1. We described the results of this new analysis above.

FigS6: I assume this is given, but as it is not specified: is the directionality of differential expression taken into account here? Or could it be significantly up in one and down in the other? Please specify / adjust color scale to allow this distinction.

This is a good point. We have modified the figure to not only indicate significance but also direction and magnitude.

FigS8: Please add a scale bar for all images. 1.688 is labeled as 359 in the legend, please unify or/and explain nomenclature. Consider adding a nuclear outline based on DAPI. It looks like 1.688 is actually more different between control and SD-Mad/Iso than Rsp. Could the authors comment on this? In the text the authors mention that these experiments were done for both SD-Mad and SD-5 heterozygotes, but only the SD-Mad data are shown.

The most abundant component of 1.688 repeats is the 359bp repeat, which is used as a proxy for 1.688 and our 359-bp probe cross hybridizes with other abundant variants of 1.688 on chromosome 3. We agree, there does seem to be some differences in the 1.688 RNA FISH, however we do not yet have evidence that 1.688 is related to the drive phenotype. We have expanded that figure (now supplemental figure 7) with multiple images for each genotype to demonstrate the lack of change in Rsp and 1.688 localization. We have added an explanation of the nomenclature.

The reference to SD-5 in the text was made in error. We do not have RNA FISH images of SD-5/Iso1 heterozygotes. We've modified the text to reflect this.

FigS9B: What does the y-axis label mean? Fold change relative to what? Is this not displaying counts?

This is a good catch by the reviewer. The y-axis is mislabeled and should read "TPM". We have made this change.

To set the KipfKD/KO data in context, please give also the k value for SD-MadRev and compare the smRNA values in this context to the data displayed in F1B. Experiment: drive analysis for SD-MadRev.

Our basis for concluding that Rsp smRNA overexpression may reduce drive strength is in demonstrating that kipfKO is sufficient to rescue wild type sperm in driving backgrounds. We did not introduce KipKD (or KO) to the SD-MadRev background because this chromosome does not drive.

The note that the 3XP3-dsRed cassette needs to be flipped out for Rsp overexpression to influence drive is interesting. It would be great if the authors could show a more detailed scheme of the structure of this insertion including the directionality of the promoter relative to the Rsp fragment and the rest of cluster 38C (including dm6 coordinates perhaps). Small RNA sequencing compared to totRNA sequencing should reveal if the transcription or the processing into piRNAs of the inserted piece is affected, and if more of the 38C piRNAs are affected. Genic transcription has been previously observed to limit Rhino-dependent piRNA production from piRNA clusters (Andersen et al 2017). It might be of interest to the general piRNA community to see how cluster output is influenced through the integration of an internal genic promoter.

We agree that this is an interesting result. We have added more detail to Fig 4A to indicate directionality and genomic location of the insert in terms of dm6.

Figure panel 4A should be adjusted to include annotations of the black boxes and to give genomic locations. It is unclear what the blue brackets mean, and where exactly the insertion took place. Are the attP sites relevant for the experiments? It might be nice to see a piRNA profile over the locus, to put the levels of additional Rsp piRNAs into perspective.

We have removed the black boxes from the schematic as they were only there as an aesthetic choice. We have indicated where exactly the insertion was made. The attP sites are there for future experimental flexibility.

Minor comments

Figure 3B: fold change of satellite RNA is shown. It might be obvious that the fold change relates to KipfKO / WT but this should be stated explicitly. What is the genetic background here?

Thank you for the comment. We added information on the genetic background in the figure.

Figure legends should be extended for clarity throughout the manuscript in main and supplementary figures. All color codes and abbreviations as well as samples / genotypes and assay used should be clearly explained. Few examples include: F1B: smRNA or totalRNA? F3B: fold change relative to what? F4B: what are these data relative to? F4C: smRNA or totalRNA? S2: Is this smRNAseq? Further description of the color code in the volcano panels would be desirable. FS3: typo in A-B should be A-D. Fold changes relative to what. Etc.

Thank you for these helpful suggestions. We have edited the figure legends as suggested to improve the clarity. We appreciate the feedback.

The abbreviation for Kipferl is kipf, not kip.

Thank you for pointing this out, we have made the corrections.

I don't understand the sentence on lines 310-312.

We agree that sentence was confusing. We replaced it with:

"Identifying potential proteins that interact with Rsp may therefore provide important clues about why satellites like Rsp are targets of drive."

**Referee cross-commenting**

I agree with the other reviewer's assessments

Reviewer #3 (Significance (Required)):

General assessment

This study of a highly complex and poorly understood drive system adds a very interesting piece to the puzzle of understanding the interplay between a RanGAP duplication and a large satellite array. It's strengths lay in the use of genetics tricks to modify drive (SD-MadRev allele, KipfKO, Rsp cluster insertion). The main weakness of the study is the relatively low correlation of several observations between drive crosses to the Iso1 and Gla lines and lack of explanations thereof. Neither gene nor repeat expression seem to give a convincing overlap in any direction.

Furthermore, it is interesting that SD-Mad and SD-5 have such different dependencies on Rsp sRNA. While outside the scope of this work, it would be very interesting to see how other drive haplotypes behave: is SD-5 the exception or is it SD-Mad (as the authors have also wondered in the discussion). Such additional comparisons may clarify also the discrepancies in RNAseq.

Advance

While it has been previously shown by the same group that Rsp satellites give rise to smRNAs through the piRNA pathway, it is to my knowledge unclear how and if these smRNAs influence drive. This study thus presents a conceptual advance in that it demonstrates that the role of Rsp smRNAs is not shared among driving haplotypes.

Audience

This study is relevant for a highly specialized audience interested in meiotic drive. It contributes to the understanding of the SD system and may serve as a basis for future research in this area. In addition, results reported in Figure 4 may be of peripheral interest for the Drosophila piRNA community for technical interests.

This reviewers expertise: Drosophila, piRNA pathway, heterochromatin, sRNA

This reviewers limitations: nuclear-cytoplasmic trafficking, 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

In the presented manuscript Edvalson and Wei et al use Drosophila genetics and NGS experiments to investigate the mechanism of meiotic drive through the Segregation Distorter (SD) system. They reveal that two driving haplotypes seem to function via different mechanisms, with drive through SD-Mad but not SD-5 involving small RNAs produced from the Responder (Rsp) satellite, the target of SD drive. SD-Mad testes displaying drive are characterized by lower levels of Rsp sRNAs compared to non-drive controls as well as SD-5, and the ectopic overexpression of Rsp sRNAs through two distinct mechanisms decrease drive in SD-Mad genetic background, specifically. With this work, the authors are adding an important piece of information to the highly complex SD system, indicating that sperm killing is likely achieved by different mechanisms in different SD haplotypes, despite sharing a common driver.

Major comments

Fig1C: It might be interesting to show the fold change between SD-Mad and SD-MadRev in addition to what is displayed. Moreover, can the authors comment on what might be causing the increased smRNA counts for 38C2? Is this because R16 has particularly low 38C2 values?

Fig1/S1: Could the authors also display the Rsp smRNA counts for all Gla crosses similar to panel 1B? What is the interpretation for the increase in Rsp smRNAs in SD-5/Gla relative to R16/Gla but the lack of such an increase in the SD-5/iso1 vs R16/iso1 comparison? Do SD-Mad and SD-5 induce the same strength of drive against each of the two wildtype chromosomes? Experiments: smRNAseq for SD-MadRev/Gla.

Fig1: The authors note changes in smRNA levels for other satellites as well as piRNA clusters but do not give any interpretation to this observation. Are they meaningful? Should they be attributed to genetic background?

FigS2: Same question also for the deregulated TEs: do they share sequence features with Rsp or are they overrepresented in the clusters that change? Are these explained by differences in insertions between genotypes? Do their total RNAseq values change in any way? What do the percentages in line 162 correspond to? Number of TEs that are deregulated? At which cutoff? It might be informative to compare the data to a cross between driver and R16, or even better the SD-MadRev control. Experiments: totRNAseq for SD-MadRev crosses and optionally crosses to R16.

Figure S3: Am I reading the PCA plots right in that there are very few gene expression changes when the drivers are in iso1 background but much more in the Gla background? Comment on possible explanations for that. Please indicate the number of significantly changed genes in each comparison. Again, are these changes correlated between the two drivers or can they be attributed to genetic background of Gla vs R16? Would it be interesting to see how SD-Mad/Gla and SD-5/Gla gene expression profiles compare? Experiment: totRNAseq for SD-MadRev crosses.

In Figure S5A it seems that totalRNA levels of Rsp are strongly increased in SD-Mad/Gla but not in SD-Mad/iso1. The iso comparison (less piRNAs but same transcript) could indicate that it is actually transcription of the Rsp that is affected here. This is even pointed out in line 205 without discussion of the fact that the Gla comparison (less piRNAs but more transcript) would rather indicate that transcription is intact, but processing into piRNAs is defective. Could this be clarified using FISH as in Figure S8? If true, SD-Mad/Gla should have much more FISH signal than SD-Mad/iso1. Either way, this discrepancy should be further discussed. Experiments: comprehensive smFISH panel for all crosses (including SD-MadRev).

In general, the totRNA profiles of repeats don't seem to correlate well between the genotypes (iso vs Gla crosses, neither for SD-5 nor for SD-Mad). Is this because values are in general small and/or replicates don't correlate? Should these data even be considered? Also panels 2A and S5C are very different from each other. The additional comparison with the SD-MadRev allele crossed into both Iso1 and Gla should give additional insight. Experiment: totRNAseq for SD-MadRev crosses.

It may be generally informative to set the sRNA and RNA comparisons into perspective, for example by including the comparison of SD-Mad crosses versus SD-MadRev crosses to exclude unrelated genetic background components as much as possible.

FigS6: I assume this is given, but as it is not specified: is the directionality of differential expression taken into account here? Or could it be significantly up in one and down in the other? Please specify / adjust color scale to allow this distinction.

FigS8: Please add a scale bar for all images. 1.688 is labeled as 359 in the legend, please unify or/and explain nomenclature. Consider adding a nuclear outline based on DAPI. It looks like 1.688 is actually more different between control and SD-Mad/Iso than Rsp. Could the authors comment on this? In the text the authors mention that these experiments were done for both SD-Mad and SD-5 heterozygotes, but only the SD-Mad data are shown.

FigS9B: What does the y-axis label mean? Fold change relative to what? Is this not displaying counts?

To set the KipfKD/KO data in context, please give also the k value for SD-MadRev and compare the smRNA values in this context to the data displayed in F1B. Experiment: drive analysis for SD-MadRev.

The note that the 3XP3-dsRed cassette needs to be flipped out for Rsp overexpression to influence drive is interesting. It would be great if the authors could show a more detailed scheme of the structure of this insertion including the directionality of the promoter relative to the Rsp fragment and the rest of cluster 38C (including dm6 coordinates perhaps). Small RNA sequencing compared to totRNA sequencing should reveal if the transcription or the processing into piRNAs of the inserted piece is affected, and if more of the 38C piRNAs are affected. Genic transcription has been previously observed to limit Rhino-dependent piRNA production from piRNA clusters (Andersen et al 2017). It might be of interest to the general piRNA community to see how cluster output is influenced through the integration of an internal genic promoter.

Figure panel 4A should be adjusted to include annotations of the black boxes and to give genomic locations. It is unclear what the blue brackets mean, and where exactly the insertion took place. Are the attP sites relevant for the experiments? It might be nice to see a piRNA profile over the locus, to put the levels of additional Rsp piRNAs into perspective.

Minor comments

Figure 3B: fold change of satellite RNA is shown. It might be obvious that the fold change relates to KipfKO / WT but this should be stated explicitly. What is the genetic background here?

Figure legends should be extended for clarity throughout the manuscript in main and supplementary figures. All color codes and abbreviations as well as samples / genotypes and assay used should be clearly explained. Few examples include: F1B: smRNA or totalRNA? F3B: fold change relative to what? F4B: what are these data relative to? F4C: smRNA or totalRNA? S2: Is this smRNAseq? Further description of the color code in the volcano panels would be desirable. FS3: typo in A-B should be A-D. Fold changes relative to what. Etc.

The abbreviation for Kipferl is kipf, not kip.

I don't understand the sentence on lines 310-312.

Referee cross-commenting

I agree with the other reviewer's assessments

Significance

General assessment

This study of a highly complex and poorly understood drive system adds a very interesting piece to the puzzle of understanding the interplay between a RanGAP duplication and a large satellite array. It's strengths lay in the use of genetics tricks to modify drive (SD-MadRev allele, KipfKO, Rsp cluster insertion). The main weakness of the study is the relatively low correlation of several observations between drive crosses to the Iso1 and Gla lines and lack of explanations thereof. Neither gene nor repeat expression seem to give a convincing overlap in any direction.

Furthermore, it is interesting that SD-Mad and SD-5 have such different dependencies on Rsp sRNA. While outside the scope of this work, it would be very interesting to see how other drive haplotypes behave: is SD-5 the exception or is it SD-Mad (as the authors have also wondered in the discussion). Such additional comparisons may clarify also the discrepancies in RNAseq.

Advance

While it has been previously shown by the same group that Rsp satellites give rise to smRNAs through the piRNA pathway, it is to my knowledge unclear how and if these smRNAs influence drive. This study thus presents a conceptual advance in that it demonstrates that the role of Rsp smRNAs is not shared among driving haplotypes.

Audience

This study is relevant for a highly specialized audience interested in meiotic drive. It contributes to the understanding of the SD system and may serve as a basis for future research in this area. In addition, results reported in Figure 4 may be of peripheral interest for the Drosophila piRNA community for technical interests.

This reviewers expertise: Drosophila, piRNA pathway, heterochromatin, sRNA

This reviewers limitations: nuclear-cytoplasmic trafficking, 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

I have only one request: I found it unclear whether the authors were referring to small RNAs or their precursor (long RNA). By reading the text carefully, I could deduce that Fig1A/Table S2 represent the small RNA sequencing, while FigS3A represents total RNA seq (detecting precursor). However, the labeling in the Fig1A and Table S2 only says 'piRNA cluster' or 'Rsp' (without clarifying 'piRNA from piRNA cluster' or 'piRNA from Rsp'), and it took quite some time for me to understand which Fig/data is smallRNA vs. longRNA.

Referee cross-commenting

I agree with other reviewers' comments, which all seem to be reasonable.

Significance

This manuscript by Edvalson et al. describes their study on SD (segregation distorter) meiotic drive system, examining the role of piRNA derived from Rsp satellite. Although the exact mechanism of drive is still unknown, this study represents a significant step forward in understanding SD-mediated drive.

By using two SD alleles (SD-5 and SD-Mad), they show that Rsp-derived piRNA is depleted in SD-Mad. The authors used total RNA sequencing/small RNA sequencing mutants and carefully designed controls (such as deletion of Sd-RanGAP) to reach the model that Rsp-derived piRNA is involved in SD-Mad-mediated drive. The result that kipferl depletion (that lead to sat DNA expression) rescues SD-Mad's drive phenotype is very interesting. This supports that the decreased Rsp piRNA indeed corresponds to SD-Mad-mediated drive. They further back up this idea by overexpressing Rsp.

Interestingly, SD-5 was not impacted by changes in Rsp expression. Based on this result, the authors state that there are mechanistic variations in the same (SD) drive system. This statement is certainly justified by the data, but I cannot help wondering there might be a unifying mechanism that explains both SD-5 and SD-Mad. I am not suggesting to edit the manuscript or add the discussion: but do they have any speculations on this? For example, SD-5 is simply epistatic to Rsp piRNA production? For example, SD-RanGAP > SD-Mad (some gene on SD-Mad inversion) > Rsp piRNA production > SD-5 > sperm killing?

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: Edvalson and colleagues use transcriptomics, cell biology and genetics to study variation between segregation distorter (meiotic drive) strains and find several important results. These include apparent suppression of small RNAs mapping to responder (the drive target) in one of the lines, a general pattern of differential expression consistent with the drive mechanism being upstream of sperm individualization (where defects have been seen previously), and genetic confirmation that perturbing Rsp expression can influence the strength of drive.

Major comments: I found the total RNA sequencing experiment a bit oddly presented. This is partly because it was in the middle of the results (might fit better first), partly because few specific genes were discussed (this might be appropriate given then question, but maybe the question should be more clearly stated), and the complexity of the approach (WCGNA + PANGEA) and how it all fits together. I suggest working to clarify the main points of this section (which are a bit different than the main focus of the Rsp work).

Minor comments:

Abstract - Probably worth mentioning Sd-RanGAP here, even if you are using it as a straw man. I agree that the specific mechanism is not known, but some of the genetics are established.

Line 80 and elsewhere - it would be helpful to be specific here - you are looking at both small and total RNA

Fig 1B - is there a reason not to show the values of the replicates here? It would be more transparent.

Line 139 - does the experimental design control for 1.688 genomic copy number? Where is it located?

144-146 - this could be written clearer, and I think it should only refer to 1C, not 1B. Part of the issue is that there are several repeats not discussed, and it isn't clear what is happening with them. I suggest expanding this description so it is more clear.

Line 161 - what do you mean (specifically) by "repetitive loci"?

193-203 - This is an important finding that is somewhat lost in trying to keep track of WCGNA and PANGEA and the different Modules. I suggest clarifying to drive home the point that differential expression appears to start prior to individualization, which suggests and earlier mechanism of drive.

Fig 3B & 3C, Fig 4 - same as 1B, is there a reason not to show the actual data points?

Line ~245 - was the same experiment done with SD-5? (as you do below for Rsp overexpression)

Referee cross-commenting

I agree with the comments as well.

Significance

This is a strong paper that moves the field forward, even if it leaves questions still to be answered (why the difference between drivers? what is the mechanism? how is rsp interacting with drive?

Several findings move the field forward: the Rsp small RNA results, the differential expression hinting at a molecular mechanism that is upstream of sperm individualization.

The audience is moderately broad. Genetic conflict is gaining in general interest, but aspects of this will be mostly interesting to the hardcore drive crowd.

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Minor comments 1) The authors suggest that the weak 4th protomer in the HCMV UL52 3-mer map is a consequence of flexibility. This may be the case, but it may also be the case that the class is polluted with 4-mer particles leading to reduced occupancy. Erasing the weak density and running a multi-model 3D classification providing the erased 3-mer and a 4-mer starting map may separate these.

We performed additional analysis (i.e., 3-mer and 4-mer particles were combined into a multi-class ab initio reconstruction followed by multi-class heterogenous refinement) and found that the original 3-mer map was a mixture of 3-mer and 4-mer states.

We have updated Fig. 2a, Supplementary Fig. 2, Supplementary Fig. 3, Supplementary Table 1, Supplementary Movie 1, and removed the discussion of the weak protomer in the 3-mer map from the results section. We have updated our EMDB and PDB depositions accordingly.

• 2) I found the supplemental figure to show the DNA in the tripentamer map too small, this is an interesting finding and should be shown more clearly.*

We have increased the size of Supplementary Fig. 6 and moved the figure caption to another page to accommodate this enlargement.

Reviewer #2

*Major issues 1) There is a high probability that the tripentamer is an artifact of the cross-linking. Because of this, it'd be great to know more about the cross-linking reaction, ideally mass spec identification and quantification of cross-links. This would also address the authors' speculation of contacts that stabilize the tripentamer. *

Crosslinking is a commonly used technique to stabilize complexes that are observed through other means but do not survive the cryo-EM vitrification process. In an EMSA experiment (Supplementary Fig. 4a), UL32 binds 30 bp DNA and migrates slower than when bound to a 10 bp probe, consistent with formation of a supra-pentameric complex. The samples in the EMSA gels are not crosslinked. Additionally, an SDS-PAGE gel of the crosslinked product used for cryo-EM showed tight bands at molecular weights expected for oligomers, supporting specific crosslinking (Supplementary Fig. 4b). These results suggest that crosslinking stabilizes a species that can form but is relatively unstable in solution.

Moreover, the author's claim "However, mutation of K532A/C535A reduced infectious virion production by half (Fig. 4b), suggesting that the tripentamer interface may play a role in the viral life cycle." Seems to be an overreach. Perhaps this is semantics but the data just show that these residues play a role in viral replication (albeit not a huge role based on the modest effect).

We have modified the title of the results section (Line 216-217) to state that "Residues at the tripentamer interfaces contribute to infectious virion production in HSV-1" as well as Line 234 and 241 to indicate that the residues play a role in the viral life cycle.

2) The density for the potential DNA does not look very convincing, although it still remains the strongest hypothesis. The authors should try to strengthen their argument. Does this putative DNA contact residues that they show are necessary for viral replication? Showing seq conservation on the structure could help their argument for the shared function of DNA-binding.

The DNA likely contacts conserved residues at the base and midsection of the central channel (residues R302, R301, R293, K289, R580, R579, R572; see Fig. 6a). We have shown that these residues are important for the production of infectious virions (Fig. 6c): even a single point mutation (R572A) decreased production of infectious virus particles by more than 90%, and double and triple point mutants (R579A/R580A, K289A/R293A/R301A) eliminated production of infectious virus. Sequence conservation of these charged residues in the central channel regions is shown in Supplementary Fig. 1d, f.

3) My last major issue is stylistic and concerns the descriptions of cryoEM structures. I found that the paper was a bit of challenge to read when the authors would introduce each structure. It was a bit of a slog to get through. Descriptions of the structures veered off into overly detailed comparisons that required constant comparison with the figure and didn't really advance my understanding past "the outer surfaces of the three orthologs are different." This masked the more interesting aspects of the authors' findings. Perhaps this could be summarized in supplementary figures or a table. Because this is a stylistic suggestion, the authors should feel free to ignore this request.

We appreciate the reviewer's concerns about accessibility, but we are excited that these structures allowed us to thoroughly describe the convergent and divergent structural features across the Herpesviridae and hope that our in-depth analysis will allow for detailed mechanistic follow-up.

*Minor comments 1) The descriptions of structure determination in the text were often unclear. For example, "In the 3-mer map, a poorly-resolved fourth protomer is visible at low contour levels, suggesting that an additional protomer is present but highly flexible in this class (Supplementary Fig. 3a)." Alternatively, it could be that the classification algorithm wasn't able to fully separate particles that were 3-mers from the 4mers. *

The reviewer is correct. As described above (Reviewer #1 comment 1), we performed additional analysis and found that the original 3-mer map was a mixture of 3-mer and 4-mer states. We have updated Fig. 2a, Supplementary Fig. 2, Supplementary Fig. 3, Supplementary Table 1, Supplementary Movie 1, the EMDB and PDB depositions, and removed the discussion of the weak protomer in the 3-mer map from the results section.

*When describing the structure determination of the HSV1 accessory factor, the authors describe no other particles other than the tripentamer. Were there other particles observed? It'd be a bit surprising that all of the protein adopted the tripentamer state. *

We agree that this result is striking. We picked particles using a 'blob picker' to avoid introducing template bias and found that the tripentamer is the predominant species. Below we show the results of 2D classification of blob picked particles (classes sorted by particle number; obvious junk classes excluded for clarity). There is one class that suggests a pentamer, but template picking with a pentamer template (based on ORF68) did not yield a pentamer class.

Additionally, as we describe in the results section and show in Supplementary Fig. 6a, further processing of the consensus UL32 map showed that 60% of particles formed a complete tripentamer (i.e., 15-mer) while other the remaining 40% formed incomplete tripentamers, missing one or more protomers (e.g., 17% of particles formed a 14-mer).

Was symmetry applied, particularly for the tripentamer that appears to have C-3 symmetry? This is in materials and methods but not clear why it isn't mentioned when describing the structure determination and results.

No symmetry was applied in the reconstruction for either UL32 or UL52. While we previously noted this in the methods section and in Supplementary Table 1, we have added this information to the results section (Line 169-170), the Fig. 3 legend, and cryo-EM processing figures (Supplementary Figures 2, 5, 6) for clarity.

2) Throughout the paper, the authors use the word "remodel" to describe structural differences between orthologs. However, this word usually carries the implication of conformational rearrangement within a protein, and not across orthologs. Please consider a different description.

We agree with the reviewer and have removed the term "remodel" throughout the manuscript text (i.e., Lines 116, 118, 120, 122, 302, 306) and from Supplementary Figures 1, 3, and 5.

3) Figure 2F is confusing and difficult to interpret. It seems that the main point is that these interfaces are conserved, which might be more easily displayed as a standard sequence conservation score mapped onto the structure. I'm also not sure that this figure is necessary as a main figure and could be supplemental.

We agree that the conservation could also be shown this way and have added labels to universally conserved residues of the protomer interface to Supplementary Fig. 1b, c. We have also moved Fig. 2f to the supplement (now Supplementary Fig. 2g).

• 4) The authors write "UL32 bound to the shortest probe tested (10 bp, Supplementary Fig. 4a)." This implies that ONLY the shortest probe is bound and that others are not bound. Consider rephrasing.*

We have rephrased to clarify at all probes tested, included the shortest, bound DNA (Line 153).

• 5) Frustum is misspellt. ;)*

Thank you. Spelling has been corrected (Line 185).

6) In the discussion, the authors speculate that the variability of the outer surface is due to "virus- or host-specific interactions". I'm confused by "host-specific interactions", because the host is the same for all three viruses. Perhaps the authors mean that the different accessory factors could interact with different host factors? If so, are the authors making a Red Queen argument? If so, it'd be pretty cool to do dN/dS analysis to test that hypothesis.

The reviewer is correct in that all three viruses (HSV-1, HCMV, KSHV) infect the same host; however, they replicate in different cell types, which could potentially express different host factors. We have no evidence to support this hypothesis and intended to propose that UL32 and UL52 may be interacting/co-evolving with other viral factors required for genome packaging. We have clarified Line 308 to generalize that "these regions are involved in virus-specific interactions".

To me, this window into evolution of this factor is the biggest advance of the work, and tbh I felt that the authors could lean into this a bit more in the discussion section. Are there any differences in the packaging mechanisms of the different herpes families that can be related to their different behavior? Any other molecular evolution analyses (e.g. dN/dS ratio analysis) that could inform their study?

We agree that understanding the evolution of the packaging accessory factor is an interesting future area of research. There are differences in capsid structure and occupancy of capsid-associated factors across the herpesvirus family (PMID: 34696343). However, we lack a mechanistic (or structural) understanding of viral genome packaging components across the herpesviruses, raising the possibility that there are differences in packaging mechanisms.

Interestingly, the further diverged alloherpesviruses and malacoherpesviruses (other families in the order Herpesvirales) do not appear to encode a factor with similar predicted structure to the Herpesviridae packaging accessory factor (PMID: 41902279). It is unclear how the mechanism of packaging differs in the Orthoherpesviridae and whether replication in mammalian/avian/reptilian cells places additional evolutionary pressure on the viral genome packaging mechanism.

Reviewer #3

Major comments

*1) [I]t is not clear whether the structures presented in the manuscript reflect those produced during HCMV or HSV-1 infection. *

We agree with the reviewer that it is important to consider to what extent purified biomolecules resemble their in vivo counterparts. This criticism can be applied to any ex situ structural analysis. However, our experimental structures allowed us to make testable observations, including the correct assignment of structurally important zinc fingers and the identification of functionally important residues in the central channel.

2) HCMV UL52 was presented to form two distinct structures, a 3-mer and a 4-mer (Fig. 2a). However, the authors acknowledge that the 3-mer is actually a 4-mer when the threshold for the cryo-EM map is lowered. The density is also visible in the PDB validation report for the 3-mer; EMD-74418.

Reviewers #1 and #2 were also curious about the 3-mer. As described above, we performed additional analysis that showed that the original 3-mer map was a mixture of 3-mer and 4-mer states. We have updated Fig. 2a, Supplementary Fig. 2, Supplementary Fig. 3, Supplementary Table 1, Supplementary Movie 1, EMDB and PDB depositions, and removed the discussion of the weak protomer in the 3-mer map from the results section.

*Given that ORF68, BFLF1, and UL32 (Didychuk et al., 2021) form complete pentamer rings, with BFLF1 forming stacked rings, it would seem odd for a protein with conserved function to deviate from a pentamer configuration, suggesting that the structures reported do not reflect the natively produced and functional protein. *

We agree that this is a surprising finding; we initially anticipated that UL32 and UL52 would also form stable pentameric rings. While this study does not resolve a complete mechanism for this factor, it does provide the first structural evidence for the implications of their poor sequence conservation and lack of complementarity.

Furthermore, this is not the first example of a conserved herpesvirus factor that possesses different oligomeric states across different subfamily homologs. As mentioned in the discussion, herpesvirus encode a sliding clamp processivity factor (HSV-1 UL42/HCMV UL44/KSHV ORF59) that shares a common PCNA-like fold, but which has varied oligomeric state across these herpesviruses.

*3) Unlike ORF68 (Didychuk et al., 2021) and UL32 (Suppl. Fig. 4), dsDNA binding experiments were not performed with UL52. Could the partial pentamers simply be poorly formed due to expression in insect cells (mammalian cells were used for protein purification in Didychuk et al., 2021), absence of dsDNA, or inappropriate buffer conditions? Moreover, were the EM grid and vitrification parameters optimized? Grid geometries and chemistries can have profound effects of protein stability especially in the context of the air-water interface, leading to degradation of protein complexes (Glaeser, 2018; D'Imprima et al., 2019). Does UL52 form complexes with dsDNA? Data are shown for the HSV-1 packaging accessory factor. Perhaps dsDNA would stabilize the UL52 pentamer. *

We have purified ORF68 and homologs from both human and insect cell expression systems, and do not observe changes in oligomeric behavior. We find that ORF68 purified as a stable pentamer from human cells (Didychuk eLife 2021) and from insect cells (this work). We have also recombinantly expressed and purified UL32 from human cells. UL32 was largely monomeric after strep affinity purification (chromatogram below, unpublished), as we report from insect cells (this work, Fig. 1c). We switched to insect cell expression systems because of the easier scalability.

Our SEC-MALS data (Fig. 1d) shows that purified UL52 does not oligomerize into a pentamer in solution, so the observed sub-pentameric (3-mer/4-mer) assemblies are unlikely to be an artifact of cryo-EM freezing conditions or the air-water interface. We have not tested if UL52 forms complexes with dsDNA, although it likely does; it is possible that this interaction would stabilize a pentamer.

4) In Didychuk et al., 2021, HSV UL32 is shown to form pentameric rings; negative stained 2D class averages were generated from tagged protein (twin strep tag), produced in mammalian cells (HEK293T), and not purified using size exclusion chromatography. In the present study HSV UL32 was not observed to form pentameric complexes "We first attempted to visualize the pentameric species by negative stain electron microscopy but were unable to identify particles of the expected dimensions." However, it is not clear why this was the case. If the pentameric structures were readily produced in previous experiments, why was cross-linking needed in the current study? As such, the tripentamer complexes seem artifactual in nature.

While a sufficient number of particles were observed in a pentameric state to do 2D class averages in the eLife paper, this was not the dominant state. The results we report in this work are consistent with those reported in the eLife paper. Reviewer #2 (comment #1) was also concerned about the possibility of a crosslinking artifact: we reproduce our response below:

"Crosslinking is a commonly used technique to stabilize complexes that are observed through other means but do not survive the cryo-EM vitrification process. In an EMSA experiment (Supplementary Fig. 4a), UL32 binds 30 bp DNA and migrates slower than when bound to a 10 bp probe, consistent with formation of a supra-pentameric complex. The samples in the EMSA gels are not crosslinked. Additionally, an SDS-PAGE gel of the crosslinked product used for EM showed tight bands, supporting specific crosslinking (Supplementary Fig. 4b). These results suggest that crosslinking stabilizes a species that can form but is relatively unstable in solution."

We have updated Line 148 to clarify this. We have also included a negative stain micrograph, below, in which UL32 pentamers (purified from insect cells) are visible in the absence of crosslinking.

5) Although the data presented in Fig. 4b suggest that interface residues, K532 and C535, might play a role in the formation of the tripentamer and have a minor role in HSV-1 replication, these experiments are incomplete. Single mutations are needed for each residue to assess their individual contribution to tripentamer formation, evidence for a loss of tripentamer formation is needed, and evidence for protein expression is needed.

We agree that we have not unambiguously defined the role of the tripentamer, the precise contributions of residues K532 and C535, or defined the contribution of the tripentamer to HSV-1 viral replication. We seek to report this novel structure to lay the basis for future mechanistic work. Reviewer #2 (comment 1) also questioned the role of these residues in HSV-1 replication, and we addressed this by modifying the title of the results section (Line 216) to state that "Residues at the tripentamer interfaces contribute to infectious virion production in HSV-1" as well as Line 246 and 253 to indicate that the residues play a role in the viral life cycle.

Please refer to Supplementary Fig. 7e for a western blot showing that these mutants do not impact UL32 expression. We included explicit references to UL32 expression on Lines 239 and 288.

*6) In the previous negative stain electron micrographs reported by Didychuk et al., 2021, were the higher order tripentamer complexes seen? *

We did not observed tripentamers in the Didychuk et al. 2021 negative dataset. Tripentamer formation may be concentration dependent. Negative stain EM carried out at nanomolar concentrations would likely cause dissociation of tripentamers, but cryo-EM and EMSA in our work were carried out at micromolar concentrations and were able to capture the higher order tripentamer.

• 7) Formation of disulphide bonds between cysteine residues in vitro is not indicative of complexes forming in vivo during replication. What evidence is there for disulphide bond formation between packaging accessory factor pentamers for KSHV, EBV, and LCMV? In the present study, the disulphide bond could form due to proximity as a result of the cross-linking and the presence of molecular oxygen rather than a bona fide enzyme catalysed reaction during herpesvirus replication to generate packaging accessory factor tripentamers. *

We agree that it is unlikely that disulfide bonds form during infection and have removed this speculation from the manuscript (Line 343-346).

8) The DNA densities in Suppl. Fig. 6e to 6g are curious. As noted by the authors, the 30mer dsDNAs do not traverse through the central cavity of the pentamer. They appear to make contact with neighboring pentamers, again suggesting that these complexes are artefacts from cross-linking. This should be discussed more thoroughly.

Please refer to above discussion of crosslinking and Supplementary Fig. 4.

9) Previously proposed functional roles for ORF68 include a scaffold for terminase assembly, association of the terminase with the portal, generation of initial free ends, or coordination with other replication machinery (Didychuk et al., 2021). Presuming that the new structures for HCMV UL52 and HSV-1 UL32 occur naturally, how do they fit with the previously proposed functional roles of the herpesvirus packaging accessory factor? A more in-depth discussion of this would be valuable.

The common core fold and pentamer/pentamer-like assemble are common features, as is the conserved, positively-charged central channel. We have added additional discussion of this.

*Minor comments A lack of page numbers and line numbers made reviewing this manuscript more challenging than necessary. *

We have included page numbers and line numbers in the revised manuscript.

*As noted in the 'General comments' section above, ORF68 (3.37Å) and BFLF1 (3.60Å) both form pentamers (Didychuk et al., 2021) and were produced in mammalian systems HEK293T cells. Protein purification in the present study was performed in insect (SF9 or High Five) cells. Does this affect complex stability. Also, the tag was retained for UL32 in Didychuk et al., 2021; could this provide stability of the pentamer in the original studies? *

As discussed above, we have no evidence to suggest that expression in human vs. insect cell expression systems dramatically changes oligomerization behavior (Reviewer #3, comment 3). N-terminal purification tags were also retained in this study for structural work but were removed for SEC-MALS, which shows that UL32 is likely in concentration dependent equilibrium between (unstable) pentamers and monomers.

Suppl. Fig. 3 is missing.

We apologize for this oversight and have included Supplementary Fig. 3.

*"UL52 has two regions remodeled" The use of the word 'remodeled' is not appropriate in this context as it implies a single protein can form two shapes under different conditions rather than distinct structures between two disparate proteins; UL52 compared to ORF68. This should be rephrased. *

This was also noted by reviewer 2, and we have removed the term "remodel" throughout the manuscript text (i.e., Lines 134, 138, 140, 337, 341) and from Supplementary Figures 1, 3, and 5.

*What is the density in the central core of UL52 (Fig. 2a; Suppl. Fig. 2e)? Was any form of focused classification performed to establish the identity of the density within the central pseudocavity? *

As noted in the manuscript, this density could be which could be attributed to co-purified protein or nucleic acid, or part of the unresolved, negatively charged loop (residues 82-181) interacting with the positively charged central channel. We have done additional analysis of the central channel density (3D classification with a focus mask) and do not resolve any distinct densities, suggesting that the density is very dynamic.

*Does UL52 bind to dsDNA? To support the hypothesis that the herpesvirus packaging accessory factor has conserved functions across the three subfamilies dsDNA binding experiments should be performed. *

We have not done this experiment. We think that demonstrating this finding for two of the three herpesvirus subfamilies is sufficient.

There is no discussion about how these data relate to the previous functional model for ORF68 presented in Didychuk et al., 2021. Do the new data alter the previous functional models?

The precise mechanistic contribution of the packaging accessory factor remains unknown, and our data do not delineate between the proposed potential roles described in Didychuk et al. 2021. Importantly, our structural information, demonstration of pentameric ring formation, and significance of the positively charged central channel show that the core function of this factor is likely conserved across the virus family. This was not known before our work.

*There are some interesting grammatical phrases; please address throughout the manuscript. One example - "...a notable shared aspiration..." Proteins do not have aspirations. Please use a more formal scientific statement. *

We have updated the language on Line 327.

*Fig. 4b - Statistical analyses missing. Please provide. *

Fig. 6c - Statistical analyses are missing. Please provide. Protein folding/expression data missing; see Fig. 5C showing mutations that result in poor protein expression.

Suppl. Fig. 7f - Statistical analyses absent.

Statistical analysis of the viral complementation in Figs. 4b and 6c has been included. Note that the viral yields reported in Supplementary Fig. 7f were used to calculate complementation efficiency in Figs. 4b and 6c. Protein expression of mutants shown in Fig. 6c was previously included in Supplementary Fig. 7e and is referenced on Lines 288 and 293.

*Suppl. Fig. 2 and 5 - FSC curves have oddities, especially in the corrected curves. The cryo-EM resolution estimates calculated by CryoSPARC for the UL52 '3-mer' and 4-mer, and UL32 tripentamer are likely overestimated. In the PDB validation files each of the deposited structures has a warning for the resolution estimate "The value from deposited half-maps intersecting FSC 0.143 CUT-OFF 4.31 differs from the reported value 3.32 by more than 10 %", suggesting that the resolution estimates are inaccurate. The authors should provide a resolution estimate using loose masks and generate FSC curves using another software program such as RELION's postprocess to provide resolution estimates. *

Thank you for bringing this to our attention. The differences in the resolution estimates are a known issue and are highly influenced by the tightness of the mask. In the revised manuscript we have updated the FSC curves to not include auto-tightened masks and revised our resolution estimates. This slightly changed the resolution to 3.29 Å for both UL52 3-mer and 4-mer and to 3.09 Å for the UL32 consensus map. Please also see the local resolution estimation maps in Supplementary Figures 2e and 5e for an illustration of the range of resolutions in each map.

Suppl. Fig. 6f and 6g - Is there any visible density that might resemble the EGS crosslinking reagent?

We do not expect to observe density for EGS due to the long flexible linker (~16 Å) between the two reactive groups.

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 manuscript describes the cryo-EM structures of a conserved, necessary, herpesvirus genome packaging accessory factor for human cytomegalovirus (HCMV), UL52, and herpes simplex virus type-1 (HSV-1), UL32. Herpesvirus packaging accessory factors have unknown function but bind dsDNA. The UL52 and UL32 structures revealed a 5-fold symmetry similar to the previous X-ray crystallography structure for Kaposi's Sarcoma-associated herpesvirus (KSHV) ORF68 and the cryo-EM structure of Epstein-Barr virus (EBV) BFLF1. However, HCMV UL52 was reported to form two structures, a 3-mer and 4-mer whereas, HSV UL32 formed a supercomplex of trimeric pentamers (tripentamer) produced by dsDNA binding and crosslinking. Similar to previous studies with ORF68, mutagenesis of HSV-1 UL32 demonstrated the importance of zinc finger residues C297, C308, C544, and H581 for core fold stability and positively charged residues H563, R572 in the central channel in the pentamer for HSV-1 recovery in virus complementation assays. In addition, mutagenesis of K532 and C535 at the tripentamer interface helix reduced virus complementation by 50%. These findings have significant overlap and similarities to previously published experiments and confirm the properties of ORF68 and BFLF1, demonstrating the conserved nature of the required packaging accessory factor for herpesviruses.

Major comments.

The manuscript is generally well written with beautifully presented cryo-EM figures. Unfortunately, the new data seem to muddy the water rather than provide clarification about the role or function of the herpesvirus packaging accessory factor. Furthermore, it is not clear whether the structures presented in the manuscript reflect those produced during HCMV or HSV-1 infection. HCMV UL52 was presented to form two distinct structures, a 3-mer and a 4-mer (Fig. 2a). However, the authors acknowledge that the 3-mer is actually a 4-mer when the threshold for the cryo-EM map is lowered. The density is also visible in the PDB validation report for the 3-mer; EMD-74418. Given that ORF68, BFLF1, and UL32 (Didychuk et al., 2021) form complete pentamer rings, with BFLF1 forming stacked rings, it would seem odd for a protein with conserved function to deviate from a pentamer configuration, suggesting that the structures reported do not reflect the natively produced and functional protein. Unlike ORF68 (Didychuk et al., 2021) and UL32 (Suppl. Fig. 4), dsDNA binding experiments were not performed with UL52. Could the partial pentamers simply be poorly formed due to expression in insect cells (mammalian cells were used for protein purification in Didychuk et al., 2021), absence of dsDNA, or inappropriate buffer conditions? Moreover, were the EM grid and vitrification parameters optimized? Grid geometries and chemistries can have profound effects of protein stability especially in the context of the air-water interface, leading to degradation of protein complexes (Glaeser, 2018; D'Imprima et al., 2019). Does UL52 form complexes with dsDNA? Data are shown for the HSV-1 packaging accessory factor. Perhaps dsDNA would stabilize the UL52 pentamer.

In Didychuk et al., 2021, HSV UL32 is shown to form pentameric rings; negative stained 2D class averages were generated from tagged protein (twin strep tag), produced in mammalian cells (HEK293T), and not purified using size exclusion chromatography. In the present study HSV UL32 was not observed to form pentameric complexes "We first attempted to visualize the pentameric species by negative stain electron microscopy but were unable to identify particles of the expected dimensions." However, it is not clear why this was the case. If the pentameric structures were readily produced in previous experiments, why was cross-linking needed in the current study? As such, the tripentamer complexes seem artifactual in nature. Although the data presented in Fig. 4b suggest that interface residues, K532 and C535, might play a role in the formation of the tripentamer and have a minor role in HSV-1 replication, these experiments are incomplete. Single mutations are needed for each residue to assess their individual contribution to tripentamer formation, evidence for a loss of tripentamer formation is needed, and evidence for protein expression is needed. In the previous negative stain electron micrographs reported by Didychuk et al., 2021, were the higher order tripentamer complexes seen?

Formation of disulphide bonds between cysteine residues in vitro is not indicative of complexes forming in vivo during replication. What evidence is there for disulphide bond formation between packaging accessory factor pentamers for KSHV, EBV, and LCMV? In the present study, the disulphide bond could form due to proximity as a result of the cross-linking and the presence of molecular oxygen rather than a bona fide enzyme catalysed reaction during herpesvirus replication to generate packaging accessory factor tripentamers.

The DNA densities in Suppl. Fig. 6e to 6g are curious. As noted by the authors, the 30mer dsDNAs do not traverse through the central cavity of the pentamer. They appear to make contact with neighboring pentamers, again suggesting that these complexes are artefacts from cross-linking. This should be discussed more thoroughly.

Previously proposed functional roles for ORF68 include a scaffold for terminase assembly, association of the terminase with the portal, generation of initial free ends, or coordination with other replication machinery (Didychuk et al., 2021). Presuming that the new structures for HCMV UL52 and HSV-1 UL32 occur naturally, how do they fit with the previously proposed functional roles of the herpesvirus packaging accessory factor? A more in-depth discussion of this would be valuable.

Minor comments.

A lack of page numbers and line numbers made reviewing this manuscript more challenging than necessary.

As noted in the 'General comments' section above, ORF68 (3.37Å) and BFLF1 (3.60Å) both form pentamers (Didychuk et al., 2021) and were produced in mammalian systems HEK293T cells. Protein purification in the present study was performed in insect (SF9 or High Five) cells. Does this affect complex stability. Also, the tag was retained for UL32 in Didychuk et al., 2021; could this provide stability of the pentamer in the original studies?

Suppl. Fig. 3 is missing.

"UL52 has two regions remodeled" The use of the word 'remodeled' is not appropriate in this context as it implies a single protein can form two shapes under different conditions rather than distinct structures between two disparate proteins; UL52 compared to ORF68. This should be rephrased.

What is the density in the central core of UL52 (Fig. 2a; Suppl. Fig. 2e)? Was any form of focused classification performed to establish the identity of the density within the central pseudocavity?

Does UL52 bind to dsDNA? To support the hypothesis that the herpesvirus packaging accessory factor has conserved functions across the three subfamilies dsDNA binding experiments should be performed. There is no discussion about how these data relate to the previous functional model for ORF68 presented in Didychuk et al., 2021. Do the new data alter the previous functional models?

There are some interesting grammatical phrases; please address throughout the manuscript. One example - "...a notable shared aspiration..." Proteins do not have aspirations. Please use a more formal scientific statement.

Fig. 4b - Statistical analyses missing. Please provide.

Fig. 6c - Statistical analyses are missing. Please provide. Protein folding/expression data missing; see Fig. 5C showing mutations that result in poor protein expression.

Suppl. Fig. 2 and 5 - FSC curves have oddities, especially in the corrected curves. The cryo-EM resolution estimates calculated by CryoSPARC for the UL52 '3-mer' and 4-mer, and UL32 tripentamer are likely overestimated. In the PDB validation files each of the deposited structures has a warning for the resolution estimate "The value from deposited half-maps intersecting FSC 0.143 CUT-OFF 4.31 differs from the reported value 3.32 by more than 10 %", suggesting that the resolution estimates are inaccurate. The authors should provide a resolution estimate using loose masks and generate FSC curves using another software program such as RELION's postprocess to provide resolution estimates.

Suppl. Fig. 6f and 6g - Is there any visible density that might resemble the EGS crosslinking reagent?

Suppl. Fig. 7f - Statistical analyses absent.

References.

Didychuk AL, Gates SN, Gardner MR, Strong LM, Martin A, Glaunsinger BA. A pentameric protein ring with novel architecture is required for herpesviral packaging. Elife. 2021 Feb 8;10:e62261. doi: 10.7554/eLife.62261. PMID: 33554858; PMCID: PMC7889075.

D'Imprima E, Floris D, Joppe M, Sánchez R, Grininger M, Kühlbrandt W. Protein denaturation at the air-water interface and how to prevent it. Elife. 2019 Apr 1;8:e42747. doi: 10.7554/eLife.42747. PMID: 30932812; PMCID: PMC6443348.

Gardner MR, Glaunsinger BA. Kaposi's Sarcoma-Associated Herpesvirus ORF68 Is a DNA Binding Protein Required for Viral Genome Cleavage and Packaging. J Virol. 2018 Jul 31;92(16):e00840-18. doi: 10.1128/JVI.00840-18. PMID: 29875246; PMCID: PMC6069193.

Glaeser RM. PROTEINS, INTERFACES, AND CRYO-EM GRIDS. Curr Opin Colloid Interface Sci. 2018 Mar;34:1-8. doi: 10.1016/j.cocis.2017.12.009. Epub 2017 Dec 22. PMID: 29867291; PMCID: PMC5983355.

Significance

General assessment: The strengths of this manuscript are the structural information provide by the cryo-EM maps for the HCMV UL52 and HSV-1 UL32 and the mutagenesis studies that corroborate previous studies for the packaging accessory factor for gammaherpesviruses KSHV and EBV. However, there are limitations. These are centered on whether the structures are representative of UL52 and UL32 complexes produced during replication rather than over expression in insect cells and stabilization using chemical cross-linking.

There is a lack of novelty in the context of the herpesvirus packaging factor. The pentameric architecture, DNA binding, zinc fingers (4), and charged residues required for DNA binding were conclusively demonstrated in previous studies (Gardner and Glaunsinger, 2018; Didychuk et al., 2021). Thus, the novelty comes from the different pentameric structures; UL52 4-mer and UL32 tripentamer. However, if these are artefactual structures due to the expression system (mammalian versus insect) used, air-liquid interface induced protein instability, or cross-linking, the novelty is lost. That's not to say the data are not informative for the herpesvirus community.

Advance: The advance in this manuscript is the new structural information for the UL52 and UL32. Even if the higher order complexes are potential artefacts, high resolution structure information for the subunit is especially informative. The mutagenesis data for UL32 are also informative in that the provide important information about a conserved and necessary protein needed for herpesvirus replication and has the potential to be used as a novel druggable target.

Audience: The manuscript will appeal to specialized and broad audiences and could influence research into antiviral therapies for herpesviruses. My field of expertise is herpesvirology, structural biology, and cryogenic electron microscopy modalities,

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

Bailey et al investigate DNA packaging accessory factors of various herpesviruses. The central findings are the cryo-EM structures of the accessory factors from HSV1 and HCMV. Combined with the corresponding author's previous structure of the KSHV accessory factor, these new findings now provide a window into packaging in all three families of human herpesviruses. They reveal that the overall structure of a ring of pentameric symmetry is conserved, the overall oligomeric stabilities are not conserved across all herpesviruses. Moreover, the authors have the important finding that basic residues in the ring pore are required for viral replication. Overall, this study represents a strong extension of the authors' study of packaging accessory proteins, with solid data and very few concerns to be addressed.

Major issues.

1) There is a high probability that the tripentamer is an artifact of the cross-linking. Because of this, it'd be great to know more about the cross-linking reaction, ideally mass spec identification and quantification of cross-links. This would also address the authors' speculation of contacts that stabilize the tripentamer. Moreover, the author's claim "However, mutation of K532A/C535A reduced infectious virion production by half (Fig. 4b), suggesting that the tripentamer interface may play a role in the viral life cycle." Seems to be an overreach. Perhaps this is semantics but the data just show that these residues play a role in viral replication (albeit not a huge role based on the modest effect).

2) The density for the potential DNA does not look very convincing, although it still remains the strongest hypothesis. The authors should try to strengthen their argument. Does this putative DNA would contact residues that they show are necessary for viral replication? Showing seq conservation on the sturcutre could help their argument for the shared function of DNA-binding.

3) My last major issue is stylistic and concerns the descriptions of cryoEM structures. I found that the paper was a bit of challenge to read when the authors would introduce each structure. It was a bit of a slog to get through. Descriptions of the structures veered off into overly detailed comparisons that required constant comparison with the figure and didn't really advance my understanding past "the outer surfaces of the three orthologs are different." This masked the more interesting aspects of the authors' findings. Perhaps this could be summarized in supplementary figures or a table. Because this is a stylistic suggestion, the authors should feel free to ignore this request.

Minor comments

1) The descriptions of structure determination in the text were often unclear. For example, "In the 3-mer map, a poorly-resolved fourth protomer is visible at low contour levels, suggesting that an additional protomer is present but highly flexible in this class (Supplementary Fig. 3a)." Alternatively, it could be that the classification algorithm wasn't able to fully separate particles that were 3-mers from the 4mers. When describing the structure determination of the HSV1 accessory factor, the authors describe no other particles other than the tripentamer. Were there other particles observed? It'd be a bit surprising that all of the protein adopted the tripentamer state. Was symmetry applied, particularly for the tripentamer that appears to have C-3 symmetry? This is in materials and methods but not clear why it isn't mentioned when describing the structure determineation and results.

2) Throughout the paper, the authors use the word "remodel" to describe structural differences between orthologs. However, this word usually carries the implication of conformational rearrangement within a protein, and not across orthologs. Please consider a different description.

3) Figure 2F is confusing and difficult to interpret. It seems that the main point is that these interfaces are conserved, which might be more easily displayed as a standard sequence conservation score mapped onto the structure. I'm also not sure that this figure is necessary as a main figure and could be supplemental.

4) The authors write "UL32 bound to the shortest probe tested (10 bp, Supplementary Fig. 4a)." This implies that ONLY the shortest probe is bound and that others are not bound. Consider rephrasing.

5) Frustum is misspellt. ;)

6) In the discussion, the authors speculate that the variability of the outer surface is due to "virus- or host-specific interactions". I'm confused by "host-specific interactions", because the host is the same for all three viruses. Perhaps the authors mean that the different accessory factors could interact with different host factors? If so, are the authors making a Red Queen argument? If so, it'd be pretty cool to do dN/dS analysis to test that hypothesis.

Significance

This paper represents an advance in the field of genome packaging. The herpesvirus packaging mechanism is still mysterious, and the role of this accessory factor is one of the biggest gaps in knowledge. Although this study doesn't uncover the role, this provides new details into the evolution of this factor across the herpesvirus lineages. To me, this window into evolution of this factor is the biggest advance of the work, and tbh I felt that the authors could lean into this a bit more in the discussion section. Are there any differences in the packaging mechanisms of the different herpes families that can be related to their different behavior? Any other molecular evolution analyses (e.g. dN/dS ratio analysis) that could inform their study?

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

Bailey et al. present the results of their structural analysis of packaging factors encoded by the β-herpesvirus human cytomegalovirus (UL52) and the ⍺-herpesvirus herpes simplex virus type 1 (UL32). The authors have previously published structures for orthologous proteins in the γ-herpesviruses Kaposi sarcoma associated herpes virus (ORF68) and Epstein-Barr virus (BFLF-1), showing both to form pentameric rings having a positively charged central channel. Here HCMV UL52 is found to form 3-mer and 4-mer assemblies that resemble incomplete pentameric rings. The complexes are formed by a screw displacement however, having both rotation about- and translation along the central axis. The structure shows fold conservation with the previously described structures, including preservation of the positively charged central channel.

Attempts to image HSV-1 UL32 were initially unsuccessful, despite light-scattering analysis indicating the presence of pentamers. DNA binding is shown by EMSA, but these complexes were also not stable for cryo-EM analysis. Chemical crosslinking of the DNA bound complex was therefore employed, resulting in production of higher-order assemblies including one comprising three pentamers, that was successfully resolved by cryo-EM. Interestingly focussed classification analysis highlighted the presence of rod-shaped density passing through the central channels of two pentameric rings in this complex. Mutation of the interface that gives rise to the formation of tripentamers reduced progeny virion production by half, leading the authors to suggest that this complex may be a biologically important assembly.

The importance of zinc-fingers identified in these structures was probed showing that mutation abolishes protein production. Similarly, mutation of the positively charged residues lining the central channel of HSV-1 UL32 greatly reduced or completely ablated progeny virion production in an assay where either WT or mutant UL32 was transfected into cells to complement UL32 knockout virus.

Overall, I found the manuscript very easy to read and the analysis appears to be expertly performed. I have no substantive criticisms of the work and think it would be suitable for publication in its current form, or subject to some small edits.

Minor comments

The authors suggest that the weak 4th protomer in the HCMV UL52 3-mer map is a consequence of flexibility. This may be the case, but it may also be the case that the class is polluted with 4-mer particles leading to reduced occupancy. Erasing the weak density and running a multi-model 3D classification providing the erased 3-mer and a 4-mer starting map may separate these.

I found the supplemental figure to show the DNA in the tripentamer map too small, this is an interesting finding and should be shown more clearly.

Significance

Herpesviruses are important pathogens of humans and are biologically complex systems. The structural analysis of this essential packaging co-factor is an important contribution to the field. It builds on the previous work by this group concerning the packaging factors of the gamma-herpesviruses KSHV and EBV. I consider this paper to be high-quality and worthy of publication in a very good journal with a microbiology/virology, biochemistry or molecular biology focussed readership. The process of genome packaging in herpesviruses is not as well characterised as in bacteriophages (and even in that case it is not well understood). This work provides important knowledge that will support future studies on this critical process in herpesvirus replication.

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

Responses to the reviewers

First of all, we would like to thank all the reviewers for their valuable and constructive comments on our manuscript. We have considered each comment and revised the manuscript accordingly. We respond to each comment below in blue font.

To Reviewer #1

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

*Iwase et al have used multiomics and spatial transcriptomics to comprehensively map neural crest cell contributions to the mouse heart and great arteries. This careful and detailed analysis reveals changes in the transcriptional profile of neural crest cells as they give rise to different regions and cell types in the heart and great vessels. The study significantly builds on a number of recent scRNA-seq analyses of neural crest cell development and includes development of a new informatic tool for regulatory network investigation. Among the new findings documented are downregulation of Hox gene expression in intracardiac crest cells and regulation of Sox9 by Meis transcription factors. Addressing the following points would improve clarity and accessibility. *

Thank you for your encouraging feedback and comments. We have responded to your comments below.

* In Figure 1C it is difficult to visualize all the colors given the mixed contribution of NCC and nonNCC cells to mesenchyme. Please also show YFP transcript distribution in NCC versus nonNCC plots. In addition, it would be helpful to show plots for both NCC and nonNCC for Gata4 and Tbx20. *

To improve visualization, we separated EYFP-positive NCCs and EYFP-negative non-NCCs into distinct plots (Figure 1e), rather than displaying EYFP transcript distribution within a single combined plot. In addition, we have included separate plots for both NCCs and non-NCCs showing the expression of Gata4 and Tbx20 (Figure 1g, h).

 Furthermore, in the revised manuscript, we subdivided the original clusters c5 and c13 into two subclusters each, resulting in a total of 23 clusters in the UMAP shown in Figure 1. This refinement was introduced to facilitate clearer interpretation and subsequent analyses.

* The authors identify a cardiomyocyte cell cluster in their integrated NCC scRNA-seq plots. Are these cells labelled by Wnt1-Cre in the authors' own dataset? Is the trajectory analysis informative as to the steps preceding acquisition of cardiomyocyte fate? *

A total of 20 EYFP-positive NCCs in our own dataset were assigned to the cardiomyocyte cluster in the integrated UMAP. Of these, 6 cells were located within the cardiomyocyte cluster (c21), while the remaining cells were classified as pharyngeal mesenchyme but positioned in close proximity to c21 in the original UMAP (Figure 1e). Although this observation is consistent with previous reports showing that NCCs give rise to cardiomyocytes, the small number of cells precluded meaningful characterization or trajectory analysis of cardiomyocyte fate acquisition. Accordingly, we have addressed this point only briefly in the revised manuscript as follows:

Only a few NCCs were detected within cardiomyocyte clusters, which were predominantly composed of non-NCCs, consistent with previous reports demonstrating NCC differentiation into cardiomyocytes15,16. The overall number of cardiomyocytes was low, likely reflecting the restricted sampling of the cardiac outflow tract (Figure 1a).

(P6, L20–25)

* Linked with this point, is it possible that there are nonNCC cells in the integrated plots? Of note, many of the NCC genes overlap with genes that have also been shown to be expressed in mesodermal cardiac progenitors (including Osr1, Pparg, Dlk1, Tcf21, Ebf2, Tbx20, Sox9). For example, is it possible to distinguish NCC derived smooth muscle within the heart from cells originating from the second heart field that may express smooth muscle genes? Cluster 27 for example appears broadly expressed in the region of ventricular outlets in Figure 3. Comparison with YFP transcript distribution may be helpful here. *

In principle, non-NCCs were excluded from the integrated plots shown in Figure 3. However, we cannot completely rule out the possibility that a small number of non-NCC cells were inadvertently included, for example due to false-positive signals during cell isolation. In our dataset, NCCs and non-NCCs were stringently distinguished based on FACS profiles, detection of EYFP transcript reads in the RNA-seq data, and rigorous exclusion of doublets and low-quality cells.

 Regarding the distinction between NCC- and non-NCC-derived SMCs, a key challenge lies in defining comparable cell populations. We first validated the annotations of SMC clusters (C4, C23, and C27) using immunostaining for Myh11, Sost, and Reln, confirming consistency with their transcriptomic identities. We then re-clustered the SMC populations and projected non-NCC SMCs (clusters 4 and 20 in Figure 1d) onto this UMAP. These non-NCC SMCs were mapped to clusters corresponding to great artery and coronary artery SMCs (C27 and C23, respectively).

 However, we lack confidence that these projected populations are directly comparable. For example, non-NCC SMCs mapped to C23 or C27 may not necessarily represent bona fide coronary or great artery SMCs from equivalent anatomical regions, and could include other SMC subtypes such as venous SMCs or pericytes. Given the known regional heterogeneity of SMCs and the absence of strict spatial matching criteria, such comparisons would be difficult to interpret. This limitation is further compounded by the relatively small number of cells available.

 For these reasons, we focused on spatial validation of cluster annotations by immunostaining in this study, and have reserved detailed comparisons between NCC- and non-NCC-derived SMCs for future work. We believe this does not detract from the overall consistency or value of the present study.

* Can the authors add any validation of key expression patterns, for example using fluorescent in situ hybridization? *

Figures 2n-y present Xenium-based multiplexed fluorescent in situ hybridization data that validate the spatial expression patterns of marker genes characterizing NCC derivatives in pharyngeal mesenchyme, intracardiac mesenchyme, and SMC populations. In addition, we have incorporated new Xenium images highlighting key gene expression patterns in the aorticopulmonary (AP) septum at E12.5 (Figure S4), supporting the annotation of cluster C16 in the integrated UMAP as corresponding primarily to the AP septum. We have also added immunostaining data for Myh11, Sost, and Reln to further validate the annotations of SMC clusters (see the response above). Together, these data provide independent spatial confirmation of the transcriptional signatures identified in our single-cell analyses. Based on these data, we revised the relevant section of the Results as follows:

The SMC clusters, which were continuous with the pharyngeal mesenchyme via transitional populations in the UMAP, were identified by high expression of the mature SMC marker Myh11 (Figure 3i). Differential gene expression analysis further distinguished individual clusters (Figure S4a-g). Among these, C27 displayed a transcriptomic profile characteristic of the great artery SMCs, including high expression of Sost (Figure S3j). C4 was enriched for Tfap2b and Ptger4 (Figure S3j), markers of the ductus arteriosus SMCs21,22, supporting its annotation. C0 and C7 likely represent transitional states between pharyngeal mesenchyme and differentiated lineages, potentially bifurcating toward great artery SMCs or cardiac cushion mesenchyme (Figure S3a and Table S6). C23 was characterized by high expression of Gja4, a marker of coronary artery SMCs, along with pericyte markers Kcnj8 and Rgs5 (Figure 3j and Figure S3k), corresponding to the cluster similarly annotated by Chen et al14. In addition, C23 was also distinguished from C4 and C27 by its expression of Reln (Figure S4d).

• Immunostaining supported these cluster annotations. Sost expression is observed in great artery SMCs but not in ductus arteriosus and coronary artery SMCs, whereas Myh11 expression was higher in ductus arteriosus and coronary artery SMCs than in aortic SMCs (Figure S4f-m). Furthermore, Reln expression was restricted to coronary artery SMCs (Figure S4n-s).*

(P8, L30 – P9, L14)

* Please elaborate on the decoded Hox code patterns that appear to be indicative of arch origins. Do the results allow determination of whether the trajectories to different cardiac fates inferred in Figure 3D differ in different arches? *

• *

In response to the reviewer’s suggestion, we have further elaborated on the decoded Hox code patterns indicative of pharyngeal arch origin and examined whether trajectories toward distinct cardiac fates differ between arch-derived NCC populations (see new supplemental figure).

To further delineate Hox code patterns associated with pharyngeal arch origin, we stratified the integrated UMAP by distinct Hox expression profiles (Figure S9). Cells expressing any Hox2 paralog, but lacking Hox3–5 paralogs, were defined as PA2-derived preotic NCCs, whereas cells expressing any of Hox3–5 paralogs were classified as PA3/4/6-derived postotic NCCs. Preotic, postotic, and Hox-negative populations were then projected onto the integrated UMAP across developmental stages (E10.5–E14.5). Trajectory inference indicated that transitions toward intracardiac mesenchyme occur earlier in preotic cells (E10.5) than in postotic cells (E11.5), consistent with their known sequential migration into the cardiac cushion8. From E12.5 to E14.5, postotic cells showed a progressive emergence of the aorticopulmonary septum–associated cluster C16 from transitional states. Notably, the proportion of Hox-negative cells increased within intracardiac mesenchyme, except in C16 where Hox expression was retained, supporting the notion that Hox genes are broadly downregulated in cushion-associated intracardiac NCCs (Figure 4k, S9).

(P11, L24 – P12, L4)

* The authors need to explain why the authors place an arrow from mesenchymal cluster 18 to 23 in Figure 3D while the trajectory analysis in 3C predicts the opposite direction. *

RNA velocity analysis of scRNA-seq data is fundamentally based on splicing dynamics. The original framework assumes that transcriptional induction and repression persist long enough for cells to reach active (transcribing) or inactive (silenced) steady-state equilibrium. However, this assumption is often violated during cell differentiation, where transient cell populations frequently exhibit rapidly changing mRNA levels that do not reach steady-state equilibrium. To address this limitation, the scVelo method was developed (Bergen et al., 2020, Nature Biotechnology), and we applied this approach to the integrated NCC datasets in the present study. This analysis successfully inferred directional flows from the pharyngeal mesenchyme toward SMCs and intracardiac mesenchyme through transitional states (Figure 3c,d). However, as the reviewer correctly pointed out, the analysis predicted a directional flow from C23 to C18, apparently opposite to the biological directionality supported by previous findings (see P9, L15–19).

 We consider this discrepancy to reflect intrinsic limitations of RNA velocity analysis. As discussed by Bergen et al., in systems containing multiple lineages and cellular processes, differences in gene regulatory networks among heterogeneous cell states can generate multiple trajectories in phase space owing to distinct splicing kinetics. In addition, incompletely captured splicing kinetics may represent only a limited portion of the overall dynamics, particularly near terminal differentiation states. In such cases, phase portraits of unspliced versus spliced transcripts may appear nearly linear rather than curved, potentially leading to erroneous assignment of positive or negative RNA velocities. Consistent with this limitation, we obtained opposite directionalities between C18 and C23 depending on whether steady-state or dynamical models were applied and according to different parameter settings. Through these repeated computational re-evaluations of lineage directionality, we concluded that RNA velocity analysis is suitable for capturing the global landscape of differentiation flow, but that accurate inference of local lineage directionality may require careful model selection and parameter optimization to ensure consistency with established biological evidence.

 To avoid arbitrariness and potential confusion, we removed the arrow between C18 and C23 from the revised Figure 3d. Instead, we now describe the observed continuity between these populations in the Discussion section as follows:

Notably, the UMAP revealed a continuum between C23 and C18 within the intracardiac mesenchyme population. Given previous findings that the proximal coronary artery SMCs originate from preotic NCCs8 and that pericytes give rise to coronary artery SMCs23, this connection likely represent a differentiation trajectory from intracardiac mesenchyme to coronary artery SMCs via a pericyte-like intermediate stage.

(P9, L15–19)

The continuity between intracardiac mesenchyme and coronary artery SMCs through a pericyte-like intermediate state is consistent with previous developmental studies showing that proximal coronary artery SMCs originate from preotic NCCs and may arise through pericyte intermediates8,21.

(P18, L24-27)

* The authors nicely show downregulation of Hox gene expression in NCC cells entering the heart. Can they add discussion of any insights into this from prior studies of loss or gain of Hox gene function? *

• *

We have added the following discussion on the roles of anterior Hox genes in cardiovascular development, together with appropriate references on loss or gain of Hox gene function:

• The regional identities of pharyngeal NCCs that contribute to cardiac development are established by Hox genes and their associated regulatory networks. Genetic studies have demonstrated essential roles for the anterior Hox genes in patterning the pharyngeal arch artery system and semilunar valve structures. Loss of Hoxa1 and Hoxb1 results in severe defects in pharyngeal arch artery development42, whereas ectopic or sustained expression of Hoxb1 in NCCs disrupts cardiovascular morphogenesis and causes malformations of the great arteries and semilunar valves43. Likewise, Hoxa3 contributes to proper patterning of the pharyngeal arch region and its NCC-derived derivatives44,45. These findings underscore the importance of precise spatial and temporal regulation of Hox genes during cardiovascular development. Among pharyngeal NCCs contributing to cardiac development, cushion-independent NCC derivatives (great artery SMCs and the aorticopulmonary septum) retain their origin-specific Hox-codes. In contrast, cushion-associated NCC derivatives (coronary artery SMCs and valvular/subvalvular interstitial cells) downregulate Hox expression and transition toward region-specific GRNs involving TFs such as Tbx20 and Gata4, whose expression is known to be induced by BMP signaling in cardiomyocytes46,47. Bmp2 and Bmp4 are expressed in the regions of the pericardial reflection traversed by NCCs en route to the cardiac cushion48. Together, these observations suggest that appropriate repression of Hox programs, coupled with activation of cardiac-specific regulatory networks, is required for normal differentiation of cushion-associated NCC derivatives.*

(P18, L30 – P19, L16)

• Roux, M. et al. Hoxa1 and Hoxb1 are required for pharyngeal arch artery development. Mech. Dev. 143, 1–8 (2017).*
• Zaffran, S., Odelin, G., Stefanovic, S., Lescroart, F. & Etchevers, H. C. Ectopic expression of Hoxb1 induces cardiac and craniofacial malformations. genesis 56, (2018).*
• Chisaka, O. & Capecchi, M. R. Regionally restricted developmental defects resulting from targeted disruption of the mouse homeobox gene hox-1.5. Nature 350, 473–479 (1991).*

• Kameda, Y., Watari-Goshima, N., Nishimaki, T. & Chisaka, O. Disruption of the Hoxa3 homeobox gene results in anomalies of the carotid artery system and the arterial baroreceptors. Cell Tissue Res. 311, 343–352 (2003).*

  In addition, we previously generated conditional Hoxa2 overexpression mice and demonstrated that ectopic Hoxa2 expression in Hox-negative PA1 cranial neural crest derivatives induced PA2-like structures, indicating a partial homeotic transformation (Kitazawa et al., Developmental Biology, 2015; 10.1016/j.ydbio.2015.04.007). Because cardiovascular phenotypes were not examined in that study, we have now resumed breeding of these mice for detailed cardiovascular phenotypic analyses.

  In parallel, we have also established knockout mice for a downstream target of Hoxa2, which are expected to complement the Hoxa2 gain-of-function model and provide further insight into the regulatory mechanisms underlying cardiac NCC differentiation and patterning. Although we have obtained preliminary observations from these models, a comprehensive analysis is still ongoing, and we therefore prefer to reserve these results for a future study with more detailed investigation.*

• Figure 3Y could be simplified to more clearly distinguish the two types of Meis binding sites. For example, it may be helpful to reorder the mesenchymal cell types based on Hox expression. *

To improve clarity and better distinguish the two types of Meis binding sites, we have reordered the heatmap of motif enrichment based on the hierarchical clustering with the updated JASPAR2024 database. In parallel, we have revised the heatmap of transcription factor gene expression to provide a more consistent and interpretable presentation. These diagrams were now presented as Figures 4k and 4l of the revised manuscript.

* The authors provide nice in vitro and in vivo evidence for an upstream role of Meis transcription factors in regulating Sox9 expression. Can the authors identify from the enhancer sequence (or their transcriptomic dataset) any of the non-Hox transcription factors that Meis may be working with here? Please discuss the significance of Sox9 expression in epicardium driven by the same enhancer. Might this regulation also operate in second heart field progenitor cells where both genes are expressed? It is not evident in Figure 7 that Sox9-EGFP is also expressed in epicardium. *

The distal Sox9 enhancer containing Meis2 binding site that we identified (chr11-112850240-112851186) also contains several consensus motifs including predicted Hand2- and Nfatc1-binding sites. However, to our knowledge, these transcription factors have not been reported as non-Hox partners of Meis proteins. Their recruitment to this enhancer, as well as potential cooperative interactions with Meis transcription factors, were not examined in the present study and remain subjects for future investigation.

 As the reviewer pointed out, Sox9 was expressed not only in intracardiac NCCs but also in the epicardium (revised Figure S13). Consistent with this expression pattern, distal *Sox9* enhancer was accessible in both intracardiac mesenchyme and epicardial cells (revised Figure S13). Especially, *Wt1*low/*Sox9*high mesenchymal cells, likely derived from the epicardium via epithelial-mesenchymal transition, also exhibit chromatin accessibility at this enhancer comparable to that observed in NCC-derived mesenchymal cells. These findings suggest that the same regulatory element may function across multiple cardiac lineages.

 We have addressed these points in the revised manuscript as follows, including additional supporting data in the supplementary figures.

*Enhancer activity in the epicardium corresponds to Sox9 expression and an open chromatin peak at the putative distal enhancer region in clusters 22 and 5 in Figure 1d, which represent Wt1high epicardial cells and intracardiac mesenchyme likely including Wt1low epicardial EMT derivatives, respectively (Figure S13). *

(P16, L28–32)

• *

• Could this approach yield similar data for Osr1? Please clarify if there is any experimental evidence supporting the predicted negative regulation of Sox9 by Osr1 in the heart illustrated in Figure 8. *

There are currently no experimental data demonstrating Sox9 repression by Osr1 in the heart. However, such an effect has been reported in tongue and limb mesenchyme (Liu et al., PNAS, 2013), as noted in P16, L5–7. Although direct experimental validation, such as Osr1 overexpression in cardiopharyngeal NCCs, would provide stronger evidence, preparation of this specific NCC lineage is difficult. Given this limitation, we instead performed in silico gene perturbation analysis using CellOracle, which predicted antagonistic roles for Sox9 and Osr1 during lineage bifurcation from pharyngeal NCCs. Because this regulatory relationship has not yet been experimentally validated in the cardiac context, we revised the illustration accordingly by adding a question mark to indicate the hypothetical nature of this antagonism (Figure 9 in the revised version).

* Concerning the links between valve mesenchyme and skeletogenic programs it would be relevant to cite the earlier work of Lincoln and Yutzey (reviewed in PMID: 16643886): *

• *

We cited the suggested work in the relevant portion of the Discussion section as follows:

By analogy, Sox9high/Scxhigh NCCs at the base of semilunar valves may form a structural attachment unit linking cushion tissues to valvular leaflets57.

(P20, L30–32)

• Lincoln, J., Lange, A. W. & Yutzey, K. E. Hearts and bones: Shared regulatory mechanisms in heart valve, cartilage, tendon, and bone development. Dev. Biol. 294, 292–302 (2006).* *
• In order to increase accessibility of the dataset the authors are encouraged to include a browser link. *

We agree with the reviewer that improving dataset accessibility is important for facilitating exploration of NCC diversity. Therefore, we have already uploaded our original fastq sequence files and count matrices in the DDBJ and GEO servers. In addition, we will upload our gene expression datasets projected onto the UMAP in UCSC cell browser, enabling readers to more easily visualize and interrogate the dataset.

Minor points:

*13. The authors could rephrase the title since the term topographical genetic switch is unclear. *

• *

We thank the reviewer for this suggestion. To improve clarity, we revised the title as follows:

Hox–Meis-relayed spatial gene regulatory transition underlies cardiopharyngeal neural crest diversification revealed by multimodal analysis * 14. In the introduction, with reference to the De Bono study, please note that Tbx1 was shown to regulate pharyngeal NCC differentiation stage transitions non-cell autonomously. *

According to the reviewer’s suggestion, we rephrased the relevant section of the introduction as follows:

De Bono et al. elaborated the transition of pharyngeal NCCs through multiple differentiating stages toward SMC fates, identifying Tbx2 and Tbx3 as key TFs in this process13. They also showed that Tbx1, the gene for 22q11.2 deletion syndrome, regulates pharyngeal NCC differentiation stage transitions non-cell autonomously13.

(P5, L1–4)

Reviewer #1 (Significance (Required)):

*Iwase et al have used multiomics and spatial transcriptomics to comprehensively map neural crest cell contributions to the mouse heart and great arteries. This careful and detailed analysis reveals changes in the transcriptional profile of neural crest cells as they give rise to different regions and cell types in the heart and great vessels. The study significantly builds on a number of recent scRNA-seq analyses of neural crest cell development and includes development of a new informatic tool for regulatory network investigation. Among the new findings documented are downregulation of Hox gene expression in intracardiac crest cells and regulation of Sox9 by Meis transcription factors. *

Again, thank you for giving us the opportunity to strengthen our manuscript with your valuable comments and queries. We have worked hard to incorporate your feedback and hope that these revisions persuade and satisfy you.

To Reviewer #2

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

Summary *In this manuscript, Iwase et al. cleverly make use of different modalities, spatial transcriptomics and single-cell omics datasets, in conjunction with a well-established Wnt1-Cre;R26R-EYFP line to trace neural crest cells (NCCs) contributing to the cardiovascular system during embryonic development in the mouse. By doing so, the authors identified a bifurcation between cardiac NCCs contributing to the OFT cushions and forming the aorticopulmonary (AP) septation complex. Thus, the authors split "intracardiac NCCs" into two different NCC programs/compartments, even though both reside in the broad OFT region. The NCCs that enter and associate with the OFT cushions undergo a Hox off transition (Hox-positive to Hox-negative once intracardiac), with a corresponding shift in Meis binding and GRN wiring. The authors propose these cells pass through a Meis2-Sox9-Scx "skeletogenic progenitor-like" intermediate and contribute to semilunar valves and coronary artery smooth muscle. By contrast, the NCCs assigned to aorticopulmonary septum (APS) formation and great vessel smooth muscle retain a distinct Hox codes. *

Thank you for your encouraging feedback and comments. We have responded to your comments below.

Major Comments *1. The manuscript would benefit from clearer delineation between the different NCC contributions, particularly for non-specialist readers. The distinction between (i) CNCCs in OFT cushions and (ii) CNCCs forming the aorticopulmonary (AP) septation complex is not adequately explained. While both populations contribute to OFT septation, according to the authors, they represent distinct compartments with different developmental trajectories. The authors could clarify this using anatomically labelled hearts at the stages they conduct their analysis, along with additional text and schematics explaining what is meant by each compartment. This would greatly enhance the accessibility of the manuscript. *

To clarify the distinction between the two components of intracardiac NCCs, cushion-associated mesenchyme and cushion-independent aorticopulmonary (AP) septum, we substantially revised the description of cluster characterization (P9, L24–P10, L6). We added new figures (Figure S5) showing their spatial relationships and distinct gene expression signatures, including E12.5 Xenium data demonstrating Vegfc expression in cushion-associated NCCs and Tcf24 expression in AP septum NCCs.

 In addition, we revised the schematic diagram of intracardiac NCC distribution in Figure 4m and added Figure S9, which spatially delineates three major NCC trajectories with distinct Hox codes: (1) migration of preotic NCCs into the outflow tract cushions, (2) migration of postotic NCCs into the outflow tract cushions, and (3) protrusion of the AP septum from the dorsal wall of the aortic sac. We also incorporated additional explanatory text and cited relevant review articles on cardiac outflow tract development. We hope that these revisions substantially improve the clarity and accessibility of the manuscript, particularly for non-specialist readers.

• The Methods state that "Decomposition of cell clusters of scRNA-seq was performed by RCTD to map them onto the Xenium dataset," but this description is insufficient. The authors should clarify whether RCTD was applied separately for each developmental stage (i.e., E11.5 scRNA-seq reference for E11.5 Xenium, E12.5 reference for E12.5 Xenium), or whether a pooled reference was used across stages. This clarification is important because RCTD performance depends critically on the correspondence between reference and target datasets. Using a multi-stage integrated reference to deconvolve stage-specific spatial data could introduce artifacts, as cell states and cluster compositions vary considerably across developmental timepoints. *

The pooled scRNA-seq data of NCCs or non-NCCs at E11.5 and E12.5 were used for RCTD to decompose for the spatial allocation in the Xenium dataset. We agree with the reviewer that a multi-stage integrated reference may introduce artifacts. In fact, datasets at E11.5 and E12.5 were similarly distributed in UMAP space and exhibited similar transcriptomic signatures, whereas those at E14.5 and E17.5 demonstrated different characteristics in the integrated UMAP in Figure. 1c. Therefore, we used only E11.5 and E12.5 datasets for RCTD decomposition. We added the bellow sentence in Method section.

The pooled scRNA-seq data of NCCs or non-NCCs at E11.5 and E12.5 were used for RCTD.

(P33, L6)

• *

• Figure 2j-m needs annotations and schematics. It is currently very difficult to identify the different compartments. See Figure C in Chen et al. for an example of this approach. In addition, what are the fine clusters from 1 to 20? Which ones are NCC-derived? *

We added anatomical annotations to the revised Figure 2a, b, j–m to facilitate identification of the different compartments. The fine clusters labeled 0–20 in the previous Figure 2j–m corresponded to the multiome clusters (NCC and non-NCC) shown in Figure 1d. In the revised manuscript, the same dataset was re-clustered into 23 clusters (0–22), which were subsequently used for decomposition analysis to predict cell compartments with maximum likelihood. We also revised the color scheme of the segmented cells in Figure 2j–m to improve visual distinction between compartments and facilitate interpretation of the spatial distribution patterns.

 Putative NCCs were identified through the following procedure:

1. EYFP expression was estimated across 39 cell types in the Xenium dataset by integration with the single-cell multiome dataset (including both NCCs and non-NCCs) using Tangram.
2. Xenium clusters enriched for EYFP expression, defined as clusters whose mean estimated EYFP level exceeded the threshold corresponding to the 65th percentile across all spots, and consistent with known neural crest derivatives were extracted as putative NCC populations.
3. For each spot within these EYFP-enriched Xenium clusters, RCTD was used to estimate the corresponding multiome cluster identity. Thus, the diagrams in Figure 2j–m indicate the most likely multiome subpopulation assignment for each putative NCC spot, rather than categorizing the multiome clusters themselves as NCC- or non-NCC-derived. This description has been included in the Method section (P33, L6–14).

* The panels show pharyngeal markers, OFT/intracardiac markers, and SMC markers in Figures 2n-y, but could the authors show the proportion of NCC-derived (YFP+) cells for each cluster? Could the authors also map only the YFP+ cells on the Xenium data? It would be useful to see the proportion of YFP-positive (NCC-derived) cells for each delineated compartment. YFP-positive cells appear to exist at the boundary between LV and RV in the septum, this observation would benefit from proper quantification. *

Our Xenium analysis could not detect EYFP signals; therefore, as noted above, we estimated EYFP expression by integrating the scRNA-seq and Xenium datasets using Tangram. To clarify the relative enrichment of distribution for each Xenium cluster, we summarized the mean estimated EYFP expression as bar plots (Revised Figure S2q, r). Consistent with their established neural crest origin, Xenium clusters annotated as neuron (38), ganglion (32), and Schwann cell (17) showed high estimated EYFP expression. In addition, Xenium clusters 15 (SMC) and 25 (cushion mesenchyme) were also enriched for EYFP expression.

 Unfortunately, currently available algorithms for Xenium data analysis do not reliably allow visualization or extraction of gene expression profiles exclusively from putative EYFP-positive cells. Instead, we improved the visualization of the spatial distribution of putative EYFP-positive cells by replacing centroid-based signal display with segmentation-based rendering (Revised Figure 2f, g).

 As pointed out by the reviewer, putative YFP-positive cells appear to be present at the boundary between the LV and RV within the interventricular septum. However, the estimated EYFP signals in this region were substantially lower than those observed in well-established neural crest derivatives, suggesting that most of these signals likely represent background noise, although a minor population of sparsely distributed neural crest-derived cells cannot be excluded. More precise characterization of potential neural crest derivatives in the ventricular region will require future investigation.

* Figure 3 is confusing because it integrates data from multiple overlapping stages (E8.5 to P7). While the authors identify distinct compartments, pharyngeal mesenchyme, intracardiac mesenchyme, and SMCs, it is unclear why stages beyond E14.5 and E17.5 (corresponding to the initial single-cell omics and Visium analysis) were included. Although leveraging additional datasets is a clever approach, the integration of data from such disparate developmental timepoints confounds interpretation. For example, Cluster 6 appears to include cells from both P7 and E12.5 stages. Given that the Visium data represent hearts at E14.5 and E17.5, it is problematic to map clusters derived from other stages onto these spatial datasets. *

We integrate data from multiple overlapping stages (E8.5 to P7) in Figure 3 to capture diachronic cell identity and also stage-specific features, particularly within intracardiac mesenchyme and SMC populations. This approach enabled us to assess core lineage relationships, including trajectories linking pharyngeal mesenchyme to intracardiac mesenchyme and SMCs, each comprising multiple distinct subpopulations. These findings support the biological relevance of the integration as a framework for understanding lineage relationships across developmental time. However, we agree with the reviewer that it is problematic to map clusters derived from other stages onto these spatial datasets. Indeed, we restricted the spatial mapping analysis to cells from E14.5 and E17.5 within the integrated dataset, thereby ensuring consistency with the developmental stages represented in the Visium data. To address concerns about this issue, we have clarified this point in the Methods section by adding the underlined words in the following sentence.

Spatial mapping of scRNA-seq data onto Visium sections was performed using only E14.5 and E17.5 datasets and the RCTD algorithm, as described above.

(P 34, L12–13)

• *

• Throughout the manuscript, the authors describe "lineage relationships" between cell populations, but these are in fact developmental trajectories inferred computationally (via UMAP connectivity and RNA velocity), not true lineage relationships. This distinction is critical and should be explicitly stated. *

• *

We agree with the reviewer that the present “lineage relationships” described in the original manuscript were primarily inferred from computational analyses, including UMAP connectivity and RNA velocity, rather than being directly demonstrated by lineage-tracing experiments. We also acknowledge that some RNA velocity results were not fully consistent with known in vivo developmental trajectories.

 We strengthened the biological validation of the inferred trajectories by incorporating extensive spatial verification of gene expression using immunohistochemistry and Xenium in situ hybridization analyses to confirm the identity and localization of each cell population. We also refined the computational analyses to better resolve regional differences in NCC dynamics and added comprehensive schematic illustrations based on established models of heart development from previous studies, with appropriate citations throughout the manuscript.

 In the revised manuscript, we have added much spatial verification of gene expression by immunohistochemistry and Xenium in situ hybridization data to confirm the identity of each cell population, and also include proper citation in the appropriate context. We also improved computational analysis to clarify the regional difference in NCC dynamics with comprehensive schematic illustration based on heart developmental processes established by previous literatures. In response to the reviewer’s comment and these revisions, we have carefully rephrased the relevant descriptions to clarify that the observed relationships represent computationally inferred developmental trajectories rather than definitive lineage relationships as follows.

(original) RNA velocity analysis in conjunction with developmental context, revealed lineage relationships among these groups (Figure 3c, d).

(revised) RNA velocity analysis in conjunction with developmental context, inferred global lineage relationships among these groups (Figure 3c, d), consistent with developmental trajectories in vivo.

(P8, L21–23)

(original) …, we present a comprehensive map of cardiopharyngeal NCC lineages …

(revised) …, we present a comprehensive map of cardiopharyngeal NCC populations …

(P18, L1)

(original) Overall, this study proposes a new framework for understanding cardiac NCCs heterogeneity based on the association with the cardiac cushion and the accompanying transition in Hox gene expression and regulatory programs. Our findings provide a basis for systematically dissecting the developmental diversity of cardiac NCCs.

(revised) Overall, this study proposes a new framework for understanding cardiac NCCs heterogeneity based on developmental route, Hox-code retention, and region-specific regulatory programs. Importantly, the developmental relationships and differentiation pathways described here are inferred from integrated computational analyses, including transcriptomic similarity, UMAP connectivity, and RNA velocity, rather than direct lineage-tracing experiments. Within this framework, our findings suggest distinct differentiation trajectories leading to great artery, ductus arteriosus, and coronary artery SMCs, as well as the aorticopulmonary septum and valvular/subvalvular mesenchyme.

(P20, L33 – P21, L7)

We also rephrased additional relevant sections throughout the manuscript in accordance with the reviewer’s comment (Please see below).

* For example, the statement "we propose that intracardiac NCCs within C10 and C22 differentiate via C2 into valvular (C21) and subvalvular (C18) interstitial cells" should be framed as a computational inference, not an established lineage relationship. Without clonal lineage tracing data, these claims cannot be verified. *

• *

We have changed the word “propose” to “infer” (P10, L24).*

Similarly, the claim that "NCCs contributing to the AP septum are distinct from other intracardiac NCCs in that they do not populate the cardiac cushions but remain continuous with NCCs populating the distal outflow tract cushion" lacks direct lineage evidence. What experimental data support this assertion? *

We agree with the reviewer that the original statement lacked direct lineage evidence, although the anatomical distinction between the outflow tract cushions and the AP septum—a protruding structure arising from the dorsal wall of the aortic sac—is well established. In the revised manuscript, we therefore removed this speculative statement and rewrote the section to more accurately describe the developmental process with appropriate references. The spatial and temporal features of AP septum formation are now also illustrated in Figure S9a (UMAP plots and schematic illustration) and described as follows:

*From E12.5 to E14.5, postotic cells showed a progressive emergence of the aorticopulmonary septum–associated cluster C16 from transitional states. * (P11, L32 – P12, L1)

In addition, we further validated the identity of the C16 cluster as AP septum–associated NCCs by incorporating additional marker analyses together with spatial verification using Xenium in situ analysis. In addition to Penk and Sfrp2, which were previously reported by Chen et al., C16 also highly expressed Postn, similar to other intracardiac clusters. Furthermore, C16 was characterized by relatively high expression of Tcf24 and low expression of Vegfc. Based on the reviewer’s comment and these additional experimental data, we revised the relevant Results section as follows:

(original) Within the intracardiac mesenchyme group, C16 exhibited high expression of Penk and Sfrp2 (Figure S3i and Table S6), corresponding to the cluster annotated as the aorticopulmonary septum in the previous study by Chen et al14. This annotation was further supported by enriched expression of Hox4 and Hox5 paralogs, consistent with its origin between PA4 and PA6 (Figure 3t, u). The aorticopulmonary septum originates as a protrusion from the dorsal wall of the aortic sac and is primarily derived from NCCs6,18–20. This septal structure fuses with the distal outflow tract cushions to divide the common arterial trunk into the aortic and pulmonary channels. Notably, NCCs contributing to this septum are distinct from other intracardiac NCCs in that it does not populate the cardiac cushions but remain continuous with NCCs populating the distal outflow tract cushion, suggesting that C16 represents this distinct NCC-derived population.

(revised) C16 was distinguished by high expression of Penk and Sfrp2 (Figure S3i and Table S6), corresponding to the cluster annotated as the aorticopulmonary septum in the previous study by Chen et al14. This cluster also exhibited robust expression of mesenchymal markers, including Postn, similar to other intracardiac clusters (Figure S3i). In addition, C16 showed relatively high expression of Tcf24 and low expression of Vegfc compared with the other intracardiac clusters (Figure S5a-e). These gene expression features of the aorticopulmonary septum were further validated by Xenium in situ hybridization (Figure S5f-j).

• Unlike other intracardiac NCCs that populate the distal outflow tract cushions, the aorticopulmonary septum originates as a protrusion from the dorsal wall of the aortic sac and is primarily derived from NCCs residing in PA4 and PA66,22–24. This septal structure subsequently fuses with the distal outflow tract cushions to partition the common arterial trunk into the aortic and pulmonary channels. Consistent with this developmental origin, C16 was enriched for the expression of Hox4 and Hox5 paralogs (Figure 3t, u), indicating that NCCs in this population retain their Hox code, in contrast to other intracardiac NCCs, in which most Hox genes were downregulated (see later details).*

(P9, L24 – P10, L6)

* In addition, the authors state that what De Bono et al. identified as "outflow smooth muscle" corresponds in their dataset to early intracardiac mesenchymal clusters C2, C10, and C22, present as early as E10.5, when mesenchymal NCC derivatives express immature SMC markers, and that these cells later differentiate into coronary artery SMCs around E14.5 (C23) as well as other non-muscle components. This claim is not verified, whether these cells are indeed the ones differentiating into coronary artery SMCs is based solely on computational inference from C2, C10, C22 to C23. *

• *

To validate that C23 corresponds to coronary artery SMCs, we performed additional immunostaining analyses for Myh11, Sost, and Reln, which distinguish great artery SMCs, ductus arteriosus SMCs, and coronary artery SMCs, respectively, consistent with their transcriptomic identities. Regarding differentiation of NCC-derived mesenchymal cells into coronary artery SMCs, we previously demonstrated using chick–quail chimera experiments and specific Cre-reporter mouse lines that proximal coronary artery SMCs originate from preotic NCCs rather than postotic NCCs (Arima Y et al. Nature Communications 3:1267, 2012). In addition, coronary artery SMCs have been reported to differentiate through pericyte intermediates (Volz KS et al., eLife 4:1–22, 2015). Consistent with these findings, our present lineage-tracing analyses using Sox9-CreERT2; Ai14 and Scx-CreERT2; Ai14 mice demonstrated that progenies of Sox9high and Scxhigh intracardiac cushion mesenchymal cells contributed to the coronary artery SMCs as well as the surrounding mesenchyme.

 We agree with the reviewer that the relationship between the early intracardiac mesenchymal clusters (C2, C10, and C22) and C23 is primarily inferred from computational trajectory analyses and is not demonstrated by direct clonal lineage tracing. Accordingly, we revised the Discussion to avoid overstatement and to clarify that these lineage relationships are inferred based on computational analyses together with prior experimental findings and the additional validation data described above. The revised text is as follows:

• *

(original) Our integrated map incorporates previously published lineage analyses of cardiac NCCs at early and late stages13,14, providing continuity through complementary single-cell and spatial transcriptomic data, although our interpretation of certain clusters differs from those of prior studies. For example, the cell population identified by De Bono et al. as outflow smooth muscle13corresponds in our dataset to early intracardiac mesenchymal clusters C2, C10, and C22, present as early as E10.5, when mesenchymal NCC derivatives express immature SMC markers. These cells later differentiate into coronary artery SMCs around E14.5 (C23) as well as other non-muscle components. Despite such differences in interpretation, the integrated map robustly captures lineage relationships, supported by accumulated developmental and anatomical evidence regarding cardiac outflow tract formation, particularly in relation to the outflow tract cushion.

(revised) Our integrated map incorporates previously published lineage analyses of cardiac NCCs at early and late stages13,14, providing continuity through complementary single-cell and spatial transcriptomic data. The present study further extends these datasets by resolving the heterogeneity of intracardiac mesenchymal populations and their lineage relationships. For example, the cell population identified by De Bono et al. as outflow smooth muscle13 corresponds in our dataset to early intracardiac mesenchymal clusters expressing immature SMC markers, which subsequently diverged into multiple derivatives including coronary artery SMCs. In addition, we identified distinct SMC populations corresponding to great artery SMCs, ductus arteriosus SMCs, and coronary artery SMCs, each characterized by unique molecular signatures such as Sost, Tfap2b/Ptger4, and Reln/Gja4, respectively. The continuity between intracardiac mesenchyme and coronary artery SMCs through a pericyte-like intermediate state is consistent with previous developmental studies showing that proximal coronary artery SMCs originate from preotic NCCs and may arise through pericyte intermediates8,21. Together, these findings provide a refined framework for understanding the diversification of cardiac NCC derivatives during outflow tract remodeling.

(P18, L14–29)

* Claims regarding marker expression in specific compartments (for example Hapln1 and Postn in cushions) require additional spatial validation at higher resolution than what is currently provided by the Visium data. Moreover, it is unclear whether these data are single-cell resolution; the authors need to clarify this. HCR staining would be ideal to confirm these expression patterns. Currently, all conclusions are based solely on gene expression without orthogonal spatial confirmation. At minimum, the authors should provide references from the literature supporting these expression patterns. *

As reviewer suggested, the confirmation of spatial context for the gene expression patterns of scRNA-seqs data is important to validate. We further investigated the spatial expression patterns through Xenium in situ hybridization system. Among cardiac mesenchyme subpopulation, Postn was dominantly expressed, however, Tcf24 was specifically expressed in AP septum (C16) not in cushion (C2, 10, 18, 21 and 22) in scRNA-seq data. On the other hand, Vegfc was expressed except in the AP septum. To confirm these opposing expression patterns, we newly added Figure S5, showing Tcf24 and Vegfc expression revealed by Xenium.

According to the reviewer’s suggestion, we added the sentence in the revised manuscript as follows:

C16 was distinguished by high expression of Penk and Sfrp2 (Figure S3i and Table S6), corresponding to the cluster annotated as the aorticopulmonary septum in the previous study by Chen et al14. This cluster also exhibited robust expression of mesenchymal markers, including Postn, similar to other intracardiac clusters (Figure S3i). In addition, C16 showed relatively high expression of Tcf24 and low expression of Vegfc compared with the other intracardiac clusters (Figure S5a-e). These gene expression features of the aorticopulmonary septum were further validated by Xenium in situ hybridization (Figure S5f-j).

(P9, L24–31)

* Could the authors confirm the absence of the Sox9high/Scxhigh population in AP septum descendants? *

Sox9high/Scxhigh NCCs are enriched not only in intracardiac NCC clusters C2, C10, and C14, but also in the AP septum-associated cluster C16, as stated in the manuscript as follows:

In the integrated UMAP, Sox9high/Scxhigh NCCs were enriched in C2, C10, C14 and C16.

(P17, L7–8)

However, based on our previous finding that proximal coronary artery SMCs originate from preotic rather than postotic NCCs (Arima Y et al. Nature Communications 3:1267, 2012), we infer that the intermediate population contributing to coronary artery SMCs is more likely derived from intracardiac NCC clusters C2 and C10 than from the pharyngeal arch 4/6-derived AP septum-associated cluster C16. To clarify this interpretation, we have added the following statement to the final paragraph of the Results section:

Together with our previous report that proximal coronary artery SMCs originate from preotic rather than postotic NCCs8, these results suggest that the intermediate population contributing to coronary artery SMCs likely represents a subset of Hox-downregulated intracardiac NCCs corresponding to clusters C2 and C10.

(P17, L29–33)

Minor Comments *Could the authors better justify their choice of stages (E11.5 to E17.5) for the single-cell multiomic assay? Given that OFT cushions are already populated by NCCs by E10.5 and that AP septum formation is already underway at this stage (see Development (2007) 134(8): 1593-1604), the rationale for beginning at E11.5 should be explicitly stated. *

We agree that NCCs have already populated the OFT cushions and that AP septum formation is underway by E10.5. Our selection of stages from E11.5 to E17.5 was intended primarily to enable synchronous comparisons between pharyngeal and intracardiac NCC populations across developmental stages, rather than to capture the earliest spatiotemporal events of cardiopharyngeal NCC lineage establishment. We have clarified this rationale in the revised manuscript by revising the statement as follows:

• *

(original) To elucidate the spatiotemporal dynamics of cardiopharyngeal NCC lineages, we performed single-cell multiome analysis on pharyngeal and cardiac tissues from E11.5 and E12.5 and ...

• *

(revised) To compare pharyngeal and intracardiac NCCs across developmental stages and characterize their temporal changes, we performed single-cell multiome analysis on pharyngeal and cardiac tissues from E11.5 and E12.5 and ...

(P6, L3-5)

• *

• *

*Spelling errors in Figure 2e: "ambious" should be "ambiguous"; "atrium venticle" should be "atrium/ventricle"; "ventricle" is misspelled in several locations. The clustering resolution is very high, yielding many clusters that are difficult to distinguish based on the colour code alone. What does "ventricle_CL" refer to? What is the "marginal layer"? A clearer legend or supplementary table defining each cluster would be helpful. *

To improve the clarity of the high-resolution clustering, we added Xenium cluster numbers to Figures 2c and 2d, enabling clusters to be distinguished not only by color but also by their anatomical locations. We also revised the cluster annotations in Figure 2e and corrected all spelling errors, including “ambiguous,”. In addition, we replaced “ventricle_CL” with “Cardiac muscle 1” and the formerly misspelled “atrium ventricle CM” with “Cardiac muscle 2.” The annotation “marginal layer” was also revised to “Spinal cord, marginal layer” for clarity.

*Figure 2j: The image is very dim. *

• *

We have improved the image quality and added regional annotations to enhance visibility in the revised figure.

* Figure 3b: What do the numbers correspond to? Additionally, "mesenchyme" is misspelled. *

We have revised Figure 3b to clarify different characteristics of mesenchymal subpopulations. We have also corrected the misspelling of “mesenchyme.”

**Referees cross-commenting**

*I also agree with the other reviewers' comments. Many thanks. *

Reviewer #2 (Significance (Required)):

*The overall approach is sound and the datasets generated are valuable resources for the field. The manuscript presents interesting findings regarding NCC heterogeneity in cardiac development.

As I understand the authors' model: OFT cushion-associated NCCs enter and associate with the OFT cushions, undergoing a Hox-positive to Hox-negative transition, accompanied by a shift in Meis binding and GRN rewiring. These cells pass through a Meis2-Sox9-Scx "skeletogenic progenitor-like" intermediate state and contribute to semilunar valves and coronary artery smooth muscle. Aorticopulmonary septum NCCs, by contrast, retain distinct Hox codes (Hox4/5 enrichment) rather than becoming Hox-negative, and contribute to the AP septum and great vessel smooth muscle.

Strengths: The integration of multiple omics modalities with lineage tracing is a powerful approach, and the identification of a Hox-dependent bifurcation in NCC fate is a novel conceptual advance.

Limitations: The reliance on computational trajectory inference without orthogonal lineage validation, combined with the integration of datasets spanning very different developmental stages, limits the strength of some conclusions. The analysis also required more precise anatomical annotations to facilitate accessibility to the readers - to visualise better the distinguishable contribution of the cardiac NCCs to the OFT.

Advance The study extends knowledge in the field by providing novel mechanistic insight into neural crest diversification in the context of cardiovascular development. The nature of the advance is primarily mechanistic, identifying a Hox-Meis regulatory switch that distinguishes cushion-associated from cushion-independent NCC lineages.

Audience This work will be of interest to a specialised audience interested in neural crest cells and developmental biologists using omics approaches to address cell fate diversification in complex tissues.

Reviewer Expertise Developmental biology, lineage analysis, mouse genetics. I do not have the expertise to assess the computational methods used in this paper. *

Again, thank you for giving us the opportunity to strengthen our manuscript with your valuable comments and queries. We have worked hard to incorporate your feedback and hope that these revisions persuade and satisfy you.

To Reviewer #3

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

Summary: *Iwase et al. presents a comprehensive multiomics analyses of cardiac neural crest cell (NCC) differentiation during cardiopharyngeal development. Using Wnt1Cre;R26R-EYFP mice, the authors isolated NCCs and non-NCCs at E11.5, E12.5, E14.5 and E17.5 stages and performed single-cell RNA-seq, ATAC-seq, spatial transcriptomics analyses. Spatial resolution of NCC-derived populations was achieved using Xenium (E11.5-E12.5) and Visium (E14.5-E17.5) platforms. Integration of single cell and spatial datasets identified distinct NCC-derived populations with defined spatial organization within the pharyngeal and intracardiac regions. The study concludes that Hox gene patterning underlies NCC subpopulation identity during cardiopharyngeal development and reveals a developmental transition from Hox-dependent to Hox-independent transcriptional regulation. Furthermore, the authors identify a Meis2-Sox9-Scx gene regulatory network making a skeletogenic progenitor-like intermediate that contribute to coronary smooth muscle and semilunar valve formation.

While the dataset is comprehensive and technically strong, several key conclusions are not always convincingly supported by enough data. As a result, some claims appear speculative and would benefit from additional experimental validation to strengthen the proposed developmental models.

I would strongly encourage authors to consider the following points to provide additional details that will strengthen their study: *

Thank you for your encouraging feedback and comments. We have responded to your comments below.

Major comments:

*- The authors should provide detailed FACS gating strategies and sorting conditions used to selectively isolate EYP-positive and EYP-negative NCC populations, including representative plots and information on exclusion criteria (e.g., doublets, dead cells). *

• *

We added the sorting gates to the revised Figure S1 and described the detailed FACS gating strategy and sorting conditions in the revised Methods section as follows:

• *

(original) EYFP-positive and -negative single cells were sorted using a FACSAria II or FACSMelody (BD Biosciences), freshly processed or cryopreserved.

• *

(revised) Single-cell suspensions were stained with 7-AAD (BD Pharmingen) for 3 min at 4℃, and EYFP-positive and -negative single cells were sorted using a FACSAria II or FACSMelody (BD Biosciences). The sorting strategy was as follows:

*Step 1. All events were gated by forward scatter (FSC) and side scatter (SSC) including area (A), height (H), and width (W) to obtain FSC singlets and remove doublets. *

*Step 2. FSC singlets were gated for the 7-AAD negative fraction to isolate viable cells. *

*Step 3. Viable cells were gated to isolate EYFP-positive NCCs or EYFP-negative non-NCCs. *

Sorted cells were freshly processed or cryopreserved for the following procedure.

(P28, L21–30)

* - Although the authors isolated nuclei for both scRNA-seq and ATAC-seq, the number of cardiomyocytes within the EYFP-negative population is unexpectedly low. The authors should clarify potential technical or biological reasons for this underrepresentation (e.g., nuclei isolation efficiency, sorting strategy, filtering criteria, or developmental stage-specific composition). *

• *

The unexpectedly low proportion of cardiomyocytes within the EYFP-negative population likely reflects the restricted sampling region used in this study. Specifically, we dissected cardiopharyngeal tissue and the outflow tract region rather than the whole heart as shown in Figure 1a, which likely introduced a sampling bias that reduced the representation of cardiomyocytes in the dataset.

Although the FACS gating strategy could potentially influence the recovery of specific cell types, we consider this possibility unlikely because cardiomyocyte populations were successfully detected in our dataset. In addition, during the quality-control process for scRNA-seq data, we applied a mitochondrial gene threshold of 25% to exclude low-quality cells. While mature cardiomyocytes typically exhibit high mitochondrial gene expression, embryonic cardiomyocytes at the analyzed developmental stages are immature and therefore were unlikely to be disproportionately excluded by this criterion. Consistent with this interpretation, distinct cardiomyocyte clusters remained detectable after filtering, indicating that cardiomyocytes were retained through the quality-control process.

Based on these considerations, we conclude that the low abundance of EYFP-negative cardiomyocytes primarily reflects the limited anatomical region sampled. We have clarified this point in the revised Results section as follows:

The overall proportion of cardiomyocytes was low, likely reflecting the restricted sampling of the cardiac outflow tract region (Figure 1a).

(P6, L23–25)

* - In figure 1, the authors present results from unsupervised clustering of 9,420 cells into 21 distinct clusters, many of which are broadly labeled as "mesenchymal cells". The authors should refine this nomenclature by providing more specific annotations or defining criteria, as this broad classification limits interpretability of the identified subpopulations. *

• *

In the revised manuscript, we further refined the clustering analysis by subdividing the original clusters C5 and C13 into two subclusters each, resulting in a total of 23 clusters in the UMAP shown in Figure 1d. This refinement improved the resolution and interpretability of the identified cell populations. In addition, we replaced the broad “mesenchymal cells” annotation with more specific classifications, including pharyngeal mesenchymal cells, intracardiac mesenchymal cells, and smooth muscle–like cells. To further clarify lineage relationships, we also provided UMAPs separately displaying NCC and non-NCC populations in Figure 1e, as suggested by the reviewer.

*- To integrate spatial annotated Xenium datasets with scRNA-seq data, the authors used Tangram, enabling estimation of the spatial distribution of EYFP-positive NCCs within the pharyngeal region of E11.5 and 12.5 embryos. However, the E11.5 section show a relatively low number of EYFP-positive cells (Figure 2f). The authors should clarify whether this reflects technical limitations (e.g., probe design, segmentation efficiency, or integration parameters) or biological factors and explain how this affects interpretation of the spatial analyses. *

*- The author used RCTD tool to decompose the scRNA-seq dataset into NCC and non-NCC components and mapped these onto the Xenium dataset. However, panels j and l in Figure 2 show low signal in the E11.5 sections. The authors should clarify whether this reflects technical limitations of the RCTD deconvolution, differences in sampling, or biological factors, and discuss how this result impacts of interpretation of the spatial mapping results at this stage. *

• *

In the original manuscript, estimated EYFP expression levels were visualized using a minimum cutoff of 0.1, with the remaining values mapped onto a 100-step color scale. However, this approach resulted in apparently weak signal intensity in the E11.5 sections because high-level noisy signals, including signals detected in the atrial lumen and outside the embryo, broadened the dynamic range of the visualization. In the revised manuscript, we applied an upper cutoff at the 90th percentile to reduce the influence of these noisy signals, resulting in improved visualization of EYFP-positive regions in Figure 2f, with signal intensity now comparable to that observed in Figure 2g. In addition, we further improved the spatial visualization of putative EYFP-positive cells by replacing centroid-based signal display with segmentation-based rendering in the revised Figure 2f, g. Importantly, these revisions affected only the visualization method and did not alter the underlying analyses or conclusions, as the integration and downstream analyses were performed using the original quantitative data.

 We also clarified the procedure used for NCC estimation and spatial mapping in the Methods section as follows:

Putative NCCs were identified through the following procedure:

Step 1. EYFP expression was estimated across 39 cell types in the Xenium dataset by integration with the single-cell multiome dataset (including both NCCs and non-NCCs) using Tangram.

Step 2. Xenium clusters enriched for EYFP expression, defined as clusters whose mean estimated EYFP level exceeded the threshold corresponding to the 65th percentile across all spots, and consistent with known neural crest derivatives were extracted as putative NCC populations.

Step 3. For each spot within these EYFP-enriched Xenium clusters, RCTD was used to estimate the corresponding multiome cluster identity.

(P33, L7–16).

*- The authors integrated their data with publicly available scRNA-seq datasets of NCCs from E8.5 to P7 hearts and present results from unsupervised clustering of 67,208 cells into 28 distinct clusters. Figures 3a and 3b show that cardiomyocyte (C26) is included in NCC-derivatives. The authors should clarify whether this reflects technical issue when they made FACS. *

• *

In our original datasets in Figure 1d, only a small number of NCCs were detected within the cardiomyocyte cluster (corresponding to C26 in Figure 3b), which was otherwise predominantly composed of non-NCCs. Cardiomyocytes assigned to C26 were also present in both publicly available scRNA-seq datasets included in the integrated analysis. Previous studies have reported that a limited subset of NCCs can differentiate into cardiomyocytes (Tomita, Y. et al., J. Cell Biol. 170:1135–1146, 2005; Tamura, Y. et al., Arterioscler. Thromb. Vasc. Biol. 31:582–589, 2011). Therefore, we consider that C26 likely represents a small population of NCC-derived cardiomyocytes rather than contamination caused by technical issues during FACS isolation, although the low cell number precluded further characterization.

 To clarify this point, we added the following statement to the Results section:

• *

*Only a few NCCs were detected within cardiomyocyte clusters, which were predominantly composed of non-NCCs, consistent with previous reports demonstrating NCC differentiation into cardiomyocytes15,16. *

(P6, L20–23)

* - The authors used RNA-velocity to infer relationship among the identified clusters. However, this analysis requires particular caution given that data were generated from multiple datasets obtained under different conditions. Several conclusions drawn from the RNA-velocity analysis are not convincing, as illustrated in Figures 3c and 3d, where the inferred velocity directions appear inconsistent with the proposed developmental model (e.g., trajectory from cluster 23 toward 18 or from 4 toward 6). The authors should clarify these discrepancies, justify the integration of heterogenous datasets and reassess the interpretation of the inferred lineage relationships. *

In the default setting, the integration workflow provided by Seurat which is widely used for scRNA-seq analysis employs canonical correlation analysis (CCA). CCA effectively corrects batch effects across datasets generated from different experimental platforms. However, it sometimes causes overcorrection to attempt to forcibly integrate different cell populations that are not shared among datasets (Andreatta, Bioinformatics, 2021). To minimize overcorrection for multiple datasets obtained under different experimental conditions, we applied reciprocal principal component analysis (RPCA) method recommended for comparative integration of heterogeneous scRNA-seq datasets (Luecken et al., Nature Methods, 2021). This selection is suitable for the integration of multiple datasets provided by different independent studies as in case of the present study.

 To infer relationship among the identified clusters, we then used RNA velocity analysis of scRNA-seq data fundamentally based on splicing dynamics. The original framework assumes that transcriptional induction and repression persist long enough for cells to reach active (transcribing) or inactive (silenced) steady-state equilibrium. However, this assumption is often violated during cell differentiation, where transient cell populations frequently exhibit rapidly changing mRNA levels that do not reach steady-state equilibrium. To address this limitation, the scVelo method was developed (Bergen et al., 2020, *Nature Biotechnology*), and we applied this approach to the integrated NCC datasets in the present study. This analysis successfully inferred directional flows from the pharyngeal mesenchyme toward SMCs and intracardiac mesenchyme through transitional states (Figure 3c,d). However, as the reviewer correctly pointed out, the analysis predicted a directional flow from C23 to C18, apparently opposite to the biological directionality supported by previous findings (see P9, L15–19).

 We consider this discrepancy to reflect intrinsic limitations of RNA velocity analysis. As discussed by Bergen et al., in systems containing multiple lineages and cellular processes, differences in gene regulatory networks among heterogeneous cell states can generate multiple trajectories in phase space owing to distinct splicing kinetics. In addition, incompletely captured splicing kinetics may represent only a limited portion of the overall dynamics, particularly near terminal differentiation states. In such cases, phase portraits of unspliced versus spliced transcripts may appear nearly linear rather than curved, potentially leading to erroneous assignment of positive or negative RNA velocities. Consistent with this limitation, we obtained opposite directionalities between C18 and C23 depending on whether steady-state or dynamical models were applied and according to different parameter settings. Through these repeated computational re-evaluations of lineage directionality, we concluded that RNA velocity analysis is suitable for capturing the global landscape of differentiation flow, but that accurate inference of local lineage directionality may require careful model selection and parameter optimization to ensure consistency with established biological evidence.

 To avoid arbitrariness and potential confusion, we removed the arrow between C18 and C23 from the revised Figure 3d. Instead, we now describe the observed continuity between these populations in the Results and Discussion sections as follows:

Notably, the UMAP revealed a continuum between C23 and C18 within the intracardiac mesenchyme population. Given previous findings that the proximal coronary artery SMCs originate from preotic NCCs8 and that pericytes give rise to coronary artery SMCs23, this connection likely represent a differentiation trajectory from intracardiac mesenchyme to coronary artery SMCs via a pericyte-like intermediate stage.

(P9, L15–19)

The continuity between intracardiac mesenchyme and coronary artery SMCs through a pericyte-like intermediate state is consistent with previous developmental studies showing that proximal coronary artery SMCs originate from preotic NCCs and may arise through pericyte intermediates8,21.

(P18, L24-27)

*- The authors should provide more detail on how they identified bifurcation points and more clearly explain the transition from intracardiac mesenchyme to smooth muscle cells (SMC). Additionally, they should clarify what distinguishes the three clusters (C4, C23, C27) in terms of transcription programs, marker expression, or functional states, to better support their proposed differentiation trajectories. *

To clarify the distinctions among the three SMC clusters (C4, C23, C27), we added a heatmap showing differentially expressed genes, violin plots for the mature SMC marker Myh11, and feature plots with immunostaining images for Myh11, Sost, and Reln expression (Figure S4). These additional analyses further validate the molecular and spatial characteristics of the three SMC clusters. Based on these data, we revised the relevant section of the Results as follows:

The SMC clusters, which were continuous with the pharyngeal mesenchyme via transitional populations in the UMAP, were identified by high expression of the mature SMC marker Myh11 (Figure 3i). Differential gene expression analysis further distinguished individual clusters (Figure S4a-g). Among these, C27 displayed a transcriptomic profile characteristic of the great artery SMCs, including high expression of Sost (Figure S3j). C4 was enriched for Tfap2b and Ptger4 (Figure S3j), markers of the ductus arteriosus SMCs21,22, supporting its annotation. C0 and C7 likely represent transitional states between pharyngeal mesenchyme and differentiated lineages, potentially bifurcating toward great artery SMCs or cardiac cushion mesenchyme (Figure S3a and Table S6). C23 was characterized by high expression of Gja4, a marker of coronary artery SMCs, along with pericyte markers Kcnj8 and Rgs5 (Figure 3j and Figure S3k), corresponding to the cluster similarly annotated by Chen et al14. In addition, C23 was also distinguished from C4 and C27 by its expression of Reln (Figure S4d).

• Immunostaining supported these cluster annotations. Sost expression is observed in great artery SMCs but not in ductus arteriosus and coronary artery SMCs, whereas Myh11 expression was higher in ductus arteriosus and coronary artery SMCs than in aortic SMCs (Figure S4f-m). Furthermore, Reln expression was restricted to coronary artery SMCs (Figure S4n-s).*

(P8, L30 – P9, L14)

To further clarify the identification of bifurcation points and the transition from intracardiac mesenchyme to SMCs, we additionally stratified the integrated UMAP according to distinct Hox expression profiles and inferred lineage trajectories corresponding to different neural crest and pharyngeal arch origins (Figure S9). Based on the inferred differentiation trajectory from C18 to C23 (P9, L15–19), together with the identification of C2 and C10 as Sox9high and Scxhigh intracardiac cushion mesenchymal populations contributing to coronary artery SMCs (P17, L27–33), we incorporated these lineage relationships into the schematic model presented in Figure 9.

Minor comments:

*- The authors convincingly demonstrate a switch in Meis-binding motifs across NCC populations, supporting a model in which cardiac cushion-associated NCCs transition from Hox-dependent to Hox-independent transcriptional regulation via alternative cofactor interactions and DNA-binding preferences. However, the authors should provide evident on whether GATA motifs are enriched within Meis peaks, as this could further clarify potential cooperative interactions during this transition. *

Although GATA-binding motifs were enriched within Meis-associated open chromatin regions in intracardiac NCCs compared with many other motifs, a substantial proportion of GATA motifs were located in peaks distinct from those containing Meis motifs. This observation raises the possibility that GATA and Meis transcription factors may cooperate through interactions across separate regulatory elements to modulate enhancer activity. However, we did not directly investigate this possibility in the present study. Instead, we found that the Meis-associated peaks identified in intracardiac NCCs, including the distal Sox9 enhancer containing a Meis2-binding site (chr11:112850240–112851186), more prominently contained several other consensus motifs, including predicted Hand2- and Nfatc1-binding sites. To our knowledge, however, these transcription factors have not previously been described as non-Hox cofactors of Meis proteins. Their potential recruitment to this enhancer, as well as possible cooperative interactions with Meis transcription factors during intracardiac NCC differentiation, was not examined in the current study and remains an important subject for future investigation.

* - In Figure 5 panels g, j and k are difficult to interpret. The authors should provide clearer annotations, labeling, or additional explanations to improve readability and facilitate understanding of the data. *

• *

We added the annotations to UMAP in Figure 6h (and 6i) corresponding to Figure 4a and included color bars in Figure 6k as well as in 6i. To further improve readability and facilitate understanding of the data, we added the explanation of the perturbation scores in the legends for Figure 6i and k.

(original)

*(g) Pseudotime trajectory analysis of integrated NCC clusters inferred using CellOracle. *

*(h-k) Sox9 (h, i) and Osr1 (j, k) knockout simulation presented as altered differentiation vector flows (h, j) and perturbation scores which was inner product of perturb simulation (i, k). *

• *

(revised)

*(g) Pseudotime developmental flow of integrated NCC clusters from the neural tube, inferred using CellOracle and projected onto the UMAP space shown in Figure 4a. *

*(h-k) Sox9 (h, i) and Osr1 (j, k) knockout simulations presented as altered differentiation vector flows (h, j) and perturbation scores, defined as the inner product between the simulated perturbation vectors and the original developmental flow (i, k). Green and magenta color bars indicate normal developmental flow and reverse flow induced by perturbation of the indicated genes, respectively. *

(P41, L3–L9)

*- In Figure 6, results support the role of hexameric Meis-binding motif-containing region as a distal enhancer of Sox9. The authors should provide additional results from a ChIP-qPCR experiment to further validate this model. *

• *

We attempted ChIP-seq experiments on O7-1 neural crest cell line using two different anti-Meis antibodies. However, we were unable to detect specific binding of Meis proteins to this enhancer region, although the luciferase assays clearly demonstrated the enhancer activity that was significantly attenuated by deletion of the Meis-binding motif. This discrepancy may reflect differences between endogenous chromatin and plasmid-based reporter contexts, including epigenetic modifications and chromatin accessibility. We are now investigating experimental conditions that would allow direct verification of endogenous Meis binding to this region.

* - Panel l in Figure S3 requires better annotation. *

• *

We added annotations including the aorta, pulmonary valve, left coronary artery, and its septal branch.

* - Correct the typo errors in Figure 5a. *

• *

The typographical errors “consercvation” and “Visuzalization” were corrected to “conservation” and “Visualization”, respectively.

* - The authors should refer to previous studies showing the role of Hoxa1 and Hoxb1 in the development of great arteries or semilunar valves. *

We have added the following discussion on the roles of anterior Hox genes in cardiovascular development, together with appropriate references:

• The regional identities of pharyngeal NCCs that contribute to cardiac development are established by Hox genes and their associated regulatory networks. Genetic studies have demonstrated essential roles for the anterior Hox genes in patterning the pharyngeal arch artery system and semilunar valve structures. Loss of Hoxa1 and Hoxb1 results in severe defects in pharyngeal arch artery development42, whereas ectopic or sustained expression of Hoxb1 in NCCs disrupts cardiovascular morphogenesis and causes malformations of the great arteries and semilunar valves43. Likewise, Hoxa3 contributes to proper patterning of the pharyngeal arch region and its NCC-derived derivatives44,45. These findings underscore the importance of precise spatial and temporal regulation of Hox genes during cardiovascular development. Among pharyngeal NCCs contributing to cardiac development, cushion-independent NCC derivatives (great artery SMCs and the aorticopulmonary septum) retain their origin-specific Hox-codes. In contrast, cushion-associated NCC derivatives (coronary artery SMCs and valvular/subvalvular interstitial cells) downregulate Hox expression and transition toward region-specific GRNs involving TFs such as Tbx20 and Gata4, whose expression is known to be induced by BMP signaling in cardiomyocytes46,47. Bmp2 and Bmp4 are expressed in the regions of the pericardial reflection traversed by NCCs en route to the cardiac cushion48. Together, these observations suggest that appropriate repression of Hox programs, coupled with activation of cardiac-specific regulatory networks, is required for normal differentiation of cushion-associated NCC derivatives.*

(P18, L30 – P19, L16)

• Roux, M. et al. Hoxa1 and Hoxb1 are required for pharyngeal arch artery development. Mech. Dev. 143, 1–8 (2017).*
• Zaffran, S., Odelin, G., Stefanovic, S., Lescroart, F. & Etchevers, H. C. Ectopic expression of Hoxb1 induces cardiac and craniofacial malformations. genesis 56, (2018).*
• Chisaka, O. & Capecchi, M. R. Regionally restricted developmental defects resulting from targeted disruption of the mouse homeobox gene hox-1.5. Nature 350, 473–479 (1991).*
• Kameda, Y., Watari-Goshima, N., Nishimaki, T. & Chisaka, O. Disruption of the Hoxa3 homeobox gene results in anomalies of the carotid artery system and the arterial baroreceptors. Cell Tissue Res. 311, 343–352 (2003).*

**Referees cross-commenting**

*Having read the comments of the other reviewers, I totally agree with them. All our comments converge and should allow the authors to improve their manuscript. *

Reviewer #3 (Significance (Required)):

*The study provides high-resolution spatial and temporal mapping of NCC-derived populations and proposes mechanistic insights into Hox-dependent versus Hox-independent transcriptional regulation, as well as a Meis2-Sox9-Scx gene regulatory network contributing to smooth muscle and semilunar valve formation.

Strengths and limitations: The datasets are rich and well-integrated, offering valuable resources for the field. However, several key conclusions rely on correlative analyses and heterogeneous datasets, making some claims speculative. Technical details, such as FACS gating, low representation of cardiomyocytes, and interpretation of RNA velocity, require further clarification, which currently limits the strength of the mechanistic inferences.

Advance: This work advances the understanding of NCC lineage diversification and gene regulatory dynamics in cardiopharyngeal development, particularly highlighting potential transcriptional switches and intermediate progenitor states that guide structural formation in the heart.

Audience: The study will be of interest to researchers in developmental biology, cardiovascular biology, and single-cell multi-omics, particularly those studying neural crest cell differentiation and cardiac morphogenesis.*

Again, thank you for giving us the opportunity to strengthen our manuscript with your valuable comments and queries. We have worked hard to incorporate your feedback and hope that these revisions persuade and satisfy you.

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:

Iwase et al. presents a comprehensive multiomics analyses of cardiac neural crest cell (NCC) differentiation during cardiopharyngeal development. Using Wnt1Cre;R26R-EYFP mice, the authors isolated NCCs and non-NCCs at E11.5, E12.5, E14.5 and E17.5 stages and performed single-cell RNA-seq, ATAC-seq, spatial transcriptomics analyses. Spatial resolution of NCC-derived populations was achieved using Xenium (E11.5-E12.5) and Visium (E14.5-E17.5) platforms. Integration of single cell and spatial datasets identified distinc NCC-derived populations with defined spatial organization within the pharyngeal and intracardiac regions. The study concludes that Hox gene patterning underlies NCC subpopulation identity during cardiopharyngeal development and reveals a developmental transition from Hox-dependent to Hox-independent transcriptional regulation. Furthermore, the authors identify a Meis2-Sox9-Scx gene regulatory network making a skeletogenic progenitor-like intermediate that contribute to coronary smooth muscle and semilunar valve formation.

While the dataset is comprehensive and technically strong, several key conclusions are not always convincingly supported by enough data. As a result, some claims appear speculative and would benefit from additional experimental validation to strengthen the proposed developmental models.

I would strongly encourage authors to consider the following points to provide additional details that will strengthen their study:

Major comments:

• The authors should provide detailed FACS gating strategies and sorting conditions used to selectively isolate EYP-positive and EYP-negative NCC populations, including representative plots and information on exclusion criteria (e.g., doublets, dead cells).
• Although the authors isolated nuclei for both scRNA-seq and ATAC-seq, the number of cardiomyocytes within the EYPF-negative population is unexpectedly low. The authors should clarify potential technical or biological reasons for this underrepresentation (e.g., nuclei isolation efficiency, sorting strategy, filtering criteria, or developmental stage-specific composition).
• In figure 1, the authors present results from unsupervised clustering of 9,420 cells into 21 distinct clusters, many of which are broadly labeled as "mesenchymal cells". The authors should refine this nomenclature by providing more specific annotations or defining criteria, as this broad classification limits interpretability of the identified subpopulations.
• To integrate spatial annotated Xenium datasets with scRNA-seq data, the authors used Tangram, enabling estimation of the spatial distribution of EYFP-positive NCCs within the pharyngeal region of E11.5 and 12.5 embryos. However, the E11.5 section show a relatively low number of EYFP-positive cells (Figure 2f). The authors should clarify whether this reflects technical limitations (e.g., probe design, segmentation efficiency, or integration parameters) or biological factors and explain how this affects interpretation of the spatial analyses.
• The author used RCTD tool to decompose the scRNA-seq dataset into NCC and non-NCC components and mapped these onto the Xenium dataset. However, panels j and l in Figure 2 show low signal in the E11.5 sections. The authors should clarify whether this reflects technical limitations of the RCTD deconvolution, differences in sampling, or biological factors, and discuss how this result impacts of interpretation of the spatial mapping results at this stage.
• The authors integrated their data with publicly available scRNAs-eq datasets of NCCs from E8.5 to P7 hearts and present results from unsupervised clustering of 67,208 cells into 28 distinct clusters. Figures 3a and 3b show that cardiomyocyte (C26) is included in NCC-derivatives. The authors should clarify whether this reflects technical issue when they made FACS.
• The authors used RNA-velocity to infer relationship among the identified clusters. However, this analysis requires particular caution given that data were generated from multiple datasets obtained under different conditions. Several conclusions drawn from the RNA-velocity analysis are not convincing, as illustrated in Figures 3c and 3d, where the inferred velocity directions appear inconsistent with the proposed developmental model (e.g., trajectory from cluster 23 toward 18 or from 4 toward 6). The authors should clarify these discrepancies, justify the integration of heterogenous datasets and reassess the interpretation of the inferred lineage relationships.
• The authors should provide more detail on how they identified bifurcation points and more clearly explain the transition from intracardiac mesenchyme to smooth muscle cells (SMC). Additionally, they should clarify what distinguishes the three clusters (C4, C23, C27) in terms of transcription programs, marker expression, or functional states, to better support their proposed differentiation trajectories.

Minor comments:

• The authors convincingly demonstrate a switch in Meis-binding motifs across NCC populations, supporting a model in which cardiac cushion-associated NCCs transition from Hox-dependent to Hox-independent transcriptional regulation via alternative cofactor interactions and DNA-binding preferences. However, the authors should provide evident on whether GATA motifs are enriched within Meis peaks, as this could further clarify potential cooperative interactions during this transition.
• In Figure 5 panels g, j and k are difficult to interpret. The authors should provide clearer annotations, labeling, or additional explanations to improve readability and facilitate understanding of the data.
• In Figure 6, results support the role of hexameric Meis-binding motif-containing region as a distal enhancer of Sox9. The authors should provide additional results from a ChiP-qPCR experiment to further validate this model.
• Panel l in Figure S3 requires better annotation.
• Correct the typo errors in Figure 5a.
• The authors should refer to previous studies showing the role of Hoxa1 and Hoxb1 in the development of great arteries or semilunar valves.

Referees cross-commenting

Having read the comments of the other reviewers, I totally agree with them. All our comments converge and should allow the authors to improve their manuscript.

Significance

The study provides high-resolution spatial and temporal mapping of NCC-derived populations and proposes mechanistic insights into Hox-dependent versus Hox-independent transcriptional regulation, as well as a Meis2-Sox9-Scx gene regulatory network contributing to smooth muscle and semilunar valve formation.

Strengths and limitations: The datasets are rich and well-integrated, offering valuable resources for the field. However, several key conclusions rely on correlative analyses and heterogeneous datasets, making some claims speculative. Technical details, such as FACS gating, low representation of cardiomyocytes, and interpretation of RNA velocity, require further clarification, which currently limits the strength of the mechanistic inferences.

Advance: This work advances the understanding of NCC lineage diversification and gene regulatory dynamics in cardiopharyngeal development, particularly highlighting potential transcriptional switches and intermediate progenitor states that guide structural formation in the heart.

Audience: The study will be of interest to researchers in developmental biology, cardiovascular biology, and single-cell multi-omics, particularly those studying neural crest cell differentiation and cardiac morphogenesis.

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, Iwase et al. cleverly make use of different modalities, spatial transcriptomics and single-cell omics datasets, in conjunction with a well-established Wnt1-Cre;R26R-EYFP line to trace neural crest cells (NCCs) contributing to the cardiovascular system during embryonic development in the mouse. By doing so, the authors identified a bifurcation between cardiac NCCs contributing to the OFT cushions and forming the aorticopulmonary (AP) septation complex. Thus, the authors split "intracardiac NCCs" into two different NCC programs/compartments, even though both reside in the broad OFT region. The NCCs that enter and associate with the OFT cushions undergo a Hox off transition (Hox-positive to Hox-negative once intracardiac), with a corresponding shift in Meis binding and GRN wiring. The authors propose these cells pass through a Meis2-Sox9-Scx "skeletogenic progenitor-like" intermediate and contribute to semilunar valves and coronary artery smooth muscle. By contrast, the NCCs assigned to aorticopulmonary septum (APS) formation and great vessel smooth muscle retain a distinct Hox codes.

Major Comments

1. The manuscript would benefit from clearer delineation between the different NCC contributions, particularly for non-specialist readers. The distinction between (i) CNCCs in OFT cushions and (ii) CNCCs forming the aorticopulmonary (AP) septation complex is not adequately explained. While both populations contribute to OFT septation, according to the authors, they represent distinct compartments with different developmental trajectories. The authors could clarify this using anatomically labelled hearts at the stages they conduct their analysis, along with additional text and schematics explaining what is meant by each compartment. This would greatly enhance the accessibility of the manuscript.
2. The Methods state that "Decomposition of cell clusters of scRNA-seq was performed by RCTD to map them onto the Xenium dataset," but this description is insufficient. The authors should clarify whether RCTD was applied separately for each developmental stage (i.e., E11.5 scRNA-seq reference for E11.5 Xenium, E12.5 reference for E12.5 Xenium), or whether a pooled reference was used across stages. This clarification is important because RCTD performance depends critically on the correspondence between reference and target datasets. Using a multi-stage integrated reference to deconvolve stage-specific spatial data could introduce artifacts, as cell states and cluster compositions vary considerably across developmental timepoints.
3. Figure 2j-m needs annotations and shcematics. It is currently very difficult to identify the different compartments. See Figure C in Chen et al. for an example of this approach. In addition, what are the fine clusters from 1 to 20? Which ones are NCC-derived?
4. The panels show pharyngeal markers, OFT/intracardiac markers, and SMC markers in Figures 2n-y, but could the authors show the proportion of NCC-derived (YFP+) cells for each cluster? Could the authors also map only the YFP+ cells on the Xenium data? It would be useful to see the proportion of YFP-positive (NCC-derived) cells for each delineated compartment. YFP-positive cells appear to exist at the boundary between LV and RV in the septum, this observation would benefit from proper quantification.
5. Figure 3 is confusing because it integrates data from multiple overlapping stages (E8.5 to P7). While the authors identify distinct compartments, pharyngeal mesenchyme, intracardiac mesenchyme, and SMCs, it is unclear why stages beyond E14.5 and E17.5 (corresponding to the initial single-cell omics and Visium analysis) were included. Although leveraging additional datasets is a clever approach, the integration of data from such disparate developmental timepoints confounds interpretation. For example, Cluster 6 appears to include cells from both P7 and E12.5 stages. Given that the Visium data represent hearts at E14.5 and E17.5, it is problematic to map clusters derived from other stages onto these spatial datasets.
6. Throughout the manuscript, the authors describe "lineage relationships" between cell populations, but these are in fact developmental trajectories inferred computationally (via UMAP connectivity and RNA velocity), not true lineage relationships. This distinction is critical and should be explicitly stated.

For example, the statement "we propose that intracardiac NCCs within C10 and C22 differentiate via C2 into valvular (C21) and subvalvular (C18) interstitial cells" should be framed as a computational inference, not an established lineage relationship. Without clonal lineage tracing data, these claims cannot be verified.

Similarly, the claim that "NCCs contributing to the AP septum are distinct from other intracardiac NCCs in that they do not populate the cardiac cushions but remain continuous with NCCs populating the distal outflow tract cushion" lacks direct lineage evidence. What experimental data support this assertion?

In addition, the authors state that what De Bono et al. identified as "outflow smooth muscle" corresponds in their dataset to early intracardiac mesenchymal clusters C2, C10, and C22, present as early as E10.5, when mesenchymal NCC derivatives express immature SMC markers, and that these cells later differentiate into coronary artery SMCs around E14.5 (C23) as well as other non-muscle components. This claim is not verified, whether these cells are indeed the ones differentiating into coronary artery SMCs is based solely on computational inference from C2, C10, C22 to C23. 7. Claims regarding marker expression in specific compartments (for exmaple Hapln1 and Postn in cushions) require additional spatial validation at higher resolution than what is currently provided by the Visium data. Moreover, it is unclear whether these data are single-cell resolution; the authors need to clarify this. HCR staining would be ideal to confirm these expression patterns. Currently, all conclusions are based solely on gene expression without orthogonal spatial confirmation. At minimum, the authors should provide references from the literature supporting these expression patterns. 8. Could the authors confirm the absence of the Sox9high/Scxhigh population in AP septum descendants?

Minor Comments

Could the authors better justify their choice of stages (E11.5 to E17.5) for the single-cell multiomic assay? Given that OFT cushions are already populated by NCCs by E10.5 and that AP septum formation is already underway at this stage (see Development (2007) 134(8): 1593-1604), the rationale for beginning at E11.5 should be explicitly stated. Spelling errors in Figure 2e: "ambious" should be "ambiguous"; "atrium venticle" should be "atrium/ventricle"; "ventricle" is misspelled in several locations. The clustering resolution is very high, yielding many clusters that are difficult to distinguish based on the colour code alone. What does "ventricle_CL" refer to? What is the "marginal layer"? A clearer legend or supplementary table defining each cluster would be helpful. Figure 2j: The image is very dim. Figure 3b: What do the numbers correspond to? Additionally, "mesenchyme" is misspelled.

Referees cross-commenting

I also agree with the other reviewers' comments. Many thanks.

Significance

The overall approach is sound and the datasets generated are valuable resources for the field. The manuscript presents interesting findings regarding NCC heterogeneity in cardiac development.

As I understand the authors' model: OFT cushion-associated NCCs enter and associate with the OFT cushions, undergoing a Hox-positive to Hox-negative transition, accompanied by a shift in Meis binding and GRN rewiring. These cells pass through a Meis2-Sox9-Scx "skeletogenic progenitor-like" intermediate state and contribute to semilunar valves and coronary artery smooth muscle. Aorticopulmonary septum NCCs, by contrast, retain distinct Hox codes (Hox4/5 enrichment) rather than becoming Hox-negative, and contribute to the AP septum and great vessel smooth muscle.

Strengths: The integration of multiple omics modalities with lineage tracing is a powerful approach, and the identification of a Hox-dependent bifurcation in NCC fate is a novel conceptual advance.

Limitations: The reliance on computational trajectory inference without orthogonal lineage validation, combined with the integration of datasets spanning very different developmental stages, limits the strength of some conclusions. The analysis also required more precise anatomical annotations to facilitate accessibility to the readers - to visualise better the distinguishable contribution of the cardiac NCCs to the OFT.

Advance

The study extends knowledge in the field by providing novel mechanistic insight into neural crest diversification in the context of cardiovascular development. The nature of the advance is primarily mechanistic, identifying a Hox-Meis regulatory switch that distinguishes cushion-associated from cushion-independent NCC lineages.

Audience

This work will be of interest to a specialised audience interested in neural crest cells and developmental biologists using omics approaches to address cell fate diversification in complex tissues.

Reviewer Expertise

Developmental biology, lineage analysis, mouse genetics. I do not have the expertise to assess the computational methods used in this paper.

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

Iwase et al have used multiomics and spatial transcriptomics to comprehensively map neural crest cell contributions to the mouse heart and great arteries. This careful and detailed analysis reveals changes in the transcriptional profile of neural crest cells as they give rise to different regions and cell types in the heart and great vessels. The study significantly builds on a number of recent scRNA-seq analyses of neural crest cell development and includes development of a new informatic tool for regulatory network investigation. Among the new findings documented are downregulation of Hox gene expression in intracardiac crest cells and regulation of Sox9 by Meis transcription factors. Addressing the following points would improve clarity and accessibility.

1. In Figure 1C it is difficult to visualize all the colors given the mixed contribution of NCC and nonNCC cells to mesenchyme. Please also show YFP transcript distribution in NCC versus nonNCC plots. In addition, it would be helpful to show plots for both NCC and nonNCC for Gata4 and Tbx20.
2. The authors identify a cardiomyocyte cell cluster in their integrated NCC scRNA-seq plots. Are these cells labelled by Wnt1-Cre in the authors' own dataset? Is the trajectory analysis informative as to the steps preceding acquisition of cardiomyocyte fate?
3. Linked with this point, is it possible that there are nonNCC cells in the integrated plots? Of note, many of the NCC genes overlap with genes that have also been shown to be expressed in mesodermal cardiac progenitors (including Osr1, Pparg, Dlk1, Tcf21, Ebf2, Tbx20, Sox9). For example, is it possible to distinguish NCC derived smooth muscle within the heart from cells originating from the second heart field that may express smooth muscle genes? Cluster 27 for example appears broadly expressed in the region of ventricular outlets in Figure 3. Comparison with YFP transcript distribution may be helpful here.
4. Can the authors add any validation of key expression patterns, for example using fluorescent in situ hybridization?
5. Please elaborate on the decoded Hox code patterns that appear to be indicative of arch origins. Do the results allow determination of whether the trajectories to different cardiac fates inferred in Figure 3D differ in different arches?
6. The authors need to explain why the authors place an arrow from mesenchymal cluster 18 to 23 in Figure 3D while the trajectory analysis in 3C predicts the opposite direction.
7. The authors nicely show downregulation of Hox gene expression in NCC cells entering the heart. Can they add discussion of any insights into this from prior studies of loss or gain of Hox gene function?
8. Figure 3Y could be simplified to more clearly distinguish the two types of Meis binding sites. For example, it may be helpful to reorder the mesenchymal cell types based on Hox expression.
9. The authors provide nice in vitro and in vivo evidence for an upstream role of Meis transcription factors in regulating Sox9 expression. Can the authors identify from the enhancer sequence (or their transcriptomic dataset) any of the non-Hox transcription factors that Meis may be working with here? Please discuss the significance of Sox9 expression in epicardium driven by the same enhancer. Might this regulation also operate in second heart field progenitor cells where both genes are expressed? It is not evident in Figure 7 that Sox9-EGFP is also expressed in epicardium.
10. Could this approach yield similar data for Osr1? Please clarify if there is any experimental evidence supporting the predicted negative regulation of Sox9 by Osr1 in the heart illustrated in Figure 8.
11. Concerning the links between valve mesenchyme and skeletogenic programs it would be relevant to cite the earlier work of Lincoln and Yutzey (reviewed in PMID: 16643886):
12. In order to increase accesibility of the dataset the authors are encouraged to include a browser link.

Minor points:

1. The authors could rephrase the title since the term topographical genetic switch is unclear.
2. In the introduction, with reference to the De Bono study, please note that Tbx1 was shown to regulate pharyngeal NCC differentiation stage transitions non-cell autonomously.

Significance

Iwase et al have used multiomics and spatial transcriptomics to comprehensively map neural crest cell contributions to the mouse heart and great arteries. This careful and detailed analysis reveals changes in the transcriptional profile of neural crest cells as they give rise to different regions and cell types in the heart and great vessels. The study significantly builds on a number of recent scRNA-seq analyses of neural crest cell development and includes development of a new informatic tool for regulatory network investigation. Among the new findings documented are downregulation of Hox gene expression in intracardiac crest cells and regulation of Sox9 by Meis transcription factors.

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 all reviewers for the valuable feedback and critical insight on our study. We acknowledge the concern that the manuscript, in its initial form, appeared descriptive and did not provide the mechanistic insight inferred from the current data. In the revised manuscript, we will (i) more clearly delineate what mechanistic inferences can be drawn from the existing data, (ii) expand our discussion of the caspase-independent mechanisms, and (iii) incorporate additional experiments/analyses aimed at identifying downstream effectors that mediate the observed phenotypes. In this revision plan, we have included six new figures addressing some of the major issues raised by reviewers.

1. Specifically, to address questions about mechanistic insight, we generated stable ACSL1:HaloTag expressing hESCs. Currently presented as Figure 1A for reviewers____. __ACSL1 is a critical enzyme that catalyzes the first step of fatty acid oxidation at the outer mitochondrial membrane. Our previous analysis and work from the Opferman lab demonstrated that ACSL1 contains a BH3-like domain. Thus, we examined the effects of MCL-1 inhibition on the mitochondrial localization of this enzyme. Our findings pinpoint that MCL-1 inhibition is causing the displacement of ACSL1 from the mitochondria (__Figures 1B-C for reviewers). Our interpretations of the effects of MCL-1 inhibition are 2-fold: 1) as we show in our data, MCL-1 inhibition causes disruption of the mitochondrial cristae, altering the microenvironment for fatty acid oxidation, and 2) as seen in cancer cells, the MCL-1 inhibitor may also displace ACSL1 from the mitochondria. In the new version of the manuscript, we will focus on these 2 mechanisms as mechanistic outcomes of MCL-1 inhibition.
2. We have included data of cells treated with Perhexilin (CPT1/2 inhibitor), and Etomoxir (CPT1a inhibitor) (Figure 2 for reviewers). This experiment determines whether direct perturbation the FAO pathway mimics the effects of the MCL-1i.
3. We have assayed the effects of MCL-1 inhibition on oxygen consumption rates in NPCs. Currently presented as Figure 3 for reviewers.
4. We will perform MCL-1:MICOS proximity ligation assays and/or immunoprecipitation assays to determine whether MCL-1 inhibitors disrupt the association of MCL-1 with MICOS. Preliminary data suggesting an association (albeit, very weak) are shown in Figure 4 for reviewers. __Reviewer #1____ (Evidence, reproducibility and clarity (Required)): __

Summary: This study claims that beyond its canonical anti-apoptotic function, MCL-1 has essential non-apoptotic roles in human neurodevelopment. Pharmacologic inhibition of MCL-1 in human neural stem cells disrupts mitochondrial inner membrane architecture by destabilizing cristae and the OPA1-MICOS complex, leading to swollen mitochondria with disorganized cristae. These structural defects impair fatty acid oxidation and lipid droplet homeostasis, linking cristae integrity to metabolic competence. Independently of apoptosis or proliferation, MCL-1 inhibition selectively depletes intermediate neural progenitors, indicating a direct role in lineage progression. Overall, the work positions MCL-1 as a key regulator of mitochondrial structure-metabolism coupling that instructs neural progenitor identity and human neurogenesis.

Overall: The study does a good job of using (in most assays) caspase inhibition (e.g., QVD treatment) to block apoptotic responses induced by MCL-1 inhibition. As a result, many of the phenotypes caused by inhibition are likely to be independent of caspase activation. As a result, this manuscript would be of interest to researchers that study the topics of the BCL-2 family and cell death signaling, mitochondrial bioenergetics and dynamics, neurodevelopment, and cellular metabolism. However, as currently presented the manuscript is only descriptive and lacks mechanistic insight.

We thank Reviewer 1 for the insightful evaluation of our work. We are encouraged that the reviewer finds the study relevant to investigators in the fields of BCL-2 family biology, mitochondrial dynamics and bioenergetics, neurodevelopment, and cellular metabolism. We also thank the reviewer for pointing out the need to increase the mechanistic insight of our findings. As mentioned above, in the revised manuscript, we are proposing to address this.

Major Concerns:

1) The authors only use a single MCL-1 inhibitor and never use other non-targeting BH3-mimetics (such as venetoclax) as negative controls. This seems like a missed opportunity to demonstrate that the phenotypes observed are MCL-1 dependent.

This is an excellent point. We will include venetoclax (ABT-199) to examine their effect on intermediate progenitors (TBR2 +) and early born neurons (BIII tubulin +).

2) There is no mechanism proposed in this study other than reliance upon QVD as not affecting the phenotypes. As submitted, the manuscript only can speculate that these phenotypes are due to non-apoptotic roles of MCL-1 inhibition. The authors have missed an opportunity to explore MCL-1's non-apoptotic functions directly.

Mechanistically, we propose MCL-1 is acting in 2 ways: 1) as we show in our data, MCL-1 inhibition causes disruption of the mitochondrial cristae, altering the microenvironment for fatty acid oxidation, and 2) as seen in cancer cells, MCL-1 inhibitors may also displace ACSL1 from the mitochondria.

In the past few weeks, since receiving the initial reviews, we have focused on testing the 2nd possibility, since the accumulation of lipids was also seen in cancer cells (see PMID: 38503284). We have successfully generated stable ACSL1:HaloTag expressing hESCs (Figure 1A for reviewers). Our findings included here, ACSL1 is displaced from the mitochondria by MCL-1 inhibition in NPCs (Figures 1B-C for reviewers).

Other concerns exist that weaken the impact of the study.

1. Figure 1 should include the fact that QVD inhibition (shown in Sup Fig 2) does not obviate the phenotype induced by pharmacological inhibition of MCL-1 on mitochondrial morphology. We would like to clarify that QVD does prevent the phenotypes induced by MCL-1 inhibition on mitochondrial morphology. In Fig1B, we report an increase in volume and surface area at 24h and 48h along with a decrease in mitochondrial content at 48h when NPCs were treated with MCL-1i only. However, NPCs co-treated with QVD in Supp Fig 2B did not exhibit any significant morphological phenotypes on average or at min/max values. Reviewer 1 may be referring to Fig 1B's corresponding min/max values presented in Supp Fig 2A where we reported an increase in __max __volume.

Figure #

Volume

Surface Area

Fig 1B (MCL-1i only, avg values)

Increase (avg vol)

increase (avg)

Supp Fig 2B (MCL-1i+QVD)

no change

no change

Supp Fig 2A (MCL-1i only, max/min values)

increase (max vol)

no change (max)

For clarity, we will move Supplementary Fig 2A into Supplementary Fig 1.

Figure 2 would benefit from evidence that caspase inhibition does not repress the phenotype on mitochondrial cristae morphology (volume and area). Furthermore, the FIB-SEM data are very hard to appreciate as the size precludes visualization of individual mitochondria.

While we included the visualization of the segmented mitochondria and cristae (Figure 2C), as well as snapshots through the z-stack for segmented cristae only (Figure 2E) and segmented mitochondria separately (Supp Figure 3A) in the original manuscript, we are also now attaching the FIB-SEM 3D reconstruction videos (New Supplementary Videos 1-2 for reviewers) (1. Mito and cristae, 2. Cristae only, 3. Mito only) for ease of visualization purposes.

Figure 3 reports that MIC60 and OPA1 appear to be downregulated in response to MCL-1 inhibition, but these appear to be more significant only when QVD is added. Why would the phenotype be obscured in the non-QVD setting (Fig. 2B&C). How does MCL-1 inhibition lead to changes in MIC60/MICOS/OPA1? This seems quite preliminary at this point.

In Figures 3B and 3C, we report decreased protein levels of short-form OPA1 and MIC10 only, not MIC60. We argue that our data with QVD shows that the cell death function of MCL-1 (i.e., inhibiting cell death effectors from initiating the caspase cascade) is not the main trigger of the phenotypes we report (cristae dysregulation and fatty acid oxidation disruption), however, cells without a functional cristae and/or defects in FAO, may not be able to survive long-term. Thus, QVD treatment preserves these cells that may not survive the dismantling of such an essential structure. To confirm this, we have performed immunofluorescence of cleaved caspase 3 (Figure 5 for reviewers). These results show that indeed MCL-1 inhibition at the time points of our study doesn't result in increased activation of Caspase-3. We reported similar results of MCL-1 inhibition in oligodendrocyte precursor cells (Gil and Hanna et al., Glia, 2025, PMID: 41420072)

The loss of MIC60 and OPA1 should repress electron transport chain function, are such impacts observed in the cultured cells? This could be shown by assessing oxygen consumption, etc. Such data would enhance the authors' conclusion that MCL-1 inhibition leads to defects in mitochondrial physiology*. *

We completely agree with this comment by Reviewer 1. In our revision, we will include an assessment of mitochondrial oxygen consumption rate, using the Seahorse analyzer (mitochondrial stress test), of NPCs treated with MCL-1i. Preliminary data (n=3) are currently presented as Figure 3 for reviewers. Interestingly, these data show a more nuanced cellular response. Consistent with our conclusion that MCL-1 inhibition does not cause apoptotic cell death, MCL-1i did not affect mitochondrial respiration at baseline. The specific deficits appear in spare respiratory capacity and maximal respiration, meaning cells can sustain routine mitochondrial function but lose the ability to respond to increased energetic demand. This suggests MCL-1 loss creates a mitochondrial reserve deficiency rather than a generalized bioenergetic failure. The results with caspase inhibitors show a near-zero OCR across both 24h and 48h timepoints, and significant reductions in maximal respiration, spare respiratory capacity, and non-mitochondrial OCR. Remarkably, these conditions are not detrimental to newborn neurons, as shown in Figure 7. This is very interesting because it suggests that, under severe bioenergetic failure, neural stem cells (PAX6+) can differentiate into newborn neurons in a TBR2-independent manner. More relevant to this study, our results unequivocally demonstrate that TBR2-positive cells depend on the non-apoptotic function of MCL-1

In Figure 4, the differences between transcripts (qPCR data) and protein (immunoblot) data are often confusing and not well explained. Why do the authors propose that mRNA expression is decreasing whereas the protein expression is increasing? Example CPT1. Furthermore, it is unclear what these data mean functionally? Is this reflective of enhanced lipid oxidation or simply a response to inhibition of fatty acid oxidation? Clarification of the impact of these findings is necessary.

We agree with Reviewer 1 that the results could be hard to interpret. However, the effects of MCL-1 inhibitors on the transcription of fatty acid oxidation genes have been widely cited by the work of Opferman and Walensky (PMID: 36198266). We speculate that the effects on transcription are triggered by mitochondrial signaling. The mechanistic insight into this phenomenon would be an interesting next step.

In the case of CPT1, we addressed this comment and found that the difference is due to differential expression of isoforms The RT-qPCR shown in Figure 4, is on CPT1c, whereas the western blot is on CPT1a. Unfortunately, after trying several products, we determined that there are no good antibodies for CPT1c. Thus, since we can't compare gene and protein expression, we will include CPT1a RT-qPCR data to complement the western blot.

The increase in lipid droplet number induced by MCL-1 inhibition has been previously documented, but it is unclear whether this increase is related to an inability to oxidize lipid (defective fatty acid oxidation) that leads to increases in the cellular abundance or whether this indicates that MCL-1 inhibition leads to enhanced storage. Do other inhibitors of fatty acid oxidation lead to similar increases in lipid droplet size and abundance? Does QVD inhibition affect this phenotype?

This is a great point raised by Reviewer 1, and one we have also wondered about. We conducted an experiment using C16 BODIPY to address this point (Figure 6 for Reviewers). We observed no changes in C16 lipid droplet accumulation in count, volume, or surface area when cells were treated with MCL-1 inhibitor for 24 hours total with or without a starvation period in the last 6 hours of treatment. However, we observed significant pan-lipid droplet accumulation in the same conditions. This contrast suggests that FAO of exogenous LC-fatty acids is not reliant on MCL-1. This finding does not discount from the requirement of MCL-1 for other FAO processes especially given the major limitation of how much C16 BODIPY (fluorescent palmitate) can be administered to the cells (10µM) which was 10-fold less than what we exogenously supplied to the cells for the pan-BODIPY experiment (100µM, see Figure 5). It is entirely possible that this small dose was not enough to detect any lipid droplet accumulation.

We have now also included experiments using etomoxir and perhexiline to assess their effects on TBR2/PAX6 (Figure 2 for reviewers). The results indicate that inhibiting the FAO pathway does not fully mimic the effects of MCL-1i on TBR2. However, we show that MCL-1i displaces ACSL1 from the mitochondria, a step that is upstream of CPT1/2. We suggest a model in which the coordinated non-apoptotic function of MCL-1 at the outer mitochondrial membrane promotes ACSL1 activity and, in the inner mitochondrial membrane, regulates mitochondrial cristae morphology. While our data point to this model, we are limited by the tools to investigate it further, but it will be a great direction for future experiments.

For Figure 6, while these data may be very meaningful, as presented they are very hard to appreciate. Insets that show the neuronal populations would help to convey the point that the differentiation is impacted. Also, are there other methods that could confirm these observations (qPCR to show changes in differentiation).

We agree with Reviewer 1. In the new version of the manuscript, we will include panels that zoom into the cell populations we quantified. The current panels will go to a new Supplemental figure. We will also add the TUBB3 to the qPCR panel in the new version.

Figure 7 is also very hard to appreciate. What is the reader to see? Can these be quantified? It seems that QVD may be rescuing in this figure, does this suggest that MCL-1 inhibition might be inducing death. All of this needs to be quantified.

We will provide quantification of BIII tubulin branching, and it will be included next to the images provided.

BCL-XL has also been implicated in affecting mitochondrial electron transport chain function (See PMID: 19255249, 21926988, 21987637). Can BCL-XL inhibitors affect any of the phenotypes associated here?

We will include experiments to test the effect of BCL-2 and BCL-XL inhibitors on TBR2 cells to address this comment.

Please be carefully avoid using the term "MCL-1 loss", when talking about pharmacological inhibition. Only genetic ablation (e.g. knockout, silencing, etc.) should be termed loss.

We have now removed the reference to MCL-1 loss in line 199.

__*Reviewer #1 (Significance (Required)):

The study advances in human cells the impacts of MCL-1 inhibition. They replicate many impacts previously observed in mouse systems and refine analyses to impacts on MICOS complex, lipid droplet storage, and neuronal differentiation. While these findings are important and would be well received by a wide audience, the study fails to provide almost any mechanistic insight into how these phenotypes are being induced. The only common theme is that blocking caspase activation in many assays fails to block the phenotype.

*__

__Reviewer #2_ (Evidence, reproducibility and clarity (Required)): _*

Summary: This manuscript by Hanna et al. investigates non-apoptotic roles of MCL-1 in human neural stem cells and connects MCL-1 inhibition to mitochondrial cristae formation and beta-oxidation. Connecting these roles to brain development, the authors also show a reduction in the number of progenitor cells upon MCL-1 inhibition, independently of caspase activity. Throughout their work, the authors make use of an impressive array of imaging techniques. While the methods used offer sufficient evidence to connect MCL-1 inhibition to cristae architecture, the mechanistic underpinnings of this effect remain unexplored. *__

We thank Reviewer 2 for the thoughtful and positive assessment of our manuscript. We appreciate the reviewer's recognition that our study reveals non-apoptotic roles of MCL-1 in human neural stem cells. We are also grateful for the acknowledgment of the imaging approaches employed, which allowed us to connect MCL-1 function to cristae architecture with multiple complementary techniques. We acknowledge the reviewer's point that the mechanistic basis by which MCL-1 influences cristae structure remains insufficiently defined. In the revised manuscript, we will clarify the limitations of the current data, expand our discussion of potential mechanisms, and incorporate additional analyses to identify downstream effectors that mediate these structural and metabolic changes.

Major comments:

- In Fig. 1B, the very same representative images are shown for both conditions (DMSO and S63845) at 48 hours.

We deeply appreciate Reviewer 2 for catching this unintentional duplication that occurred during figure preparation. We have now corrected this issue.

- For Western Blot analysis, it looks like the authors only quantified the band density of their proteins of interest without considering varying levels of control protein (Actin) levels. Normalizing the protein levels to actin would account for any differences in loaded protein amounts (although a Ponceau staining might be preferable still to exclude this). This is especially relevant for Fig. 4E, where actin levels visibly differ between the conditions.

All WB quantifications were normalized to Actin (this detail is now added to the y-axis of all band density graphs and figure legends). In addition, we will transform the data to a logarithmic scale to "normalize" for gel-to-gel variability.

- The authors offer evidence that MCL-1 inhibition impedes proteolytic cleavage of OPA1-L into the OPA-1-S isoforms, yet do not explore the mechanism behind this. Since OPA1 is cleaved by both OMA1 and YME1L, determination of the levels of these proteases could help shed some light on the mechanism leading to cristae reorganization.

We will follow up on Reviewer 2's comment with a WB analysis of OMA1 and YMEL in cells treated with an MCL-1 inhibitor.

- Generally speaking, while the authors show all those effects (cristae defects, FAO dysfunction) upon MCL-1 inhibition, it would be interesting to see whether any of those effects can be rescued by blocking FA import e.g. through carnitine palmitoyl- transferase 1a (CPT1a) inhibition with etomoxir to understand if they are downstream of altered Fa supply. This could affect cristae morphology through altered Cardiolipin biogenesis.

This is an excellent point, which was also raised by reviewer 1. We have now included experiments using etomoxir and perhexiline to assess their effects on TBR2/PAX6 (Figure 2 for Reviewers). As mentioned above, the results indicate that inhibiting the FAO pathway does not fully mimic the effects of MCL-1i on TBR2. However, we show that MCL-1i displaces ACSL1 from the mitochondria, a step that is upstream of CPT1 and 2. We suggest a model in which the coordinated non-apoptotic function of MCL-1 at the outer mitochondrial membrane promotes ACSL1 activity and, in the inner mitochondrial membrane, regulates mitochondrial cristae morphology. While our data point to this model, we are limited by the tools to investigate it further, but it will be a great direction for future experiments. The suggestion of Reviewer 2 that the effects on FAO could impact cardiolipin biogenesis is a very exciting possibility. However, difficult to test with the tools available.

- In line 262 the authors discuss that mitochondria lose metabolic function upon MCL-1 inhibition. This claim would require additional experiments. While the authors look at lipid droplet accumulation and FAO enzymes, there are many more aspects to mitochondrial metabolic function that should be investigated. While measuring the oxygen consumption rate via Seahorse might require additional resources (optional), measurements of ATP production, ROS generation or determination of the mitochondrial membrane potential should be feasible.

We fully agree with Reviewer 2's comment, which was also raised by Reviewer 1. In our revision, we will include an assessment of the mitochondrial oxygen consumption rate of NPCs treated with MCL-1i, measured using the Seahorse analyzer (mitochondrial stress test). These data are presented as Figure 3 for reviewers. Interestingly, these data show a more nuanced cellular response. While MCL-1i does not globally collapse mitochondrial respiration at baseline, the specific deficits appear in spare respiratory capacity and maximal respiration, meaning cells can sustain routine mitochondrial function but lose the ability to respond to increased energetic demand. This suggests MCL-1 loss creates a mitochondrial reserve deficiency rather than a generalized bioenergetic failure. The results with caspase inhibitors show a near-zero OCR across both 24h and 48h timepoints, and significant reductions in maximal respiration, spare respiratory capacity, and non-mitochondrial OCR. These conditions are detrimental for TBR2-positive NPCs (Figure 6) , but not for newborn neurons (Figure 7).

- While the authors "propose a model in which MCL-1 associates with MICOS", they do not offer direct scientific to support this hypothesis. Co-immunoprecipitation experiments or e.g. proximity ligation assays would better support the proposed model.

We agree with this statement. Preliminary, we have performed proximity ligation assays and immunoprecipitation analyses to test for this interaction (see below and ____Figure 4 for reviewers), and the results indicate an interaction, albeit very weak. In the revised version of the manuscript, we will attempt to repeat these experiments with MCL-1i.

- While Fig. 7 shows representative images, quantification e.g. for the truncation of neuronal processes is missing.

We will provide quantification of BIII tubulin branching, which will be included alongside the images provided.

- In lines 219f. the authors state that they "observed a significant downregulation of PAX6 and EOMES at 24 hours that was not rescued by QVD co-treatment". While there is still a trend towards a downregulation, there is no statistical significance anymore. In fact, PAX6 levels almost mirror those of SOX2 which is not described as "downregulated" by the authors. In order to be more consistent, I would suggest rephrasing this part, or at least reword it to be less absolute.

In the new version, we will clarify that while QVD rescued TBR2 and PAX6 transcript levels at 24h, it did not rescue them at 48h. We will also mention the downregulation of SOX2 at 48h that persists with co-treatment.

- Brinkmann et al. (2025) also investigated cristae structure upon MCL-1 deletion in vivo and found no effect when MCL-1 was replaced with other Bcl-2 family members. It would be interesting to combine MCL-1 inhibition with overexpression of MCL-1 versus BCL-XL to reconsolidate some of the discrepant findings.

While this is a great suggestion for future studies, there are some complications. Specifically, it is likely that the inhibitor may also target the overexpressed MCL-1 and thus, a mutant form is needed.

To address this, we generated a Flag-tagged MCL-1 construct with a mutated BH3 domain, previously described by Kotschy et al. Nature 2016. We validated the construct in HeLa cells, but unfortunately the mutant protein appears to be significantly less stable than the WT construct, complicating analysis of this experiment.

Minor comments:

- In Supp. Fig. 1C the MCL-1 protein is shown both to run above 37kDa (upper panel) and below 37 kDa (lower panel). Could the authors please comment on why this is the case?

The observed variation is caused by drift in the gel during electrophoresis. In Fig 1C, the protein ladder is on the edge of the gel, whereas in Fig 1E, the protein ladder is in the middle of the gel, and the last sample is on the edge and also exhibits edge drift.

- In line 64 of the introduction the authors mention clinical trials yet do not give a citation for these trials making it hard to judge whether the content of these trials is actually related to the brain.

This information is anecdotal, based on an Amgen press release.

- MCL-1 as well as ACSL-1 are sometimes written without the hyphen both in the text and figures.

We will carefully check the manuscript before submission.

- Lines 92-94 and 106-108 essentially highlight the same existing knowledge gap. Maybe the content of these two paragraphs could be combined in order to avoid repetition.

We thank Reviewer 2 for this suggestion. We will do this in the new version of the manuscript.

- In Fig. 1A, the authors provide a schematic for their experimental design. While the figure legend is very thorough, some of this information (like the days of collection) could also be included in the figure itself. The same is true for schematics in the following figures.

We agree with this and will incorporate the suggestion in the new version.

- Fig. 2A includes a typo (analyze) but would maybe also be more suitable for the supplement figures or could even be combined with Fig. 1A as not much new content is added.

We already incorporated these changes in the new version of the manuscript.

- Regarding statistical analysis, could the authors please comment on why they did not consider one-sample t-tests suitable for the cases where control values were set at 1 (e.g. Fig. 4B, C for the relative expression).

This is a valid suggestion. We will rerun RT-qPCR data using a one-sample t-test.

- In lines 247f. the authors state that "inhibition of MCL-1 leads to [...] and disassembly of the MICOS complex as well as OPA1". This sounds like OPA1 is still cleaved upon MCL-1, which is not at all what the authors showed and further discuss. Rewording of the sentence would help in avoiding any misunderstandings.

We agree with this comment and have now reworded the paragraph: "Inhibition of MCL-1 leads to structural collapse of the cristae likely due to the possible disassembly of the MICOS complex, as suggested by decreased MIC10 levels, and interruption of OPA1 cleavage, as suggested by decreased short-form OPA1, two scaffolds required for cristae maintenance."

- In lines 210f. the authors state that "quantitative imaging increased the average and maximum volume of lipid droplets". While there is definitely a trend towards an increase for the maximum volume, the increase is in fact not statistically significant. This should be reflected in the wording.

We have reworded this to "Quantitative imaging revealed a significant increase in average lipid droplet volume and a trending increase in maximum volume of lipid droplets."

- In Fig. 6 the overlap between TBR2 and PAX6 is hard to judge when printed out. Including a zoom-in may make it easier to judge.

We agree with Reviewer 2. In the new version of the manuscript, we will include panels that zoom into the cell populations we quantified. The current panels will go to a new Supplemental figure. We will also add the TUBB3 to the qPCR panel in the new version.

- In Fig. 7 the color-coding is listed in the figure legend but is missing from the figure itself. If the authors could include this, as they did for the other figures, it would further improve this figure.

We agree. We have specified the channel color in the new figure.

- Line 238 should reference Fig. 7A, as Fig 7B does not exist.

Thanks for catching this. It is already corrected

- In the figure legends the authors state that biological replicates were used. Were technical replicates also performed?

Yes, technical replicates were performed for RT-qPCR.

Reviewer #2 (Significance (Required)):____ Significance

The authors make use of a wide array of imaging techniques to further elucidate non-apoptotic roles of MCL-1. The study has the potential to offer new insights into mitochondrial biology on the level of basic research rather than translational. While the methods used offer sufficient evidence to connect MCL-1 inhibition to cristae architecture, the mechanistic underpinnings of this effect remain unexplored. Nevertheless, the study offers additional knowledge on the role of MCL-1 in human neural stem cells, whereas previous research mostly focused on cardiomyocytes or cancer cells.

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

Summary: ____ In this study, Gama et al. describe a non-canonical role for the anti-apoptotic protein Myeloid Cell Leukemia-1 (MCL-1) in mitochondrial cristae organization and suggest a role of MCL-1 in regulating metabolism and neuronal differentiation. Using fluorescence microscopy imaging and electron microscopy, the authors show changes to mitochondrial morphology upon treatment with MCL-1 inhibitor S63845. MCL-1 inhibition results in altered protein and transcript levels of some key proteins involved in mitochondrial cristae organization and fatty acid metabolism. While some of the findings are interesting and indeed point towards a non-canonical role of MCL-1, several key conclusions of the authors are not sufficiently supported by the data shown in the manuscript.

We thank Reviewer 3 for the careful evaluation of our manuscript. We appreciate the reviewer's recognition that our study identifies a potential non-canonical role for MCL-1 in mitochondrial cristae organization, metabolism, and neuronal differentiation. As with Reviews 1 and 2, we are encouraged that the reviewer finds these observations interesting and suggestive of previously unappreciated functions for MCL-1. We agree that stronger evidence is required to firmly link MCL-1 inhibition to specific changes in MICOS organization and metabolic regulation. In the revised manuscript, we will (i) more clearly distinguish between observations and mechanistic inferences, (ii) temper conclusions where appropriate, and (iii) incorporate additional analyses and controls to better substantiate the proposed model.

Major comments:

1. The authors try to disentangle the apoptotic and non-apoptotic role of MCL-1 through addition of a caspase inhibitor. However, I am not convinced that phenotypes found under the addition of caspase inhibitor are necessarily caused by non-canonical functions independent of apoptosis. It could also be that the observed changes happen upstream of caspase activation. In addition, many of the described finding, such as CPT1 expression changes, only happen in the presence of the caspase inhibitor. If one follows the logic of the authors, changes associated by non-canonical MCL-1 functions should happen under MCL-1 inhibition and caspase inhibition, but not with MCL-1 inhibition only____. __ The reviewer is right that we expected non-canonical functions to happen under MCL-1 inhibition and caspase inhibition. Our data with QVD shows that the cell death function of MCL-1 (i.e., inhibiting cell death effectors from initiating the caspase cascade) is not the main trigger of the phenotypes we report (cristae dysregulation and fatty acid oxidation disruption), however, cells without a functional cristae and/or defects in FAO, may not be able to survive long-term. Thus, QVD treatment preserves these cells that may not survive the dismantling of such an essential structure. To confirm this, we performed immunofluorescence of cleaved caspase 3 (__Figure 5 for reviewers). These results show that, indeed, MCL-1 inhibition at the time points of our study doesn't result in increased Caspase-3 activation. We reported similar results of MCL-1 inhibition in oligodendrocyte precursor cells (Gil and Hanna et al., Glia, 2025, PMID: 41420072).

The authors show no data on the viability of the cells in response to the MCL-1 inhibitor. To exclude secondary effects of the inhibitor, at least some of the results should be validated with an MCL-1 knock down.

We will include this experiment in our revised manuscript. To check the effects of MCL-1 knockdown on TBR2 positive cells, we tested 5 different ASOs for MCL-1. Knockdown efficiency with ASOs was very low (on average In Figure 1, the authors show immunofluorescence data of mitochondria and nucleus staining and conclude that MCL-1 inhibition alters mitochondrial morphology. Based on the images shown in Fig. 1, I do not think that individual mitochondria can be segmentd to measure their volume and length. In addition, some metrics such as mitochondrial content are not explained in the text or methods.

We can achieve mitochondrial segmentation with a SoRa Spinning Disk Confocal Microscope, which has a lateral (XY) resolution of approximately 120 nm to 150 nm and an axial (Z) resolution of approximately 300 nm-320 nm. All images are first denoised prior to sharpening using the Richardson-Lucy deconvolution algorithm. Additionally, the FIB-SEM data are consistent with the IF data (both show increase in mitochondrial volume and surface area).

We agree with Reviewer 3 that we need to explain some metrics in the revised version. We will specify the meaning of mitochondrial content (count of all mitochondria in FOV, not normalized to Hoechst).

In Fig. 2 B-D, the authors show TEM and FIB-SEM imaging to demonstrate alterations in the cristae architecture upon treatment with MCL-1 inhibitor. However, based on the images shown, it looks that cristae area and density is reduced under S63845 treatment in TEM images, while the FIB-SEM data come to the opposite conclusion. In addition, the quantification of cristae volume quantified as cristae volume in percentage is unclear to me.

We apologize for the confusion. No conclusions about the cristae area and density were made using the TEM data, because TEM data represent a single snapshot section of a mitochondrion without a discernible orientation. Cristae from TEM were described as "aberrant" and preliminarily revealed changes in cristae and were followed up with FIB-SEM, 3D reconstruction of intact mitochondria, and quantification of volume.

In the new version of the manuscript, we will specify that the cristae volume is normalized to the volume of its respective mitochondria (i.e., how much of the mitochondrial volume is attributed to cristae).

The change in CPT1/2 protein levels (Fig. 4) is interesting but does not directly proof that fatty acid oxidation is altered, as concluded by the authors. For this, the authors would need to directly measure fatty acid oxidation for example using Seahorse or metabolic tracing experiments. Also, to prove that the MCL-1 inhibition affects neural differentiation through fatty acid oxidation, a rescue experiment should be performed through CPT1 overexpression.

We agreed that this is an important point. We have optimized the fatty acid oxidation test using Seahorse and will make sure to include it in the revised version of the manuscript.

In Figure 6, the authors show decreased intermediate progenitor cells after MCL-1 inhibition by immunofluorescence staining. I am not convinced that this can be concluded from the data shown, since the concentration of intermediate progenitor cells is very close to the noise levels. Since the MCL-1 treated cells look much less sparse, I don't think the percentages can be compared (total counts are between 2-20). Although this data might give some indication that differentiation could be impaired, the measured effect could be very well due to lower viability of the cells. The authors need to control for this or come up with a different method for measuring differentiation.

The number of TBR2 is low, but we disagree with the reviewer's assessment of noise levels. We focused on cells expressing only TBR2 and rigorously examined this population of cells. The percentages are compared to account for the lower density of the MCL-1i-treated cultures, as the IPC counts are normalized to the Hoechst total cell count within the FOV. Moreover, the immunofluorescence images are complemented with RT-qPCR, which shows significant downregulation of EOMES (gene encoding TBR2).

Figure 7 is missing quantification

We will include this quantification in the revised version of the manuscript.

Reviewer #3 (Significance (Required)):

General assessment____: The manuscript reports an interesting finding, which suggest a non-canonical role of MCL-1 in mitochondrial remodeling, regulation of fatty acid oxidation and neuronal fate. While this finding would be highly interesting and relevant, the presented data do not sufficiently support this conclusion. Further experiments would have to be performed to proof causality. ____ Advance: Should the authors manage to proof their hypothesis by additional experiments, this would indeed advance the field on mitochondrial remodeling and its effect on neuronal differentiation by

identifying a novel molecular player. ____ Audience: mitochondrial biology, cell biology, developmental neuroscience Own expertise: mitochondrial biology, cell biology, advanced imaging 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 #3

Evidence, reproducibility and clarity

Summary:

In this study, Gama et al. describe a non-canonical role for the anti-apoptotic protein Myeloid Cell Leukemia-1 (MCL1) in mitochondrial cristae organization and suggest a role of MCL1 in regulating metabolism and neuronal differentiation. Using fluorescence microscopy imaging and electron microscopy, the authors show changes to mitochondrial morphology upon treatment with MCL1 inhibitor S63845. MCL1 inhibition results in altered protein and transcript levels of some key proteins involved in mitochondrial cristae organization and fatty acid metabolism. While some of the findings are interesting and indeed point towards a non-canonical role of MCL1, several key conclusions of the authors are not sufficiently supported by the data shown in the manuscript.

Major comments:

1. The authors try to disentangle the apoptotic and non-apoptotic role of MCL1 through addition of a caspase inhibitor. However, I am not convinced that phenotypes found under the addition of caspase inhibitor are necessarily caused by non-canonical functions independent of apoptosis. It could also be that the observed changes happen upstream of caspase activation. In addition, many of the described finding, such as CPT1 expression changes, only happen in the presence of the caspase inhibitor. If one follows the logic of the authors, changes associated by non-canonical MCL1 functions should happen under MCL1 inhibition and caspase inhibition, but not with MCL1 inhibition only.
2. The authors show no data on the viability of the cells in response to the MCL1 inhibitor. To exclude secondary effects of the inhibitor, at least some of the results should be validated with an MCL1 knock down.
3. In Figure 1, the authors show immunofluorescence data of mitochondria and nucleus staining and conclude that MCL1 inhibition alters mitochondrial morphology. Based on the images shown in Fig. 1, I do not think that individual mitochondria can be segmentd to measure their volume and length. In addition, some metrics such as mitochondrial content are not explained in the text or methods.
4. In Fig. 2 B-D, the authors show TEM and FIB-SEM imaging to demonstrate alterations in the cristae architecture upon treatment with MCL1 inhibitor. However, based on the images shown, it looks that cristae area and density is reduced under S63845 treatment in TEM images, while the FIB-SEM data come to the opposite conclusion. In addition, the quantification of cristae volume quantified as cristae volume in percentage is unclear to me.
5. The change in CPT1/2 protein levels (Fig. 4) is interesting but does not directly proof that fatty acid oxidation is altered, as concluded by the authors. For this, the authors would need to directly measure fatty acid oxidation for example using Seahorse or metabolic tracing experiments. Also, to prove that the MCL1 inhibition affects neural differentiation through fatty acid oxidation, a rescue experiment should be performed through CPT1 overexpression.
6. In Figure 6, the authors show decreased intermediate progenitor cells after MCL1 inhibition by immunofluorescence staining. I am not convinced that this can be concluded from the data shown, since the concentration of intermediate progenitor cells is very close to the noise levels. Since the MCL1 treated cells look much less sparse, I don't think the percentages can be compared (total counts are between 2-20). Although this data might give some indication that differentiation could be impaired, the measured effect could be very well due to lower viability of the cells. The authors need to control for this or come up with a different method for measuring differentiation.
7. Figure 7 is missing quantification

Significance

General assessment: The manuscript reports an interesting finding, which suggest a non-canonical role of MCL1 in mitochondrial remodeling, regulation of fatty acid oxidation and neuronal fate. While this finding would be highly interesting and relevant, the presented data do not sufficiently support this conclusion. Further experiments would have to be performed to proof causality.

Advance: Should the authors manage to proof their hypothesis by additional experiments, this would indeed advance the field on mitochondrial remodeling and its effect on neuronal differentiation by identifying a novel molecular player.

Audience: mitochondrial biology, cell biology, developmental neuroscience

Own expertise: mitochondrial biology, cell biology, advanced imaging 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 #2

Evidence, reproducibility and clarity

Summary:

This manuscript by Hanna et al. investigates non-apoptotic roles of MCL-1 in human neural stem cells and connects MCL-1 inhibition to mitochondrial cristae formation and beta-oxidation. Connecting these roles to brain development, the authors also show a reduction in the number of progenitor cells upon MCL-1 inhibition, independently of caspase activity. Throughout their work, the authors make use of an impressive array of imaging techniques.While the methods used offer sufficient evidence to connect MCL-1 inhibition to cristae architecture, the mechanistic underpinnings of this effect remain unexplored.

Major comments:

• In Fig. 1B, the very same representative images are shown for both conditions (DMSO and S63845) at 48 hours.
• For Western Blot analysis, it looks like the authors only quantified the band density of their proteins of interest without considering varying levels of control protein (Actin) levels. Normalizing the protein levels to actin would account for any differences in loaded protein amounts (although a Ponceau staining might be preferable still to exclude this). This is especially relevant for Fig. 4E, where actin levels visibly differ between the conditions.
• The authors offer evidence that MCL-1 inhibition impedes proteolytic cleavage of OPA1-L into the OPA-1-S isoforms, yet do not explore the mechanism behind this. Since OPA1 is cleaved by both OMA1 and YME1L, determination of the levels of these proteases could help shed some light on the mechanism leading to cristae reorganization.
• Generally speaking, while the authors show all those effects (cristae defects, FAO dysfunction) upon MCL-1 inhibition, it would be interesting to see whether any of those effects can be rescued by blocking FA import e.g. through carnitine palmitoyl- transferase 1a (CPT1a) inhibition with etomoxir to understand if they are downstream of altered Fa supply. This could affect cristae morphology through altered Cardiolipin biogenesis.
• In line 262 the authors discuss that mitochondria lose metabolic function upon MCL-1 inhibition. This claim would require additional experiments. While the authors look at lipid droplet accumulation and FAO enzymes, there are many more aspects to mitochondrial metabolic function that should be investigated. While measuring the oxygen consumption rate via Seahorse might require additional resources (optional), measurements of ATP production, ROS generation or determination of the mitochondrial membrane potential should be feasible.
• While the authors "propose a model in which MCL-1 associates with MICOS", they do not offer direct scientific to support this hypothesis. Co-immunoprecipitation experiments or e.g. proximity ligation assays would better support the proposed model.
• While Fig. 7 shows representative images, quantification e.g. for the truncation of neuronal processes is missing.
• In lines 219f. the authors state that they "observed a significant downregulation of PAX6 and EOMES at 24 hours that was not rescued by QVD co-treatment". While there is still a trend towards a downregulation, there is no statistical significance anymore. In fact, PAX6 levels almost mirror those of SOX2 which is not described as "downregulated" by the authors. In order to be more consistent, I would suggest rephrasing this part, or at least reword it to be less absolute.
• Brinkmann et al. (2025) also investigated cristae structure upon MCL-1 deletion in vivo and found no effect when MCL-1 was replaced with other Bcl-2 family members. It would be interesting to combine MCL-1 inhibition with overexpression of MCL-1 versus BCL-XL to reconsolidate some of the discrepant findings.

Minor comments:

• In Supp. Fig. 1C the MCL-1 protein is shown both to run above 37kDa (upper panel) and below 37 kDa (lower panel). Could the authors please comment on why this is the case?
• In line 64 of the introduction the authors mention clinical trials yet do not give a citation for these trials making it hard to judge whether the content of these trials is actually related to the brain.
• MCL-1 as well as ACSL-1 are sometimes written without the hyphen both in the text and figures.
• Lines 92-94 and 106-108 essentially highlight the same existing knowledge gap. Maybe the content of these two paragraphs could be combined in order to avoid repetition.
• In Fig. 1A, the authors provide a schematic for their experimental design. While the figure legend is very thorough, some of this information (like the days of collection) could also be included in the figure itself. The same is true for schematics in the following figures.
• Fig. 2A includes a typo (analyze) but would maybe also be more suitable for the supplement figures or could even be combined with Fig. 1A as not much new content is added.
• Regarding statistical analysis, could the authors please comment on why they did not consider one-sample t-tests suitable for the cases where control values were set at 1 (e.g. Fig. 4B, C for the relative expression).
• In lines 247f. the authors state that "inhibition of MCL-1 leads to [...] and disassembly of the MICOS complex as well as OPA1". This sounds like OPA1 is still cleaved upon MCL-1, which is not at all what the authors showed and further discuss. Rewording of the sentence would help in avoiding any misunderstandings.
• In lines 210f. the authors state that "quantitative imaging increased the average and maximum volume of lipid droplets". While there is definitely a trend towards an increase for the maximum volume, the increase is in fact not statistically significant. This should be reflected in the wording.
• In Fig. 6 the overlap between TBR2 and PAX6 is hard to judge when printed out. Including a zoom-in may make it easier to judge.
• In Fig. 7 the color-coding is listed in the figure legend but is missing from the figure itself. If the authors could include this, as they did for the other figures, it would further improve this figure.
• Line 238 should reference Fig. 7A, as Fig 7B does not exist.
• In the figure legends the authors state that biological replicates were used. Were technical replicates also performed?

Significance

The authors make use of a wide array of imaging techniques to further elucidate non-apoptotic roles of MCL-1. The study has the potential to offer new insights into mitochondrial biology on the level of basic research rather than translational. While the methods used offer sufficient evidence to connect MCL-1 inhibition to cristae architecture, the mechanistic underpinnings of this effect remain unexplored. Nevertheless, the study offers additional knowledge on the role of MCL-1 in human neural stem cells, whereas previous research mostly focused on cardiomyocytes or cancer 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

Summary: This study claims that beyond its canonical anti-apoptotic function, MCL-1 has essential non-apoptotic roles in human neurodevelopment. Pharmacologic inhibition of MCL-1 in human neural stem cells disrupts mitochondrial inner membrane architecture by destabilizing cristae and the OPA1-MICOS complex, leading to swollen mitochondria with disorganized cristae. These structural defects impair fatty acid oxidation and lipid droplet homeostasis, linking cristae integrity to metabolic competence. Independently of apoptosis or proliferation, MCL-1 inhibition selectively depletes intermediate neural progenitors, indicating a direct role in lineage progression. Overall, the work positions MCL-1 as a key regulator of mitochondrial structure-metabolism coupling that instructs neural progenitor identity and human neurogenesis.

Overall: The study does a good job of using (in most assays) caspase inhibition (e.g., QVD treatment) to block apoptotic responses induced by MCL-1 inhibition. As a result, many of the phenotypes caused by inhibition are likely to be independent of caspase activation. As a result, this manuscript would be of interest to researchers that study the topics of the BCL-2 family and cell death signaling, mitochondrial bioenergetics and dynamics, neurodevelopment, and cellular metabolism. However, as currently presented the manuscript is only descriptive and lacks mechanistic insight.

Major Concerns:

1) The authors only use a single MCL-1 inhibitor and never use other non-targeting BH3-mimetics (such as venetoclax) as negative controls. This seems like a missed opportunity to demonstrate that the phenotypes observed are MCL-1 dependent.

2) There is no mechanism proposed in this study other than reliance upon QVD as not affecting the phenotypes. As submitted, the manuscript only can speculate that these phenotypes are due to non-apoptotic roles of MCL-1 inhibition. The authors have missed an opportunity to explore MCL-1's non-apoptotic functions directly.

Other concerns exist that weaken the impact of the study.

1. Figure 1 should include the fact that QVD inhibition (shown in Sup Fig 2) does not obviate the phenotype induced by pharmacological inhibition of MCL-1 on mitochondrial morphology.
2. Figure 2 would benefit from evidence that caspase inhibition does not repress the phenotype on mitochondrial cristae morphology (volume and area). Furthermore, the FIB-SEM data are very hard to appreciate as the size precludes visualization of individual mitochondria.
3. Figure 3 reports that MIC60 and OPA1 appear to be downregulated in response to MCL-1 inhibition, but these appear to be more significant only when QVD is added. Why would the phenotype be obscured in the non-QVD setting (Fig. 2B&C). How does MCL-1 inhibition lead to changes in MIC60/MICOS/OPA1? This seems quite preliminary at this point.
4. The loss of MIC60 and OPA1 should repress electron transport chain function, are such impacts observed in the cultured cells? This could be shown by assessing oxygen consumption, etc. Such data would enhance the authors' conclusion that MCL-1 inhibition leads to defects in mitochondrial physiology.
5. In Figure 4, the differences between transcripts (qPCR data) and protein (immunoblot) data are often confusing and not well explained. Why do the authors propose that mRNA expression is decreasing whereas the protein expression is increasing? Example CPT1. Furthermore, it is unclear what these data mean functionally? Is this reflective of enhanced lipid oxidation or simply a response to inhibition of fatty acid oxidation? Clarification of the impact of these findings is necessary.
6. The increase in lipid droplet number induced by MCL-1 inhibition has been previously documented, but it is unclear whether this increase is related to an inability to oxidize lipid (defective fatty acid oxidation) that leads to increases in the cellular abundance or whether this indicates that MCL-1 inhibition leads to enhanced storage. Do other inhibitors of fatty acid oxidation lead to similar increases in lipid droplet size and abundance? Does QVD inhibition affect this phenotype?
7. For Figure 6, while these data may be very meaningful, as presented they are very hard to appreciate. Insets that show the neuronal populations would help to convey the point that the differentiation is impacted. Also, are there other methods that could confirm these observations (qPCR to show changes in differentiation).
8. Figure 7 is also very hard to appreciate. What is the reader to see? Can these be quantified? It seems that QVD may be rescuing in this figure, does this suggest that MCL-1 inhibition might be inducing death. All of this needs to be quantified.
9. BCL-XL has also been implicated in affecting mitochondrial electron transport chain function (See PMID: 19255249, 21926988, 21987637). Can BCL-XL inhibitors affect any of the phenotypes associated here?
10. Please be carefully avoid using the term "MCL-1 loss", when talking about pharmacological inhibition. Only genetic ablation (e.g. knockout, silencing, etc.) should be termed loss.

Significance

The study advances in human cells the impacts of MCL-1 inhibition. They replicate many impacts previously observed in mouse systems and refine analyses to impacts on MICOS complex, lipid droplet storage, and neuronal differentiation. While these findings are important and would be well received by a wide audience, the study fails to provide almost any mechanistic insight into how these phenotypes are being induced. The only common theme is that blocking caspase activation in many assays fails to block the phenotype.

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 time and constructive comments on our manuscript.

Reviewer #1 recognizes the importance of the question we address (namely, the early consequences of Wilms' tumour inducing mutations on kidney development in two models for different Wilms' tumour initiating mutations) and provides useful suggestions for improvement of the manuscript.

Reviewer #2 raises the concern regarding the novelty of the study. We appreciate these comments and this implies the necessity of mainly textual changes we have to do to highlight the novel aspects of our study and findings and their significance in the revision of the manuscript.

Reviewer #3 offers a generally positive assessment of the data, while suggesting that the work may be interpreted primarily from a developmental perspective rather than a Wilms' tumour-focused one. In the revision there is need to better emphasize how these perspectives are closely interconnected in the context of Wilms' tumour biology.

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

This manuscript addresses an important gap in Wilms tumor (WT) biology: what are the earliest pathogenic events following WT driver mutation induction, and how do these early developmental trajectories differ across genotypes? The authors provide a carefully staged and comparative analysis of two WT-associated genetic contexts-conditional Wt1 loss (using lineage-specific Cre drivers targeting nephrogenic (Six2-Cre) versus stromal (Foxd1-Cre) compartments, as well as a temporally controlled Wt1CreERT2 model targeting both lineages upon tamoxifen induction) and inducible LIN28B overexpression, and relate the resulting developmental phenotypes to two CSC marker paradigms derived from patient-based studies. A major strength is the precise, time-resolved description of the earliest initiating phenotypes (E12.5 and E18.5, with additional postnatal analysis for LIN28B) and the direct side-by-side comparison of how each genotype perturbs nephrogenesis. The authors conclude that Wt1 loss (especially in the nephrogenic lineage) leads to a severe developmental block accompanied by a disturbance of lineage identity ("lineage confusion"), whereas LIN28B overexpression causes a disturbed transition between uninduced and induced nephron progenitor cell (NPC) states, producing blastemal-like regions that persist postnatally. Using immunostaining for NCAM1, SIX2, CITED1, and ALDH1A2, the authors map marker combinations during normal kidney development and across mutant contexts, and propose that tumor-initiating alterations, most clearly in the LIN28B model, and more suggestively in the Wt1CreERT2 (Wt1CE) context, promote the emergence of a CSC-like population inferred to co-express all four markers (NCAM1+SIX2+CITED1+ALDH1A2+), a state not observed in normal kidneys.

We thank Reviewer #1 for this correct and complete summary of our manuscript. This reviewer recognizes the current gap in our understanding of the origins of Wilms' tumors and appreciates the approach we have chosen to start filling this gap using two different mouse models.

Overall, this study provides a particularly clear direct comparison of the earliest tumor-initiating events triggered by distinct WT-relevant driver alterations. While the manuscript does not yet offer a detailed molecular mechanistic framework explaining why these two mutations produce such divergent developmental and marker-state outcomes (which would further strengthen the work), the careful comparison and the conclusions drawn from it are meaningful and make an important contribution to our understanding of the developmental processes that can lead to Wilms tumor initiation.

We thank this reviewer for recognizing the importance of a direct comparison of the early consequences of two different Wilms' tumour mutations. We agree we do not yet provide a mechanistic framework for these differences. Although these studies are on-going, they are outside the scope of this manuscript.

*Major comment: 1. A central and highly emphasized conclusion of this manuscript is that tumor-initiating alterations induce a CSC-like population co-expressing all four markers (NCAM1, SIX2, CITED1, and ALDH1A2), and that this state is not observed during normal kidney development. Because this "quadruple-positive" population is a key mechanistic take-home message and closely linked to the overall conceptual model, the manuscript would be substantially strengthened by a direct, same-cell demonstration of co-expression of all four markers, rather than inference from consecutive sections. The authors state that they were unable to do so due to a technical limitation, namely, antibody host-species constraints that prevent co-detection of CITED1 and ALDH1A2 within the same section. *

We agree that not being able to show co-expression of all 4 CSC markers is a serious limitation for the interpretation of our data. The reviewer suggests the following alternatives:

*Several feasible approaches could address this limitation for example: - Identify an alternative antibody reagent from a different host species. *

The 'problematic' antibodies are the ones staining for ALDH1A2 and CITED1, which are both Rabbit IgG. Alternative antibodies for ALDH1A2 are all raised in rabbit, so these will not solve this problem. For CITED1 we have now identified a biotin-conjugated antibody which could be used in additional co-staining. We propose to test this antibody for the revision of this manuscript.

*- RNAscope / smFISH for in situ single-cell co-detection. *

We are aware of these techniques as alternative for antibody staining. However, we have no experience with these techniques, nor do we have access to the required technologies. After discussions with collaborators with much experience in this technique, we realized the combination of the potential extensive optimization and costs does not make this a suitable alternative for the limited samples we have available.

*- Single-cell RNA-seq (scRNA-seq) to test whether a bona fide quadruple-positive transcriptional state exists. *

This could be an option but is itself a huge project and therefore outside the scope of this manuscript. We note that the known scarcity in single cell data might still complicate the detection of each marker in individual cells, especially for low-expressed TFs like Six2 and Cited1.

*Overall, resolving this technical limitation would markedly increase confidence in one of the manuscript's most important claims and strengthen the proposed genotype-phenotype/CSC-marker framework

*

_As discussed above, we propose the t_ry the biotin-conjugated CITED1 antibody__

• It is somewhat unexpected that the Six2-specific Wt1 deletion appears to produce a more severe phenotype than the tamoxifen-inducible Wt1CreERT2 approach, which is intended to target a broader Wt1-derived lineage (both nephrogenic and stromal). The Discussion offers several plausible, non-mutually exclusive explanations for this observation (e.g., timing, recombination efficiency/mosaicism, and the rescue contribution of "escaping" wild-type cells). It would be helpful to support at least one of these explanations experimentally. For example, the authors could quantify the extent of "escape" (percentage of non-recombined cells within the lineage) across embryos/timepoints to validate that mosaicism is indeed the cause of the milder phenotype. *

We can address this experimentally by making use of the tdTomato Cre reporter that was included in our model which allows us to follow the fate of mutated and non-mutated cells in the lineage. We propose to combine Six2 antibody staining with the tdTomato signal to quantify the percentage of cells that has maintained Six2 expression and is therefore likely an escaping cell/nephron.

Minor comments 1. Please clarify whether the difference shown in Fig. 2C is statistically significant, and report n, error bars/variation, the statistical test used, and p-values (if applicable).

These details will all be added.

• The authors note the presence of some SIX2+; tdTomato+ cells in Foxd1GC control kidneys. Given the expected stromal restriction of Foxd1 lineage labeling, please clarify the likely explanation and, if possible, indicate how frequent this is.

*

The reviewer here points to the important question regarding the origin and potential overlap between the stromal and nephrogenic lineages. This is not only an important but highly relevant question for origin and biology of Wilms' tumours, but also for normal kidney development. Kobayashi et al (2014) reported some contribution of the Foxd1 lineage in the Six2 lineage. Also Magella et al (2018) found some signs for this, as did a recent pre-print (Haghighitalab et al. 2026). There is even data suggesting that (part of) the renal stromal is derived from the paraxial instead of intermediate mesoderm in chicken (Guillaume et al. 2009) with some supportive data from mouse development as well (Levinson et al. 2005). The latter is especially interesting given the commonly found ectopic muscle differentiation in WT1-mutant Wilms' tumours (Miyagawa et al. 1998; Schumacher et al. 2003; Gadd et al. 2012). However, if a common, potentially Foxd1+/Six2+ double positive, progenitor exists, it will in the normal developing kidney be present before E11.5 and therefore the data in our current manuscript, or the unpublished scMultiome data, is not informative for this. We propose to discuss this in detail in the Discussion of the manuscript, and speculate on its relevance for Wilms' tumours.

It is somewhat unexpected that Six2-specific Wt1 deletion appears to produce a more severe phenotype than the tamoxifen-inducible Wt1CreERT2 approach, which is intended to target a broader Wt1-derived lineage. The Discussion offers several plausible, non-mutually exclusive explanations (e.g., timing and/or recombination efficiency/mosaicism and the contribution of "escaping" cells). it would be helpful to support at least one of these explanations experimentally, for example by quantifying the extent of "escape" across embryos/timepoints and tamoxifen dosing.

This was addressed above.

*4. A careful proofreading pass is needed to ensure text-figure consistency, particularly for arrow annotations. For example, the Results text refers to "Fig. 1F, arrows," but arrows are not apparent in that panel. Likewise, the Results text mentions a "white filled arrow" in Fig. 2H, whereas the figure appears to show only open arrows. Please align the wording with the annotations actually shown in the figures. *

We apologize for these errors and thank the reviewer for pointing them out. These, and all other textual and graphical errors, will be corrected in the new version of the manuscript.

__Reviewer #1 (Significance (Required)): __

Overall, this study provides a particularly clear direct comparison of the earliest tumor-initiating events triggered by distinct WT-relevant driver alterations. While the manuscript does not yet offer a detailed molecular mechanistic framework explaining why these two mutations produce such divergent developmental and marker-state outcomes (which would further strengthen the work), the careful comparison and the conclusions drawn from it are meaningful and make an important contribution to our understanding of the developmental processes that can lead to Wilms tumor initiation.

We thank the reviewer for this comment, and like to emphasize that this is precisely the scope we intended with the current manuscript.

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

*Wilm's Tumor, a pediatric kidney cancer, is associated with gain or loss of activity of a number of genes including the loss of activity of the nucleic acid binding protein WT1 and gain of activity (enhanced expression at the mRNA level) of the RNA binding protein Lin28 which negatively impacts the maturation of the miicroRNA let-7, elevating levels of let-7 targets. Previous mouse studies have examined the impact of loss of Wt1 throughout within the nephron progenitor and interstitial cell compartments in capping mesenchyme that is thought to be the source of the tumor and of broad elevated expression in all kidney progenitors. *

*In this manuscript, the authors have refined the loss of Wt1 to nephron or stromal progenitors and compared the phenotype to loss of Wt1 in both lineages examining cultured kidneys over a 72 hr period, in addition to uncultured kidneys examined at e18.5. A similar analysis was performed on Lin28 mutants. The analysis itself consisted of video imaging, limited immunostaining and histochemistry. *

Reviewer #2 provides, in our opinion, a very limited overview of the contents of our manuscript. Our work presented here shows:

• A detailed analysis of effects of Wt1 loss or activation of LIN28B in the following systems:
• 5 embryonic kidneys
• 5 embryo kidneys
• P19 postnatal kidneys (for the LIN28B model)
• In vitro cultured kidneys.

• Time-laps analysis of in vitro cultured kidneys

• In the case of the Wt1 knockout this was studied in nephrogenic, stromal, and the combination of nephrogenic and stromal lineages

• Whereas our previous work (Berry et al. 2015) focused on different stages of nephron development, we now focus on the different lineages.
• For the first time we study the different marker sets for Wilms' tumour cancer stem cells in their developmental context. Important take-home messages for this are:
• The two published maker sets behave different in the normal developing kidney, and no cell types or developmental stages exist in the normal developing kidney that expresses all four markers

• In contrast, after either of the two Wilms' tumour mutations are induced, we have strong, though not yet conclusive, evidence that this event induces cells that are positive for all four CSC markers, suggesting these quadruple-positive cells could be the functional CSCs. This mutation-dependent appearance of the CSCs would be a complete different mechanism for the origin of CSCs than believed for, for instance, leukemia and colorectal cancer, where an existing cell type with stem- or progenitor cell characteristics which already express the CSC markers picks up the tumour initiating mutations and thus starts behaving as CSC. The cascade our data suggests for the Wilms' tumour CSCs is much more complex.

• To our knowledge this is the first direct and side-by-side comparison of the early effects of different Wilms' tumour mutations. This analysis clearly shows the differences in underlying biology for these two situations, and this can have important consequences for interpretation of patients data (which was historically almost always generated without knowing the initiating mutation) and opens the possibility of mutation-specific therapeutic possibilities and requirements. This is funcamentally different from the current patient stratification based on clinical outcome (favorable vs non-favorable histology) or very general molecular markers with clear biological consequences (like chr 1p status).

• With respect to the mutation-dependent accumulation of CSC markers, although in both Wt1 and LIN28B models this seems to be happening, for the LIN28B model this seems to be the result of a simple developmental block, whereas for the Wt1 mutants this appears to be a lineage conversion phenotype. This is again something that has to our knowledge never been suggested for the origin of CSCs and even in the context of normal kidney development is almost unprecedented.
• We optimize the use of the Wt1CreERT2 driver to target different lineages in the developing kidney using different timepoints for tamoxifen treatment. Not only does this have technical use, it also illustrates the complex role of Wt1 in the earliest stages of kidney development.

Although the data presented are descriptive and do not yet provide a complete molecular mechanism, we believe they offer novel, unexpected and important insights that merit publication. We acknowledge that these aspects may not have been sufficiently clear in the original version of the manuscript, and therefore not being picked up by the reviewer. In response to Reviewer #2 comments, we propose a thorough rewrite of the Discussion of the manuscript to emphasize these aspects more.

*While wholly qualitative and largely observational and descriptive, the limited data are of good quality and the conclusions drawn are reasonable. *

We thank the reviewer for their compliments on the quality and conclusions of the data. While we acknowledge the reviewer's characterization of the study as quantitative and descriptive, we respectfully do not consider this to diminish its suitability for publication. We believe the dataset provides substantial and meaningful insights (definitely not limited), and we have clarified and expanded upon the novel aspects and significance of our findings as outlined above.

*For the Wt1 study, most interesting would be in the loss of Wt1 from the NPC lineage. Clearly, there is already a significant phenotype at the time of study (E12.5) hence there is no strong insight into the earliest effects of Wt1 loss and how this might contribute to tumor formation. Quite what happens to these cells phenotypically is unclear given the limited set of markers used to look at the cells. Specific removal of Wt1 from the stromal lineage generates a milder phenotype, indicating a role for Wt1 there, but without a mechanistic analysis of the resultant products, the underlying mechanisms remain unclear. *

As discussed in our response to reviewer #1, we agree on the lack of mechanism in the current study but emphasize here as well that although this is the topic of the on-going follow-up studies this is outside the scope of the current manuscript. We refer to the same response for our proposal for additional experiments for the revised version of the manuscript.

*Wt1 removal from both lineages generated a phenotype less severe than removal from nephron progenitors (and previous data on "double lineage removal" with a Nestin1 cre), an indication that the genetic approach was not up to the task. *

Respectfully, we would like to emphasize the practical challenges associated with the use of genetically modified mouse models for developmental (and other) studies. We doubt there are many Cre drivers that do exactly what they were intended to do, do only that, and at full 100% efficiency. Many Cre drivers are, when originally described, only described for the cell type they were intended for, and any other activities or limitations are missed or ignored. One could rightfully argue that is bad science, but unfortunately this is often the reality and the starting point for many in vivo analyses. And these are only the complications regarding the behaviour of the Cre driver, and does not even touch on issues like the biological processing of tamoxifen, and the stability of already existing mRNA and protein of the gene of interest in the context of, in this case, a rapidly developing organ. Simply dismissing technical complications as 'not up to the task' is in our opinion not the way forward for studying the origin of diseases.

What is important, and what we demonstrate, is the realization of the limitations of a system, test them and where possible take them into account in the interpretation of data. In this case, instead of hiding the incompleteness of the Cre activity, we actually demonstrate this using retained staining of Wt1 and discuss this in the context of the different phenotypes. We have carefully tried not to overinterpret our data, and note that this reviewer does not give any specifics where this could be affecting our manuscript.

We also like to stress that in the context of Wilms' tumour development the incomplete activity of this Cre driver could even increase the relevance of this model, since the early stages of Wilms' tumourigenesis in the (future) patient happen in a few mutant cells in the context of a further normal developing kidney. The effect of the normal cells in our model that we speculate about could also be important in the patient, we just don't have the technical possibilities to test this yet.

*In some sense, one could regard this work as a pilot study, looking to optimize expensive and time-consuming mouse experiments to maximize insight (ie choose optimum model, address most informative time points, decide on analytical approaches). As a stand-alone paper, the work may not significantly advance our understanding of the topic. *

As argued above, in our opinion this does not do justice to the work we describe in our manuscript.

For example, can simple loss of Wt1 tells us anything about Wt? Yes Wt1 is lost in a subset, but even in these there are additional genetic mutations.

Of course even in WT1-mutant tumours there will be additional mutations found in the tumour. In fact, it has been known for a long time that WT1-mutant Wilms' tumours select for oncogenic mutations in β-catenin with a surprising preference for specific mutations affecting Ser45. However, it is clear that in these tumours the loss of WT1 is the first, rate-limiting step (Fukuzawa et al. 2004; Li et al. 2004; Zirn et al. 2006; Uschkereit et al. 2007). These β-catenin mutations are selected for in an already WT1-mutant context. If we want to understand the full biology of the WT1 mutant tumours including the β-catenin mutation, we will first need to understand the effect of only losing WT1 because that is what provide the selective pressure for the next step (oncogenic mutation in β-catenin). The work described here is an essential first step in that.

• For Lin12, there is no significant advance beyond the studies of the Daly lab. *

As argued above, this is not correct. The following aspects were not covered in the original paper describing this model:

• The in vitro analysis of control and LIN28B embryonic kidneys, including the time-lapse analysis demonstrating how the phenotype develops over time
• The expression of the Wilms' tumour cancer stem cell markers and how these change as a result of the LIN28B activation
• The direct comparison to the Wt1 loss phenotypes, and the demonstration these different mutations lead to fundamentally different biological phenotypes despite both eventually being classified as Wilms' tumours.

  I have no useful suggestions for improvement which would require a completely different approach to the problem from the start.

We respect this reviewer's opinion, but based on the above we do not agree and maintain a different interpretation.

Reviewer #2 (Significance (Required)):

*The authors set out the goal in the introduction - to obtain a better understanding of the origins of Wilm's tumor. There doesn't appear to be an insight of cancer relevant significance beyond earlier studies. *

Our work studies the very first steps in the development of Wilms' tumours. It will never be possible to study this in the (future) patient as these happen around wk 8-10 of pregnancy. By instead analyzing this in mouse models we show fundamental biological differences between different Wilms' tumour inducing mutations which is for sure relevant for patients, the interpretation of patient data (or more the difficulties with interpreting patient data if the initiating mutation or tumour class is not known). Moreover, the data provides new insights in the Wilms' tumour cancer stem cells, a preferred target for any therapy, and suggests the combination of all four known markers might be required to identify and study the true WT CSC. In our opinion such findings provide extremely relevant insights for the field.

*To a readership now/too used to analysis at genome scales (genomic, transcription), this study might appear modest. *

While we agree that genome-wide approaches can provide valuable insights, this doesn't mean that work that doesn't use them cannot provide important insights nor does it mean that every piece of work that does use them provides any new insights. We respectfully emphasize that the merit of a study should be assessed based on the data presented and their interpretation, rather than on the techniques that were used to obtain them.

The target audience is unclear.

Our target audience for this manuscript is everybody who is interested, for whatever reason, in the biology of Wilms' tumours.

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

*Wilms tumor arises from disrupted kidney development. Progenitor-like populations and cancer stem cell (CSC) fractions have been described in patient tumors, but how specific mutations alter embryonic programs to generate these states remains unresolved. *

*Pop et al. model genotype-phenotype relationships during kidney embryogenesis. Using Six2- and Foxd1-driven Cre lines, they test the effects of Wt1 loss-of-function and Lin28b gain-of-function in nephron and stromal progenitors. Through explant imaging, histology, and immunofluorescence, they define mutation-specific effects on ureteric branching, cap mesenchyme organization, stromal composition, and nephron differentiation. *

*Lineage-restricted Wt1 deletion produces distinct outcomes depending on whether nephron progenitors, stromal progenitors, or both are targeted. Lin28b overexpression causes delayed nephrogenesis and lobular organization resembling human Wilms tumor morphology, with expansion of blastemal-like populations. *

This is a correct summary of this part of our data.

These genetic removals of Wt1 and overexpression of Lin28b are useful for the field in understanding where and how Wt1 functions and whether Lin28b could be a model for Wilms' tumor.

We agree that our data on Wt1 loss focusses on the role of Wt1 in normal kidney development, how its loss disrupts normal kidney development and how this could be important for Wilms' tumourigenesis. This includes but goes beyond being only relevant for the function of Wt1, it informs on the biology of WT1-mutant Wilms' tumours.

There is in our mind no doubt whether the LIN28B model is a model for Wilms' tumours. Activating mutations in LIN28B are found in human patient tumours, and already in the original publication of this model (Urbach et al. 2014) it was convincingly shown that the phenotype in the kidneys after less than 3 weeks (when the animals have to be culled animal welfare reasons) represents early stages of Wilms' tumours. Our data presented here confirms this, and extends it with respect to the behavior of the CSC markers and the comparison to the Wt1 loss phenotypes.

Whether the use of previously defined markers NCAM and Aldeflour serve the authors well or is a distraction is to be determine but it is unclear how useful these have been for understanding WT biology thus far. The authors describe these in the developing kidney in explants and in vivo.

*Overall, the data support the view that distinct mutations generate different forms of lineage derailment but it is unclear how this links to Wilms tumor. It is better suited to dsescribe the role of an interesting protein Wt1 in kidney development and lineages therein. Connecting it to tumor biology would require further scrutiny of tumors. *

Since CSCs are, according to the cancer stem cell model, the cells in a tumour that should preferentially be targeted, the exact identification of the CSC markers is directly important for the treatment of tumours. Our data analyzes two different sets of CSC markers, we show these cells label non-overlapping cell types in the normal kidney but that after mutation induction their expression changes and potentially become co-expressed in a single cell type (see our response to Reviewer #1 for more details on this). Identifying the developmental origins of CSCs in a tumour that is the direct result of disturbance of normal embryonic development (Hohenstein et al. 2015; Li et al. 2021) can be used as an entry point into understanding the biology of these tumours. Based on this we argue that although our analysis is on embryonic kidneys, their implications are highly relevant for the actual tumours and their treatment. We propose to further emphasize this in the Introduction and Discussion of our manuscript.

*The study shows that removal of Wt1 in the stromal compartment has distinct phenotypes, which could be important for Wilms tumor biology as this is an poorly understood part of this tumor. *

As already discussed in our response to Reviewer #1, we agree this is a potential important and poorly understood part of Wilms' tumours, directly for WT1 mutant tumours which are stromal-predominant, but potentially also for other tumours. We propose to further address this in the Discussion of the manuscript.

*Major comments: *

1. • This manuscripts uses elegant genetics to scrutinize the role of Wt1 and Lin28b. These stand out as difficult to conduct experiments and are of high value. *

We thank this Reviewer's appreciation for the design, challenges and value of our data.

In contrast, the section on ALDH1A2 and ALDEFLUOR activity is less integrated with the developmental framework.

We discussed our reasons for focusing on the normal developmental context of the cells expressing the CSC markers in the previous section. Since the originally described CSC marker was activity for the AldeFluor enzymatic assay (Pode-Shakked et al. 2013) which we could not use on sections or kidney rudiments, we had to conclusively identify which ALDH isozyme is responsible for this signal in this context. There is much inconsistency about this in the literature, and whichever isozyme is important in these tumours might not be the causative factor in other tumours where AldeFluor labels the CSCs. We therefore use previously published microarray data from the group that originally identified the NCAM1/ AldeFluor combination as Wilms' tumour CSCs to identify ALDH1A2 as the culprit in this cancer type. With this knowledge we could move our analyses to antibodies, allowing co-staining with the other markers. Note that if the signal in these CSCs would have been the result of ALDH1A1 or ALDH1A3 which we show are expressed in the developing ureteric bud, the implications of this for the biology of the tumours would be totally different. We propose to discuss this aspects and its importance in more detail in the revised manuscript.

*Much is unclear here e.g, antibody validation, rationale for performing these assays in explants rather than in vivo tissue, and the shift in Aldh1a2 staining pattern between E12.5 and E18.5, including reported nuclear localization.

*

We need to correct the reviewer on this remark, part of our data is using in vivo samples (E12.5 and E18.5) as well as cultured kidney rudiments. We will clarify which technique we use in the legends of the figures. We prefer to use this combination of techniques for several reasons: 1) the additional 3D information obtained from kidney rudiments can help with identifying specific developmental stages in the developing kidney; 2) due to the different fixation more antibodies work reliably in cultured rudiments than on paraffin frozen sections; 3) this is an important extra factor in the validation of antibodies; and 4) the possibility of culturing kidney rudiments on a time-lapse imaging system allows us to study phenotypes over time (this also greatly reduces the number of animals we need to study multiple timepoints in a developing system, an important aspect for the 3Rs). A good example of this in the timelapse data shown for the nephrogenic Wt1 knockout. The extreme outwards migration of the mutant cells (we show this using the tdTomato reporter) could only be identified in timelapse experiments, but is fully consistent with the sections of the corresponding E12.5 and E18.5 in vivo sections.

We have no explanation for the shift to nuclear localization for ALDH1A2. We are not aware of any other publications showing this. We cannot rule out this is a technical artifact but based on all other expression data obtained with this antibody and their consistency with other publication we don't think this is very likely.

*It is unclear how the manuscript is strengthened by this component. NCAM1 is referenced in the context of Wilms tumor CSCs, but unlike the rest of the manuscript which is mechanistic, it is unclear whether NCAM1 represents a mechanistic node in tumor initiation or merely a surface marker used for cell isolation? If NCAM1 functions just as a proxy for a progenitor-like state rather than a driver of tumor biology surely Wilms tumors will be full of progenitors or blastemal cells and many surface markers. It is unclear what strong evidence shows NCAM1 to be useful, this distinction should be stated. *

Cancer stem cells are defined based in functional characteristics, i.e. the capability of reconstituting a complete tumour with all of its complexity after transplantation in immune-compromised mice. The markers are usually, indeed, merely proxy markers for a specific cell type in the tumour with this functional capacity. The same can be said in this case for the AldeFluor activity, it is used as CSC marker for many cancer types but we are not aware of any data on a functional role for this pathway in any of them. It would be a really interesting experiment to combine our models with an additional conditional knockout for Ncam1 or Aldh1a2 to see if the phenotype we describe here changes. The genetics of such an experiment with so many alleles are however horrendous, would come with an enormous surplus of animals and would take too long for the average project.

The developmental framework presented argues that mutation-specific lineage derailment underlies tumor formation. Marker identity alone does not define pathogenesis. Perhaps reorganize this section to align it with the lineage-confusion model or removing it altogether would make the manuscript punchier?

• *

We propose to rewrite these parts to make this more clear.

• *

• The manuscript is highly focused on the nephrogenic compartment yet removes Wt1 from the stroma as part of one of the main lines of experiments. At several occasions, stromal changes are described qualitatively but using quantitative terms. As such, the manuscript currently comes across as having a bit of a black box where we cannot see the stroma beyond H&E stains. Could there additional antibody stains for stromal markers e.g., Pdgfra, Pdgfrb, or Meis1 to better visualize this compartment and perhaps enable quantification of changes?*

We agree this lack of additional stromal markers is a limitation of the current manuscript. Our reason for so far not including these was our doubts on the usefulness and relevance for the complete renal stroma of many commonly used markers. The scarceness of detailed studies on the developing stroma was a big part of this doubt. Some preliminary tests show that Meis1 is not exclusively found in the developing stroma of the mouse kidney but is also expressed in early stages of the nephrogenic lineage, and is therefore not a good marker for this purpose. Pdgfra and Pdgfrb however seem to be expressed throughout the complete stroma and not in the other lineages. __We propose to analyze these two additional markers for the revised manuscript. __

*Minor comments: *

*Page 4, Lines 89-95: Remove the repeated sentence beginning "Although best known as a transcription factor...". *

*Page 8, Line 164: Arrows referenced in Figure 1F are not visible. *

*Page 8, Lines 164-166: The sentence may refer to Figure 1G; this figure is not otherwise cited. *

*Page 18, Lines 413-414: (Pode-Shakked et al., 2013) is cited twice. *

*Figure 2C: Error bars are missing. Indicate number of biological replicates. *

*Gene nomenclature should be consistent throughout the manuscript. A mouse protein/gene is Six2/Six2. *

*Use precise language when referring to protein detection rather than "expression." *

*Standardize corticomedullary orientation across figures. *

*Page 7, Lines 160-161: Provide immunostaining supporting WT1+/Tdtomato− stromal identity. Co-staining with Foxd1 would clarify lineage assignment. *

*At E18.5 in the Six2-driven Wt1 mutant, WT1 signal is absent despite earlier stromal WT1+ cells. Clarify the fate of these cells. *

*Comment on the lower recombination efficiency observed in Wt1CE at E11.5. *

*Page 14, Lines 321-322: Determine how long CITED1 persists in WTCE mice. Co-staining with later differentiation markers would clarify whether progenitor retention coexists with nephron maturation. *

Page 15, Lines 352-353: Clarify whether the sentence describing blastemal-like regions should reference Figure 5D.

We thank the reviewer for these correction and other minor comments. We will address them in the revised manuscript. With respect to the remark regarding the gene nomenclature, until recently we were also under the assumption that mouse proteins only have the first character as capital. However, to our surprise we recently realized the official mouse nomenclature states that the protein (but not the gene) is in fact in all capitals. We refer for this to section 1.5.2 at https://www.informatics.jax.org/mgihome/nomen/gene.shtml.

References.

Berry RL, Ozdemir DD, Aronow B, Lindstrom NO, Dudnakova T, Thornburn A, Perry P, Baldock R, Armit C, Joshi A et al. 2015. Deducing the stage of origin of Wilms' tumours from a developmental series of Wt1-mutant mice. Dis Model Mech 8: 903-917.

Fukuzawa R, Breslow NE, Morison IM, Dwyer P, Kusafuka T, Kobayashi Y, Becroft DM, Beckwith JB, Perlman EJ, Reeve AE. 2004. Epigenetic differences between Wilms' tumours in white and east-Asian children. Lancet 363: 446-451.

Gadd S, Beezhold P, Jennings L, George D, Leuer K, Huang CC, Huff V, Tognon C, Sorensen PH, Triche T et al. 2012. Mediators of receptor tyrosine kinase activation in infantile fibrosarcoma: a Children's Oncology Group study. J Pathol 228: 119-130.

Guillaume R, Bressan M, Herzlinger D. 2009. Paraxial mesoderm contributes stromal cells to the developing kidney. Dev Biol 329: 169-175.

Haghighitalab A, Nosrati F, Dehghani-Ghobadi Z, Sayed M, Ahn C, Hu Y-C, Chung E, Lim H-W, Park J-S. 2026. A knock-in Six2Cre line reveals transient interstitial potential in nephron progenitors. bioRxiv: 2026.2002.2004.703893.

Hohenstein P, Pritchard-Jones K, Charlton J. 2015. The yin and yang of kidney development and Wilms' tumors. Genes Dev 29: 467-482.

Kobayashi A, Mugford JW, Krautzberger AM, Naiman N, Liao J, McMahon AP. 2014. Identification of a Multipotent Self-Renewing Stromal Progenitor Population during Mammalian Kidney Organogenesis. Stem Cell Reports 3: 650-662.

Krishna A, Meynert A, Dolt KS, Kelder M, Mesropian A, Ewing A, Brouwers C, Claassens JW, Linssen MM, Sheraz S et al. 2026. Mutational scanning reveals oncogenic CTNNB1 mutations have diverse effects on signaling. Nat Genet 58: 366-375.

Levinson RS, Batourina E, Choi C, Vorontchikhina M, Kitajewski J, Mendelsohn CL. 2005. Foxd1-dependent signals control cellularity in the renal capsule, a structure required for normal renal development. Development 132: 529-539.

Li CM, Kim CE, Margolin AA, Guo M, Zhu J, Mason JM, Hensle TW, Murty VV, Grundy PE, Fearon ER et al. 2004. CTNNB1 mutations and overexpression of Wnt/beta-catenin target genes in WT1-mutant Wilms' tumors. Am J Pathol 165: 1943-1953.

Li H, Hohenstein P, Kuure S. 2021. Embryonic Kidney Development, Stem Cells and the Origin of Wilms Tumor. Genes (Basel) 12.

Magella B, Adam M, Potter AS, Venkatasubramanian M, Chetal K, Hay SB, Salomonis N, Potter SS. 2018. Cross-platform single cell analysis of kidney development shows stromal cells express Gdnf. Dev Biol 434: 36-47.

Miyagawa K, Kent J, Moore A, Charlieu JP, Little MH, Williamson KA, Kelsey A, Brown KW, Hassam S, Briner J et al. 1998. Loss of WT1 function leads to ectopic myogenesis in Wilms' tumour. Nat Genet 18: 15-17.

Pode-Shakked N, Shukrun R, Mark-Danieli M, Tsvetkov P, Bahar S, Pri-Chen S, Goldstein RS, Rom-Gross E, Mor Y, Fridman E et al. 2013. The isolation and characterization of renal cancer initiating cells from human Wilms' tumour xenografts unveils new therapeutic targets. EMBO Mol Med 5: 18-37.

Schumacher V, Schuhen S, Sonner S, Weirich A, Leuschner I, Harms D, Licht J, Roberts S, Royer-Pokora B. 2003. Two molecular subgroups of Wilms' tumors with or without WT1 mutations. Clin Cancer Res 9: 2005-2014.

Urbach A, Yermalovich A, Zhang J, Spina CS, Zhu H, Perez-Atayde AR, Shukrun R, Charlton J, Sebire N, Mifsud W et al. 2014. Lin28 sustains early renal progenitors and induces Wilms tumor. Genes Dev 28: 971-982.

Uschkereit C, Perez N, de Torres C, Kuff M, Mora J, Royer-Pokora B. 2007. Different CTNNB1 mutations as molecular genetic proof for the independent origin of four Wilms tumours in a patient with a novel germ line WT1 mutation. J Med Genet 44: 393-396.

Zirn B, Samans B, Wittmann S, Pietsch T, Leuschner I, Graf N, Gessler M. 2006. Target genes of the WNT/beta-catenin pathway in Wilms tumors. Genes Chromosomes Cancer 45: 565-574.

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

Wilms tumor arises from disrupted kidney development. Progenitor-like populations and cancer stem cell (CSC) fractions have been described in patient tumors, but how specific mutations alter embryonic programs to generate these states remains unresolved. Pop et al. model genotype-phenotype relationships during kidney embryogenesis. Using Six2- and Foxd1-driven Cre lines, they test the effects of Wt1 loss-of-function and Lin28b gain-of-function in nephron and stromal progenitors. Through explant imaging, histology, and immunofluorescence, they define mutation-specific effects on ureteric branching, cap mesenchyme organization, stromal composition, and nephron differentiation. Lineage-restricted Wt1 deletion produces distinct outcomes depending on whether nephron progenitors, stromal progenitors, or both are targeted. Lin28b overexpression causes delayed nephrogenesis and lobular organization resembling human Wilms tumor morphology, with expansion of blastemal-like populations. These genetic removals of Wt1 and overexpression of Lin28b are useful for the field in understanding where and how Wt1 functions and whether Lin28b could be a model for Wilms' tumor. Whether the use of previously defined markers NCAM and Aldeflour serve the authors well or is a distraction is to be determine but it is unclear how useful these have been for understanding WT biology thus far. The authors describe these in the developing kidney in explants and in vivo.

Overall, the data support the view that distinct mutations generate different forms of lineage derailment but it is unclear how this links to Wilms tumor. It is better suited to dsescribe the role of an interesting protein Wt1 in kidney development and lineages therein. Connecting it to tumor biology would require further scrutiny of tumors. The study shows that removal of Wt1 in the stromal compartment has distinct phenotypes, which could be important for Wilms tumor biology as this is an poorly understood part of this tumor. My recommendation is that the manuscript would be considered for a major revision where it is more focused on kidney biology.

Major comments:

1. This manuscripts uses elegant genetics to scrutinize the role of Wt1 and Lin28b. These stand out as difficult to conduct experiments and are of high value. In contrast, the section on ALDH1A2 and ALDEFLUOR activity is less integrated with the developmental framework. Much is unclear here e.g, antibody validation, rationale for performing these assays in explants rather than in vivo tissue, and the shift in Aldh1a2 staining pattern between E12.5 and E18.5, including reported nuclear localization. It is unclear how the manuscript is strengthened by this component. NCAM1 is referenced in the context of Wilms tumor CSCs, but unlike the rest of the manuscript which is mechanistic, it is unclear whether NCAM1 represents a mechanistic node in tumor initiation or merely a surface marker used for cell isolation? If NCAM1 functions just as a proxy for a progenitor-like state rather than a driver of tumor biology surely Wilms tumors will be full of progenitors or blastemal cells and many surface markers. It is unclear what strong evidence shows NCAM1 to be useful, this distinction should be stated. The developmental framework presented argues that mutation-specific lineage derailment underlies tumor formation. Marker identity alone does not define pathogenesis. Perhaps reorganize this section to align it with the lineage-confusion model or removing it altogether would make the manuscript punchier?
2. The manuscript is highly focused on the nephrogenic compartment yet removes Wt1 from the stroma as part of one of the main lines of experiments. At several occasions, stromal changes are described qualitatively but using quantitative terms. As such, the manuscript currently comes across as having a bit of a black box where we cannot see the stroma beyond H&E stains. Could there additional antibody stains for stromal markers e.g., Pdgfra, Pdgfrb, or Meis1 to better visualize this compartment and perhaps enable quantification of changes?

Minor comments:

Page 4, Lines 89-95: Remove the repeated sentence beginning "Although best known as a transcription factor...".

Page 8, Line 164: Arrows referenced in Figure 1F are not visible.

Page 8, Lines 164-166: The sentence may refer to Figure 1G; this figure is not otherwise cited.

Page 18, Lines 413-414: (Pode-Shakked et al., 2013) is cited twice.

Figure 2C: Error bars are missing. Indicate number of biological replicates.

Gene nomenclature should be consistent throughout the manuscript. A mouse protein/gene is Six2/Six2.

Use precise language when referring to protein detection rather than "expression."

Standardize corticomedullary orientation across figures.

Page 7, Lines 160-161: Provide immunostaining supporting WT1+/Tdtomato− stromal identity. Co-staining with Foxd1 would clarify lineage assignment.

At E18.5 in the Six2-driven Wt1 mutant, WT1 signal is absent despite earlier stromal WT1+ cells. Clarify the fate of these cells.

Comment on the lower recombination efficiency observed in Wt1CE at E11.5.

Page 14, Lines 321-322: Determine how long CITED1 persists in WTCE mice. Co-staining with later differentiation markers would clarify whether progenitor retention coexists with nephron maturation.

Page 15, Lines 352-353: Clarify whether the sentence describing blastemal-like regions should reference Figure 5D.

Significance

Overall, the data support the view that distinct mutations generate different forms of lineage derailment but it is unclear how this links to Wilms tumor. It is better suited to dsescribe the role of an interesting protein Wt1 in kidney development and lineages therein. Connecting it to tumor biology would require further scrutiny of tumors. The study shows that removal of Wt1 in the stromal compartment has distinct phenotypes, which could be important for Wilms tumor biology as this is an poorly understood part of this tumor.

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

Wilm's Tumor, a pediatric kidney cancer, is associated with gain or loss of activity of a number of gene including the loss of activity of the nucleic acid binding protein WT1 and gain of activity (enhanced expression at the mRNA level) of the RNA binding protein Lin28 which negatively impacts the maturation of the miicroRNA let-7, elevating levels of let-7 targets. Previous mouse studies have examined the impact of loss of Wt1 throughout within the nephron progenitor and interstitial cell compartments in capping mesenchyme that is thought to be the source of the tumor and of broad elevated expression in all kidney progenitors.

In this manuscript, the authors have refined the loss of Wt1 to nephron or stromal progenitors and compared the phenotype to loss of Wt1 in both lineages examining cultured kidneys over a 72 hr period, in addition to uncultured kidneys examined at e18.5. A similar analysis was performed on Lin28 mutants. The analysis itself consisted of video imaging, limited immunostaining and histochemistry.

While wholly qualitative and largely observational and descriptive, the limited data are of good quality and the conclusions drawn are reasonable. For the Wt1 study, most interesting would be in the loss of Wt1 from the NPC lineage. Clearly, there is already a significant phenotype at the time of study (E12.5) hence there is no strong insight into the earliest effects of Wt1 loss and how this might contribute to tumor formation. Quite what happens to these cells phenotypically is unclear given the limited set of markers used to look at the cells. Specific removal of Wt1 from the stromal lineage generates a milder phenotype, indicating a role for Wt1 there, but without a mechanistic analysis of the resultant products, the underlying mechanisms remain unclear. Wt1 removal from both lineages generated a phenotype less severe than removal from nephron progenitors (and previous data on "double lineage removal" with a Nestin1 cre), an indication that the genetic approach was not up to the task.

In some sense, one could regard this work as a pilot study, looking to optimize expensive and time-consuming mouse experiments to maximize insight (ie choose optimum model, address most informative time points, decide on analytical approaches). As a stand-alone paper, the work may not significantly advance our understanding of the topic. For example, can simple loss of Wt1 tells us anything about Wt? Yes Wt1 is lost in a subset, but even in these there are additional genetic mutations. For Lin12, there is no significant advance beyond the studies of the Daly lab.

I have no useful suggestions for improvement which would require a completely different approach to the problem from the start.

Significance

The authors set out the goal in the introduction - to obtain a better understanding of the origins of Wilm's tumor. There doesn't appear to be an insight of cancer relevant significance beyond earlier studies. To a readership now/too used to analysis at genome scales (genomic, transcription), this study might appear modest. The target audience is unclear.

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 addresses an important gap in Wilms tumor (WT) biology: what are the earliest pathogenic events following WT driver mutation induction, and how do these early developmental trajectories differ across genotypes? The authors provide a carefully staged and comparative analysis of two WT-associated genetic contexts-conditional Wt1 loss (using lineage-specific Cre drivers targeting nephrogenic (Six2-Cre) versus stromal (Foxd1-Cre) compartments, as well as a temporally controlled Wt1CreERT2 model targeting both lineages upon tamoxifen induction) and inducible LIN28B overexpression, and relate the resulting developmental phenotypes to two CSC marker paradigms derived from patient-based studies.

A major strength is the precise, time-resolved description of the earliest initiating phenotypes (E12.5 and E18.5, with additional postnatal analysis for LIN28B) and the direct side-by-side comparison of how each genotype perturbs nephrogenesis. The authors conclude that Wt1 loss (especially in the nephrogenic lineage) leads to a severe developmental block accompanied by a disturbance of lineage identity ("lineage confusion"), whereas LIN28B overexpression causes a disturbed transition between uninduced and induced nephron progenitor cell (NPC) states, producing blastemal-like regions that persist postnatally. Using immunostaining for NCAM1, SIX2, CITED1, and ALDH1A2, the authors map marker combinations during normal kidney development and across mutant contexts, and propose that tumor-initiating alterations, most clearly in the LIN28B model, and more suggestively in the Wt1CreERT2 (Wt1CE) context, promote the emergence of a CSC-like population inferred to co-express all four markers (NCAM1+SIX2+CITED1+ALDH1A2+), a state not observed in normal kidneys.

Overall, this study provides a particularly clear direct comparison of the earliest tumor-initiating events triggered by distinct WT-relevant driver alterations. While the manuscript does not yet offer a detailed molecular mechanistic framework explaining why these two mutations produce such divergent developmental and marker-state outcomes (which would further strengthen the work), the careful comparison and the conclusions drawn from it are meaningful and make an important contribution to our understanding of the developmental processes that can lead to Wilms tumor initiation.

Major comment:

1. A central and highly emphasized conclusion of this manuscript is that tumor-initiating alterations induce a CSC-like population co-expressing all four markers (NCAM1, SIX2, CITED1, and ALDH1A2), and that this state is not observed during normal kidney development. Because this "quadruple-positive" population is a key mechanistic take-home message and closely linked to the overall conceptual model, the manuscript would be substantially strengthened by a direct, same-cell demonstration of co-expression of all four markers, rather than inference from consecutive sections. The authors state that they were unable to do so due to a technical limitation, namely, antibody host-species constraints that prevent co-detection of CITED1 and ALDH1A2 within the same section. Several feasible approaches could address this limitation for example:
   • Identify an alternative antibody reagent from a different host species.
   • RNAscope / smFISH for in situ single-cell co-detection.
   • Single-cell RNA-seq (scRNA-seq) to test whether a bona fide quadruple-positive transcriptional state exists.

Overall, resolving this technical limitation would markedly increase confidence in one of the manuscript's most important claims and strengthen the proposed genotype-phenotype/CSC-marker framework 2. It is somewhat unexpected that the Six2-specific Wt1 deletion appears to produce a more severe phenotype than the tamoxifen-inducible Wt1CreERT2 approach, which is intended to target a broader Wt1-derived lineage (both nephrogenic and stromal). The Discussion offers several plausible, non-mutually exclusive explanations for this observation (e.g., timing, recombination efficiency/mosaicism, and the rescue contribution of "escaping" wild-type cells). It would be helpful to support at least one of these explanations experimentally. For example, the authors could quantify the extent of "escape" (percentage of non-recombined cells within the lineage) across embryos/timepoints to validate that mosaicism is indeed the cause of the milder phenotype.

Minor comments

1. Please clarify whether the difference shown in Fig. 2C is statistically significant, and report n, error bars/variation, the statistical test used, and p-values (if applicable).
2. The authors note the presence of some SIX2+; tdTomato+ cells in Foxd1GC control kidneys. Given the expected stromal restriction of Foxd1 lineage labeling, please clarify the likely explanation and, if possible, indicate how frequent this is.
3. It is somewhat unexpected that Six2-specific Wt1 deletion appears to produce a more severe phenotype than the tamoxifen-inducible Wt1CreERT2 approach, which is intended to target a broader Wt1-derived lineage. The Discussion offers several plausible, non-mutually exclusive explanations (e.g., timing and/or recombination efficiency/mosaicism and the contribution of "escaping" cells). it would be helpful to support at least one of these explanations experimentally, for example by quantifying the extent of "escape" across embryos/timepoints and tamoxifen dosing.
4. A careful proofreading pass is needed to ensure text-figure consistency, particularly for arrow annotations. For example, the Results text refers to "Fig. 1F, arrows," but arrows are not apparent in that panel. Likewise, the Results text mentions a "white filled arrow" in Fig. 2H, whereas the figure appears to show only open arrows. Please align the wording with the annotations actually shown in the figures.

Significance

Overall, this study provides a particularly clear direct comparison of the earliest tumor-initiating events triggered by distinct WT-relevant driver alterations. While the manuscript does not yet offer a detailed molecular mechanistic framework explaining why these two mutations produce such divergent developmental and marker-state outcomes (which would further strengthen the work), the careful comparison and the conclusions drawn from it are meaningful and make an important contribution to our understanding of the developmental processes that can lead to Wilms tumor initiation.

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

Summary:

Building a functional nervous system with the correct number of correctly specified and wired pre- and post-synaptic neurons requires a tight control of proliferation and cell survival during development. Focusing on the latter, Naser Alshami and colleagues in the laboratory of Alicia Hidalgo investigated the role of three neurotrophins and their respective Toll receptors in regulating neuronal survival in the visual system of Drosophila. A combination of expression and genetic loss-of-function and gain-of-function analyses revealed a novel role for the neurotrophin DNT2 (spaetzle-5) in controlling survival, dendritic spine formation, columnar axonal innervation of lamina neuron L1 expressing the receptor Toll-2.

Major comments:

• DNT-2 (spz-5) is described to be expressed in non-neuronal cells in the retina. However, the clustered arrangement of cells in cross-sections of the retina during pupal development (72h DNT-2 bottom panel) suggests that photoreceptors could be labeled by the reporter as well. This should be assessed by co-labeling experiments with Elav and/or mAb24B10. This is important, as photoreceptors may represent a presynaptic additional source of this neurotrophin to lamina neurons L1 (see below).
• The most right hand panel of Figure 2A, first row, Toll-2, does not resemble a 24h APF optic lobe, and would need to be replaced by a correctly staged younger sample. This is relevant, as the image at 48 h APF shows extensive staining in a cluster at the medulla edge, suggesting that expression in medulla neurons may be more abundant at younger stages. The expression of Toll-2 in L3 neurons is not unambiguously demonstrated at the angle shown in Figure 1E first panel. Normally, MCFO clones should be able to reveal single cells instead of L1 and L3 together.
• Related to this part of the study, the authors describe that Toll-2 is expressed by connecting neurons (L1, Mi1, Tm3...). This is solely based on RNAseq data. In the next paragraph, the authors describe that Toll-2 is expressed in L1 and DNT-2 in Mi1 (or related) medulla neurons. However, this is rather vague, as it is not clear what "related" medulla neurons refers to. As expression in Mi1 neurons is at the heart of this study, it would be essential to verify that Mi1 is included in the Toll-2-Gal4 and DNT-2 expression pattern. Indeed, in Figure 1A (right panel, row 2, 48 h APF) expression of DNT-2 in this neuron subtype may be visible. This should be verified by markers and single cell labeling such as MCFO. Finally, the authors mention that the DNT-2 ligand could "reach" the Toll-2 receptor. Here some more precision would be helpful, as it is not clear, how far neurotrophins would need to travel, and synaptic contacts may form too late for mediating survival. Furthermore, there seems to be also the possibility that there is a substantial contribution of autocrine signaling, if L1 neurons indeed express both the ligand and the receptor, which would need to be presented more clearly.
• The next set of experiments uses over-expression of DNT-2 and DNT-3 in Toll-8 expressing neurons and assessment of survival rates in lamina neurons. Survival is determined by evaluating Dcp-1 signals rather than cell numbers. During normal larval development, 7 future lamina neurons per column are formed, two of which are eliminated by apoptosis. Thus, it would be important to verify whether it is these neurons that are surviving, using for instance Elav labeling; moreover lamina neuron specific markers (see Xu et al., 2024) may enable assessment whether the surviving neurons adopt specific lamina neuron identities in excess to the normal 5. It may also need to be considered that both neurotrophins could affect proliferation, if the numbers do not add up (as drivers may be active already during the third instar larval stage, this is possible).
• Gain-of-function experiments are followed by loss-of-function experiments, assessing the impact of loss of DNTs in entire animal mutants on lamina neuron survival. As for the gain-of-function studies, it would be important to know, which lamina neuron subtypes are dying in the absence of DNT-2 and DNT-3. Indeed, Figure 4C reveals a substantial loss of L1 and L3 neurons. The over-expression experiments of the ligands using the Toll-2 driver are complicating matters, considering that the driver line is described to be a mutant in Toll-2, and the ligand is expressed in the same neuron, creating an autocrine signaling situation. Moreover, since Toll-2-Gal4 is used to drive expression in all lamina neurons L1 and L3, and entire optic lobes are assessed, the numbers should be around 1600 in wild type, but are not, suggesting that there might have been some cell death in controls. If there are more lamina neurons in the over-expression situation, it would mean that additional neurons are adopting L1 and L3 identity and thus reveal an additional direct role in cell fate determination by DNT-2 and Toll-2 or an indirect role considering the intricate interactions of lamina neurons to adopt specific identities via N signaling (see Xu et al. 2024). This would need to be assessed more in detail using for instance independent drivers or markers for lamina neuron subtypes.
• The statement, that the remaining signal is coming from macrophages would need to supported by additional markers or described more carefully (as it could be glia).
• Page. 9. The study reports a new interaction between DNT-2 and Toll-2 as its possible receptor. This is in part based on whole animal lethality, without providing quantification. This should be added.
• Interestingly, knock-down of Toll-2 in all neurons led to increased cell death of lamina neurons, which cannot be overcome by over-expression of DNT-2 [FL]. It would be interesting to assess whether cell death is even higher by providing a statistical test comparing the knock-down of Toll-2 with the knock-down of Toll-2 and the simultanous over-expression of DNT-2 in Figure 4B.
• Page 11 (also page 4). The authors describe that "connectivity of lamina neurons to medulla neurons takes place at 30-48 h APF and between medulla and lobula complex at 60-70h APF, and that expression of synaptic markers starts at 24h APF". It is not clear how the authors define connectivity (here and throughout the manuscript). While axonal and dendritic projections are established during the first half of pupal development, functional synapses are thought to be solely established from mid-pupal development onwards (early expression of synaptic markers may not be indicative of synapse formation as early as 24h APF). Moreover, there is no solid evidence yet that connections form sequentially between the lamina and medulla, and medulla and lobula and lobula plate neurons. This description would need to be adjusted. A reference for the occurrence of spontaneous activity (prior to synapse formation) should be provided: Akin et al. 2019. This is important for the subsequent interpretation that cell death overlaps with "connectivity": if this term is used to refer to synapse formation, this may not hold, and need adjustment.
• Next, the study assesses the impact of loss of Toll-2 in surviving lamina neurons L1 on columnar axonal branching. The authors observed that loss of Toll-2 or over-expression of DNT-2 leads to extension of axonal branches into neighboring columns in the medulla neuropil layer M1. Here, it would be important to assess, whether this phenotype is really due to the loss of Toll-2 or the fact that neighboring L1 neurons are missing (even just one neighbor missing may be enough to create this defect as contact-dependent repulsion may no longer work, Millard et al. 2007). Toll-2 MARCM clones would be able to address this point. Without further experiments conclusions need adjustments. Finally, the authors conclude that this is an indication for a role in connectivity, but indeed, it is not clear whether these projections lead to abnormal connections, as the thin projections may not form synapses. This statement would need to be adjusted or supported by including synaptic markers, such as Brp/or TransTango experiments.
• The study concludes with the statement that over-expression of DNT-2 FL and knock-down of Toll-2 alters dendritic spine formation of lamina neurons L1 in the lamina, it increases by over-expression and decreases in Toll-2 knock-down. This is in contrast to the axonal projections, where additional extensions are observed in both genotypes. This should be discussed, as the opposite effects may be in line with an indirect effect through loss of neighbors in medulla columns (in contrast to the lamina).
• The authors conclude in the discussion (page 15 and 17) that DNT-2 by medulla neurons Mi1 or related medulla neurons is key to promoting survival of lamina neurons L1, expressing the Toll-2 receptor. However, the manuscript does not provide any direct evidence that it is indeed Mi1 (or other medulla neurons) playing this role. Considering that DNT-2 may also be provided by photoreceptors (see above), this statement may need support by additional experiments, such as a Mi1 specific or photoreceptor-specific over-expression of DNT-2 in a DNT-2 whole animal mutant background and assessment of survival of labeled L1 neurons (e.g. using the markers Svp/Zfh1 (Xu et al. 2024). Moreover, it would be crucial to indeed determine the time point of when cells are undergoing apoptosis and to assess whether this would coincide with synaptogenesis. Such experiments would allow to comment on a potential retrograde signaling process between connecting neurons.
• The data and molecular and genetic methods are presented in detail in the Material and Method sections, as well as through pertinent supplementary tables. Similarly, the statistical analyses and sample numbers are indicated in the Material and Methods section and related supplementary tables. Table S4 does not indicate genotypes, but short versions of crosses, which should be corrected. Moreover, the list of fly lines seems to miss one line (UAS-mCherry). Concerning the antibody list, sources are not provided (unlike stated in the main text). In the Result section, differences are described simply as statistically significant decreases and increases, however, at times it would be useful to add some indication about the level of differences (percentage or fold times).

Minor comments:

• The figures are informative, however, their orientations are not as indicated horizontal or lateral, but often oblique and not consistent (e.g. Fig 1a, last four panels). This makes comparisons more difficult. The authors could possibly find more suitable optical sections and present images more consistently.
• Please provide the correct reference describing the occurrence of spontaneous activity (Akin et al. 2019).
• The manuscript would be more easily accessible to non-experts, if schematic drawings would be provided (neuron subtypes, model).
• The final paragraph of the discussion about DNT-2 keeping connecting neurons together is not easily understandable, as the term "keeping together" remains undefined.

Significance

As in the vertebrate nervous system, insect nervous systems rely on neurotrophins to control the survival and correct wiring of neurons. This study proposes a model, that advances our understanding as to how a specific neurotrophin, delivered by a post-synaptic neuron, acting via a specific receptor, controls several developmental steps of one of its pre-synaptic partners. The study assessed the expression patterns of three neurotrophins and the Toll receptor family, using the T2A-Gal4 expression system, revealing their dynamic expression in photoreceptors, specific lamina and medulla neurons, as well as tracheae and glia during pupal development. A strength of the study represents the identification of DNT-2 and Toll-2 as interacting partners, which could be assessed by the extensive set of specifically generated reagents for this study (Gal4 drivers and mutants). However, a number of statements may require additional experiments to conclusively support the proposed model, in particular the action of DNT-2 and Toll-2 in Mi1 and L1, respectively. Moreover, it would be important to critically assess the timing of events of cell death, setting up of projection patterns and synapse formation.

Audience: The outlined findings, if strengthened further, will be of interest for scientists studying nervous system development in vertebrates and invertebrates in general, and the action of neurotrophins in particular.

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 investigate the role of neurotrophins in development of the Drosophila visual system, where multiple waves of cell death eliminate populations of neurons. Using transgenic reporter lines, the authors examine expression of four spz genes, which encode ligands and four Toll receptor genes in the optic lobe. The authors argue that two of these ligands, spz-3/DNT-3 and spz-5/DNT-2, are expressed in the optic lobe concurrently with a wave of cell death. Through gain- and loss of function studies they argue that DNT-3 promotes survival through Toll-8 whilst DNT-2 promotes survival through novel interactions with Toll-2 (with a focus on the lamina neurons). Finally, the authors argue that DNT-2 and Toll-2 regulate L1 neuron targeting and size.

The involvement of neurotrophins in neuronal apoptosis has not been explored in detail in the visual system, making this a potentially interesting topic. However, there are several issues with the manuscript: The primary concern is that most of the authors' conclusions are not supported by the data presented. Many of the figures are difficult to interpret without the addition of appropriate markers. In addition, there are several inconsistencies both within the manuscript and within the context of existing literature that are not addressed. For example, the authors have not demonstrated that lamina neurons undergo apoptosis. This point is critical, as apoptosis, particularly that likely to be affected by neurotrophic factors, fine-tunes multicolumnar neuron numbers after they establish connections with their partners. On the other hand, unicolumnar neurons such as those in the lamina are not thought to need much fine-tuning because of their unicolumnar status. This distinction is not addressed in the manuscript. At present, the evidence presented does not convincingly support the authors' conclusions, and substantial further work is needed before the claims can be firmly established. Overall, these limit the credibility and significance of this study.

Major comments:

In general, the conclusions drawn about neuropil regions and specific cell identities are not supported by sufficient evidence. Across the figures, the identity of neurons and neuropils remains unclear due to the absence of subtype-specific markers or neuropil labelling that would place the data in the anatomical context of the optic lobe.

(1) The authors claim that spz and Toll genes are differentially expressed in the optic lobe throughout pupal development. However, the data presented do not convincingly support this conclusion. The primary concern is that the Gal4 reporter insertions used to infer gene expression do not match published single-cell transcriptomic datasets (as noted by the authors). This discrepancy calls into question the accuracy of these reporters (or the transcriptomics) and raises uncertainty about which dataset can be trusted. Given that these data are inconsistent, gene expression should be assessed directly by HCR to draw reliable conclusions.

Related comments: The authors claim that spz and Toll genes are expressed differentially in the optic lobe throughout pupal development. However, the characterisation of these expression patterns is lacking in the following aspects: - Time courses are inconsistent and incomplete across the different genes. - The conclusion that DNT-2/3 are expressed concurrently with cell death lacks evidence is not supported: DNT2/3 expression does not appear restricted to a clear time window and without co-staining for apoptotic markers this conclusion cannot be substantiated. - There is insufficient evidence to support the claim that spz-2 is specifically expressed in Mi1 neurons or that Toll-8 is specific to L4/L2. These claims are largely based on the transcriptomics data, which were inconsistent with the reporters in many other contexts. There are no neuron type markers costained to validate this claim. - Sample sizes are too low for some reporter lines (~2-4).

(2) The authors quantified cell death using Dcp-1 signal volume and concluded that overexpression of DNT-2 or DNT-3 reduces cell death, while loss of either increases cell death in the lamina. However, the lamina is not labelled with an appropriate marker, nor are neuropils labelled to make the structure recognisable making it unclear whether this wave of cell death is truly occurring in the lamina. This is a recurring issue throughout the manuscript. Related comments: the reported values for Dcp-1 signal volume differ dramatically between the Gal4 (~40,000) and mutant (~2,000) controls, without an explanation for this discrepancy.

(3) The claim that DNT-2 and DNT-3 promote survival via Toll-2 and Toll-8, respectively, is not sufficiently supported. The authors do not show that manipulation of each ligand specifically affects neurons expressing the relevant receptor. To substantiate this conclusion, the specificity of ligand manipulation should be tested. For example, through compensation experiments or by demonstrating that DNT-2 loss of function specifically induces cell death in Toll-2+ neurons.

(4) The authors further investigate a potential interaction between DNT-2 and Toll-2, concluding that DNT-2 promotes survival in the lamina, medulla, and lobula complex through Toll-2. However, no evidence is presented showing altered levels of apoptosis in the medulla or lobula/lobula plate following DNT-2 manipulation. The conclusion that DNT-2 promotes survival via Toll-2 is based on a Toll-2RNAi;DNT-2FL epistasis experiment, but this interpretation does not account for the reduced expression of each construct due to Gal4 titration. This issue also applies to subsequent experiments (Figs. 5 and 6). A Gal4 titration control is required for each of these experiments to exclude this confound.

(5) The authors conclude that DNT-2 production in Mi1 medulla neurons is required for L1 connectivity, survival, and morphology. However, these conclusions are based on manipulating DNT-2 and Toll-2 expression in L1 neurons, which does not directly test the requirement for DNT-2 in Mi1 neurons. In addition, the image quality in Figure 5 is problematic: brightness and contrast differ visibly between panels, making them challenging to interpret. Finally, the discussion states that "loss of function for DNT-3 (spz-3) causes lamina cell death that is not naturally compensated for by DNT-2," but no evidence is presented in this manuscript to support that conclusion.

Other general comments:

• Lack of labels for each individual image makes figures and text difficult to interpret/reference.
• Gene names not italicised in figures
• Inconsistent aspect ratio for graph axes (Obvious in figure 5C,D)
• Image borders often not completely vertical/horizontal making them appear jagged in some figures e.g. Figure1A.
• General typos e.g. (n=10 10 brains), inhibitor pf apoptosis p35
• Lobula complex should not be used to refer to lobula plate and lobula plug at this stage as they are distinct neuropils

Figure 1:

• Arrows not explained
• N-Cadherin absent from some images without explanation
• First two images of 1B are identical.
• Images in panel B at different timepoints and spz-4 is missing without explanation.

Figure 3:

• Dark spot in DNT-3FL 3A image.
• Descriptions in text often lack specific details "Dcp1+ apoptosis in the lamina and outside the lamina too." Supplementary:
• Figures S2-S6 lack labels denoting the timepoint of each figure
• Figure S7 is illegible
• Supplementary figure 8 is referred to as S9 in text and is missing genotype information as well as neuropil labels on images

Significance

Significance: This is an interesting topic of investigation, and the findings-if substantiated-would be of interest to the field of Drosophila visual system development and programmed cell death. In particular, it would be valuable to place these results in the broader context of programmed cell death as a mechanism for eliminating excess multicolumnar neurons during visual system development. However, at present the evidence presented does not convincingly support the authors' conclusions, and substantial further work is needed before the claims can be firmly established.

Expertise:

Drosophila neurogenesis, neuron specification, cell signalling

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

Alshamsi et al. investigate the role of Drosophila neurotrophins (DNTs) and their Toll receptors in regulating neuronal apoptosis during optic lobe development.

The authors provide compelling evidence that different DNTs and their Toll receptors are expressed in optic lobe neurons, and their activity regulates neuronal survival during optic lobe development. They further show that disruption of DNT/Toll signaling impacts neuronal morphologies.

Comments

ABSTRACT

"Over-expression of DNT-3 (spz-3) and DNT-2 (spz-5) could rescue natural occurring cell death, whereas their loss of function caused cell death, showing that DNT-3 and DNT-2 can, and are required to, promote cell survival during optic lobe development."

I find it more appropriate to say that OE prevents naturally occurring cell death because it inhibits a normal physiological process."rescue" would only be correct if there were an experimental or genetic loss (e.g., deletion of a survival factor) and you are restoring normal survival levels

"Importantly, DNT-2 is expressed in Mi1 neurons and Toll-2 in connecting L1 neurons. We show that DNT-2 functions in concert with Toll-2, as Toll-2 RNAi knock-down prevented the rescue of apoptosis by DNT-2 over-expression and all Toll-2+ neurons were lost in DNT-2 mutants."

I find this sentence very difficult to follow

i suggest moving "Importantly, DNT-2 is expressed in Mi1 neurons and Toll-2 in connecting L1 neurons."

to the next sentence. "by specifically investigating the Mi1 (DNT-2+) and L1 (Toll-2) synaptic partners", alterations in DNT-2 or Toll-2 expression levels impaired connectivity of L1 neurons at the M1 medulla layer and altered dendritic morphology of L1 neurons

"As DNT-3 (spz-3) and DNT-2 (spz-5) are expressed in the medulla and they could influence both lamina and medulla neurons, this suggests that their function maintaining cell survival could enable the stabilisation or alignment of connected neurons across medulla columns."

influence what? this is very vague and needs a temporal understanding of when neurons die, synapses are formed, and consider the phenotypes of the mutants and RNAi experiments.

INTRODUCTION

"Neuronal survival is maintained by neurotrophic factors secreted in limited amounts by target cells, leading to the survival of only those neurons that receive trophic support (Levi-Montalcini, 1987, Davies, 2003)."

Perhaps the authors can be more precise, e.g. One mechanism by which Neuronal survival is regulated is through neurotrophic factors secreted in limited amounts by to be synaptic partners and adjacent cells

"In this context, if neurotrophism is fundamental for nervous system development, it could have been enabled by evolutionarily conserved molecular mechanisms."

I think the authors want to suggest that given the fundamental and widespread role of neurotrophism in nervous system development, it remains unknown if it relies on evolutionarily conserved molecular players.

"Neurotrophins - NGF, BDNF, NT3, NT4 - are the main growth factors maintaining neuronal survival in the vertebrate nervous system (Levi-Montalcini, 1987, Lu et al., 2005). Importantly, they can also promote cell death, depending on context (Lu et al., 2005). They can promote cell survival via their Trk receptors and ERK and AKT downstream and via p75NTR and NFB downstream, or cell death via p75NTR, Sortilin, and JNK signalling instead (Lu et al., 2005). mechanisms."

The authors should avoid the repeated use of "they"

"There are six spz and nine Toll paralogous genes in Drosophila, which could play distinct functions. In fact, at least full-length DNT-1 and Toll-1 can promote cell death instead, and at least Toll-6 can promote either cell survival or cell death, depending on context (Foldi et al., 2017, Singh et al., 2025, Zhu et al., 2008) . Importantly, mature DNT-1 and DNT-2 with Toll-6 and Toll-7 are required for and can promote neuronal survival during circuit formation in the embryonic ventral nerve cord (McIlroy et al., 2013, Zhu et al., 2008)."

I find this paragraph difficult to follow. I suggest the following editing, which the authors might want to consider:

"There are six spz and nine Toll paralogous genes in Drosophila, which could play distinct functions. In fact, at least full-length DNT-1 and Toll-1 can promote cell death instead, while and at least Toll-6 can promote either cell survival or cell death, depending on context (Foldi et al., 2017, Singh et al., 2025, Zhu et al., 2008) . Importantly, mature DNT-1 and DNT-2 with Toll-6 and Toll-7 are required necessary and sufficient to promote for and can promote neuronal survival during circuit formation in the embryonic ventral nerve cord (McIlroy et al., 2013, Zhu et al., 2008).

"During this time (24-50h APF), connectivity between photoreceptors, lamina and medulla neurons is established; this is followed by medulla neurons connecting to lobula neurons;"

I find this sentence misleading, if not incorrect. If by connectivity the authors mean synaptogenesis, for all that is known, synaptogenesis has been shown to occur from from mid-pupal development (P50) onwards If by connectivity, the authors mean the targeting of specific neuropiles and layer organization, it is also incorrect that lamina and medulla organization precedes the connectivity between medulla and lobula neurons. These processes are all concurrent. Can the authors please clarify?

"and by 72h APF cell death has greatly diminished and synaptogenesis completes connectivity patterns, in preparation for adult eclosion at 96h APF (Millard and Pecot, 2018, Melnattur and Lee, 2011, Hadjieconomou et al., 2011, Kurmangaliyev et al., 2020) "

Kurmangaliyev et al., 2020 is probably not an appropriate citation here as it mostly deals with transcriptional programs of circuit assembly in the developing optic lobe

"Thus, 24-48h APF is a critical period to maintain necessary lamina and medulla neurons alive in the optic lobe."

Perhaps the authors want to revisit this sentence and explicitly say that 24-48h APF is a period where apoptosis defines cell numbers

"The development of the Drosophila visual system has been well described (Holguera and Desplan, 2018, Melnattur and Lee, 2011, Millard and Pecot, 2018, Hadjieconomou et al., 2011, Behnia and Desplan, 2015)."

I find that Behnia and Desplan, 2015 is not appropriate, as it is a review that describes the characterization of neuronal circuits underlying visual modalities in the fly brain, and not their development. The following reviews dealing with different aspects of neurogenesis, neuropile development and circuit formation are likely more relevant: Bakshi et al Current Opinion in Neurobiology 2025 Malin et al PNAS 2021 Ngo Dev Bio 2017

"R7 and R8, together with lamina neurons target to medulla layers M6 and M3, respectively, organizing into medullar columns that respond to the same point in visual space and maintain retinotopy."

I find this sentence misleading because lamina neurons do not target M6 and only L3 targets M3.

"Medullar interneurons also form connections across multiple layers, where each layer represents different visual features (Fischbach and Hiesinger, 2008, Millard and Pecot, 2018)."

Here, Behnia and Desplan, 2015, Matsliah et al Nature 2024, Borst and Groschner , Annu. Rev. Neurosci. 2023, and even Schnaitmann et al J Comp Physiol A Neuroethol Sens Neural Behav Physiol 2020 are better references that review feature detection and circuit organization in the optic lobe.

"Neurons within the lobula complex integrate signals from the medulla and project to the optic glomeruli in the central brain and motor outputs to enable appropriate behavior (Behnia and Desplan, 2015, Borst et al., 2020, Courgeon and Desplan, 2019b)."

I find Courgeon and Desplan, 2019b is not very appropriate here, as it reviews the coordination of neural patterning in the Drosophila visual system. More adequate and relevant manuscripts and reviews are Wu et al. elife 2016, Tanaka and Clark, 2022, Lapoetke et al. Neuron 2022 , even Zhao et al. elife 2024

"Spz-5 is well known as Drosophila neurotrophin-2 (DNT-2), and as Spz-3 has been proposed to have neurotrophin functions which we expand on and demonstrate here, we refer to Spz-3 as DNT-3 (Zhu et al., 2008, Coutinho-Budd et al., 2017, Sun et al., 2024, Ballard et al., 2014, Ulian-Benitez et al., 2017)."

This last paragraph of the introduction seems out of place, and partly redundant with page 3. Perhaps the authors would like to finish the introduction with a paragraph highlighting the major findings and conceptual advance of the manuscript? This seems to be a good and natural way of following their previous sentence "Here, we asked whether neurotrophin family ligands encoded by DNTs (spzs) and their Toll receptors could regulate cell survival during neural circuit formation, in the Drosophila pupal optic lobe."

RESULTS

I suggest the following edit: To ask whether DNTs (spzs) are expressed in the pupal optic lobe, we generated T2A-Gal4 driver lines for spz-1,3,4,5 fly lines , crossed them to 10xUASmyrGFP or 20UAS6xmCherry reporter flies, and analysed resulting progeny optic lobes during development and in the adult with anti-GFP antibodies, as required (Figure 1A).

Also, If the lines were also crossed with mcherry, mentioning anti-GFP antibodies is incomplete.

"spz-1MIO2318-T2A>myrGFP and spz-1MIO2318-T2A>6xmCherry revealed expression in a few centrifugal neurons in the lobula complex that projected to the lamina,"

At which stage?

"subsequently medulla neurons and abundant arborisations into the lobula complex and medulla."

Subsequent to what? I am sorry, but I don't understand this description.

"Expression from spz4MI5678 -T2A->myrGFP was not detected until 72h APF" Data before 72h APF is missing Where was it expressed? Which cell types? Which neuropiles?

"And then was found in the medulla and lobula complex and followed by the trachea." At which stage? What was followed by the trachea? I think the authors mean that at later staged (in the adult) expression was restricted to the trachea in both medulla and lobula

"spz-3-T2A>6xmCherry (hereby named DNT-3) was highly expressed in non-neuronal retinal cells and medulla neurons;"

At what stage?

"And subsequently in the trachea and possibly glia."

The authors could and should explain how they reach this conclusion. Given that no cell type specific markers were used, this identification was likely based on morphological features.

"Finally, DNT-2-T2A>6xmCherry (spz-5) was found in medulla neurons, which could be tentatively identified as Mi1 medulla neurons (Nern et al., 2025) by 48h APF, and this pattern was maintained."

The authors suggest these neurons are Mi1 based on what? Also, Fischbach's Cell Tissue Res (1989) seminal paper is probably worth mentioning.

"Abundant cells expressed spz-1 in the lobula complex (Figure 1B, left) and medulla (Figure 1B, right), and DNT-3 (spz-3) and DNT-2 (spz-5) in the medulla (Figure 1B) during optic lobe development. DNT-3 (spz-3) and DNT-2 (spz-5) were expressed in distinct non-neuronal cells in the retina, seen in Multi Colour Flip Out (MCFO) clones (Figure 1C)."

This sentence is misleading. These experiments allow the authors to conclude that spz+ neurons innervate these neuropiles. It does not allow the authors to conclude that spz molecules localize to the neuropiles. The authors should revise these claims in the main text and relevant figure legends

"To visualize the distribution of Tolls in the optic lobes during pupal development, we used the GAL4 lines previously described (Li et al., 2020), driving expression of the reporter myrGFP (Figure 2A)."

Same comment as above

"Using MCFO clones as well as myrGFP, we could identify some of the Toll-8+ neurons as Lawf1, feedback neurons projecting from the medulla to the lamina (Figure 2B), and L2 and L4 lamina neurons (Figure 2B, C); Toll-6+ cells to include L3 and L4 lamina neurons (Figure 2B,D); and Toll-2+ neurons as L1 lamina neurons which target to M1 and M5 medulla layers and L3 lamina neurons that project to M3 (Figure 3B,E) (Hakeda-Suzuki and Suzuki, 2014, Behnia et al., 2014)."

Hakeda-Suzuki and Suzuki, 2014, Behnia et al., 2014 are not the most appropriate references. Instead, I suggest the authors should cite Fischbach's Cell Tissue Res (1989).

"Overall, the expression in the scRNAseq dataset (Kurmangaliyev et al., 2020) of the spz ligands and Toll-8 (also known as Tollo) data were less consistent with the cell biology data, whereas the expression of Toll-1, -2 and -6 confirmed cells seen with the cell-biology based reporters"

It is perhaps more accurate to refer to the cell-biology based reporters as translation reporters, which is what T2a based Gal4 drivers are.

"Most particularly, Toll-2 mRNA (synonym 18w) was found in L1 and L3 lamina neurons over time, plus also in L5 at 24h APF, and Toll-6 mRNA was found in L2, L3, L4 over time, plus also in L1 at 24h (Supplementary Figure 1-6)."

This sentence is difficult to read and could be considered poorly written for several reasons including: - "plus also" is repetitive. "Plus" and "also" serve the same function. - Inconsistent punctuation. The lack of commas before "and Toll-6 mRNA..." makes the sentence feel unbalanced. -Vague time reference. "Over time" is imprecise. It's unclear whether it means during development, at multiple timepoints, or something else. Also regarding scRNAseq analysis: - the authors mention "We compared our reporter-based profiles with published scRNAseq datasets of the optic lobe through development (Kurmangaliyev et al., 2020, Ozel et al., 2021)" however the Ozel dataset doesn't seem to be used.

Also, from the Material and Methods section:

"The data were imported as a Seurat object, and cells corresponding to specific timepoints (e.g., 24 h, 36 h) were subsetted based on the provided metadata. Dimensionality reduction was carried out using principal component analysis (PCA), followed by Uniform Manifold Approximation and Projection (UMAP) embedding computed on the first 30 principal components. Cluster annotations provided by the original authors were used for all cluster-level analyses and visualisations."

The Kurmangaliyev dataset is already processed. I am probably missing something here, but is not obvious to me why the authors performed PCA again

"To conclude, at the time of naturally occurring cell death (0-48h APF), Toll-1 is highly expressed throughout the optic lobe; Toll-2, -6 and -8 are expressed in the medulla; Toll-8 and Toll-6 are prominently expressed in the lobula complex, and Toll-6 and Toll-2 are prominent in the lamina."

This is misleading. These experiments allow the authors to conclude that toll+ neurons innervate these neuropiles. It does not allow us to conclude that toll molecules localize to the neuropiles. The authors should revise these claims in the main text and relevant figure legends

"To ask whether DNTs can promote cell survival in developing optic lobes, we over-expressed DNT-2 (spz-5) and DNT-3 (spz-3) and visualized dying cells with the apoptotic marker anti-Dcp1 at the peak of naturally occurring cell death (24h APF)."

These experiments were done using Toll8-Gal4 and nsyb-Gal4 drivers. What's nsyb-Gal4 expression during development? Is the expression of this driver consistent with the conclusions drawn from these experiments?

"To test whether DNT-2 could promote cell survival during optic lobe development, we over-expressed full-length DNT-2FL or cleaved DNT-2CK in all neurons with nsybGAL4. This reduced the incidence of Dcp1+ apoptosis in the lamina and outside the lamina too (Figure 3C-D and Supplementary Figure S9C,D)."

How do the authors explain that DNT-2CK reduced the number of Dcp1+ cells?

"We generated DNT-3 (spz-3) loss of function mutants by P-element mobilization. DNT-2 and DNT-3 loss of function mutants caused considerable cell debris in the medulla and lobula complex, which compromised the analysis in this region, so we focused on the lamina."

Perhaps, rather than stating that the mutants caused considerable cell debris, the authors could say that the mutants displayed considerable cell debris

More importantly, I have concerns with the data from these experiments (Figure 3). Dcp1 signal volume intensity using Imaris. In all panels (A,C,E) the segmented images do not match the raw DCP1 staining, raising concerns on how much can one rely on this quantification. Could this be because the Dcp1 staining shown is a single z plane and the segmentation is a 3d rendering? The authors should carefully and robustly explain this discrepancy which is present in all images where Dcp1 signal volume intensity was quantified.

Also, could the authors explain why the quantifications in Figure 3B and 3C differ by an order of magnitude (10×) from those in panel 3D? Please look at the WT control, there is a 10X difference in signal volume intensity.

"Toll-2pTVGAL4 flies are heterozygous mutant for Toll-2, and, remarkably, in combination with DNT-2 homozygous mutants resulted in semi-lethality, revealing a functional interaction between these two genes.

I cannot entirely follow this conclusion. I understand the authors propose that the combination of partial loss of Toll-2 and full loss of DNT-2 affects viability, more than either mutation alone. Is this what they mean? Can the authors comment on the viability of DNT-2 mutants?

"Macrophages loaded with HisYFP and distributed mostly between the retina and lamina could be observed across these samples (Figure 4C), suggesting they had engulfed dead cells" How do the authors identify these YFP+ cells as macrophages?

"Together, these data show that DNT-2 functions as a ligand for Toll-2 to maintain the survival of neurons in the lamina, medulla and lobula complex during optic lobe development." While the results from Figure 4 showing that DNT-2 acts as a ligand for Toll-2 to support neuron survival are solid ( in particular panels C-F), it doesn't necessarily mean all neurons die directly due to loss of Toll-2 signaling. It is plausible that Neurons that express Toll-2 die because they lose critical survival signals. The death of these Toll-2-expressing neurons could then cause a cascade effect, where neighboring or connected neurons die indirectly due to loss of trophic support, disrupted circuits, or secondary damage. So, the observed cell death in multiple regions may be a combination of direct effects on Toll-2-positive neurons and indirect effects on other neurons. "In the Drosophila pupa, connectivity of lamina to medulla neurons takes place at 30-48h APF, and between medulla and lobula complex at 60-70h APF (Kurmangaliyev et al., 2020, Millard and Pecot, 2018, Pecot et al., 2014, Hadjieconomou et al., 2011)." I have the same comment as mentioned above regarding the timing of connectivity. If by connectivity, the authors mean the targeting of specific neuropiles and layer organization, it is incorrect that lamina and medulla organization precedes the connectivity between medulla and lobula neurons. These processes happen concurrently. Can the authors please clarify?

"Importantly, the expression of synaptic markers starts at 24h, peaks at 60h APF and spontaneous neuronal activity takes place at 48h APF, meaning that at least some neural circuits are already connected by this point (Kurmangaliyev et al., 2020)."

The correct placement of the reference to Kurmangaliyev et al. is after "peaks at 60h APF " A reference to Akin et al and Bajar et al when referring to PSINA is missing.

"Thus, the period of naturally occurring cell death overlaps with connectivity" I think the authors mean that the period of cell death is concurrent with the development of synaptic connectivity.

"L1 neurons normally project along columns that can be labelled with mAb24B10, and target to layers M1 and M5 of the medulla." The authors should mention that 24b10 labels the photoreceptors, providing a spatial reference to identify medulla columns

"Interestingly, Toll-2RNAi knock-down did not alter the phenotype caused by DNT- 112FL overexpression, and impaired targeting to the same extent as each genetic manipulation alone (Figures 5B,D)."

How do the authors interpret these results? And perhaps the authors would like to explain the rationale of overexpressing DNT-2fl in L1 neurons, that do endogenously express it.

"To conclude, these data show that DNT-2 and Toll-2 are required for appropriate connectivity of L1 neurons to target Mi1 medulla neurons at M1 medulla layer."

The authors characterize neuronal morphologies but do not directly assess connectivity using synaptic markers. While defective morphologies are likely to impact connectivity, the conclusion that DNT-2 and Toll-2 are required for appropriate connectivity should be tempered. The authors should revise their wording to reflect that their data support morphological defects rather than direct evidence of altered synaptic connectivity.

DISCUSSION

"In fact, throughout animal development, between 50% (e.g. in Drosophila) and 80% (e.g. in vertebrates) are lost to naturally occurring cell death"

"of neurons" is missing before "are lost"

"Consistently with these findings, we have shown that the survival of L1 neurons depends on DNT-2 functioning together with Toll-2."

The authors state that "the survival of L1 neurons depends on DNT-2 functioning together with Toll-2." It seems that what they intend to convey is that DNT-2 acts as a ligand for Toll-2. The text should be clarified to explicitly indicate this ligand-receptor relationship rather than implying a cooperative function.

"These data demonstrate that DNT-2 and Toll-2 function together in visual system development."

Since the authors did not use tub-GAL80 or another temporal control to restrict gene expression specifically to development, the observed phenotypes could reflect combined developmental and adult effects. Throughout the text, the authors should revise their wording to acknowledge this limitation.

"Finally, interference with the normal levels of DNT-2 and Toll-2 also impaired axon targeting and dendritic morphology, consistently with the coupling between cell survival with connectivity."

The authors state that interference with normal levels of DNT-2 and Toll-2 "impaired axon targeting and dendritic morphology, consistently with the coupling between cell survival and connectivity." This statement seems tautological, as neurons that die cannot form connections. The authors should clarify whether they are referring to a specific mechanistic link beyond this obvious consequence.

"Our findings are consistent with prior reports that had shown the maintenance of cell survival to be required during neural circuit formation."

This statement seems tautological. It is generally expected that neurons must survive in order to contribute or be part of neural circuits. The authors should clarify if they are highlighting a specific mechanistic insight beyond this obvious requirement.

"In the medulla, Dm8 medulla neurons are produced in excess and are eliminated during connectivity to their R7 inputs (Courgeon and Desplan, 2019a). This is enabled by the cell surface molecular tags DIP in yDm8 binding Dpr11 in yR7, during synaptic matching (Courgeon and Desplan, 2019a)."

The statement that "Dm8 medulla neurons are produced in excess and are eliminated during connectivity to their R7 inputs" is both unclear and inaccurate. It is not evident what is meant by "during connectivity." Moreover, Courgeon and Desplan (2019a) show that Dm8 neurons undergo cell death before or by P40, prior to synaptogenesis. The authors should correct this statement and clarify the timing and mechanism of Dm8 neuron elimination.

"Importantly, the maintenance of cell survival takes place during connectivity, and enables synaptic matching between connecting neurons."

It is unclear what is meant by "during connectivity." Moreover, both Courgeon et al. (2019a) and Xu et al. (2018, 2022) show that these neurons (e.g., Dm8, Dm12, Dm14) undergo cell death before or by P40, prior to synaptogenesis. The authors should clarify the timing and mechanism of cell survival and revise this statement accordingly.

"By contrast, it has also been proposed that apoptosis plays a minor role in cell number control during visual system development, depending instead on cell proliferation and spatial patterning through Dpp/BMP signalling (Malin et al., 2024). However, those findings were based on events taking place at the larval third instar wandering stage, when proliferation and spatial patterning are prevalent, whereas apoptosis peaks in pupa."

It seems that the authors are trying to suggest that different mechanisms control cell numbers at different developmental stages: during larval neurogenesis (L3), cell numbers are regulated primarily by proliferation and spatial patterning, whereas in the pupal stage, neuronal survival via apoptosis plays a key role. If this is the intended point, it should be stated more clearly, as the current comparison to Malin et al. (2024) is confusing and does not make this distinction explicit.

"However, Toll-2 mutant MARCM clones generated in the pupa result in a dramatic loss of lamina neuron dendrites and aberrant axonal navigation in the medulla, as well as widespread neuronal loss (Li et al., 2020)." This statement is puzzling. Neurogenesis occurs during the larval stage until P15, and MARCM requires progenitor cell division. The authors should clarify how MARCM clones were generated during pupation and provide the relevant experimental details in the Materials and Methods. "Importantly, connectivity between L1 and medulla neurons takes place between 20-48h APF, during the period of naturally occurring cell death, and spontaneous activity in the optic lobe takes place at 48h, meaning at least some circuits are connected by then (Kurmangaliyev et al., 2020)."

The authors state that "connectivity between L1 and medulla neurons takes place between 20-48h APF," but no reference is provided for this timing. To my knowledge, no study has directly demonstrated this, so the authors should either provide supporting evidence or revise this statement. Additionally, citing Kurmangaliyev et al. (2020) for spontaneous activity in the optic lobe is not appropriate for this point, as PSINA was originally described by Orkun Akin.

"When altering DNT-2 or Toll-2 levels, L1 axonal terminals in the medulla were misrouted, rather than being confined to a single column. This is reminiscent of the phenotypes caused by alterations in Dscam and Fez levels (Millard et al., 2007, Peng et al., 2018)."

The authors note that altering DNT-2 or Toll-2 levels causes L1 axonal terminal phenotypes reminiscent of phenotypes caused by changes in Dscam and Fez levels (Millard et al., 2007; Peng et al., 2018). However, they only reference these previous studies without discussing whether there could be a shared mechanism. While these comparisons are interesting, the manuscript would benefit from either a deeper discussion of potential mechanistic links or a clear statement that the comparison is purely phenotypic.

"As DNT-2 is secreted in medulla neurons and Toll-2 is expressed along neurons that connect in the medulla (e.g. L1, Mi1, Tm3, Dm9, T4), DNT-2 could help keep connecting neurons together during dynamic cellular events in development."

This sentence is poorly written and vague. It is unclear what "keep connecting neurons together" means mechanistically. Likely, DNT-2 is secreted by postsynaptic medulla neurons (e.g., Mi1), whereas Toll-2 is expressed in neurons innervating the medulla (e.g., L1, Mi1, Tm3, Dm9, T4). The authors should rephrase this sentence to clearly convey their mechanistic and cellular interpretation.

FIGURES

Figure 1 Arrowheads point to what? OL orientations should be described in the figure captions

Figure 5 The title of Figure 5 ("Altering the levels of DNT-2 and Toll-2 modifies L1 axon targeting at medulla M1 layer") is misleading. The correct layer targeting is preserved; what changes is the pattern of unicolumnar innervation. When DNT-2 or Toll-2 levels are altered, L1 neurons innervate multiple columns rather than maintaining their normal single-column specificity. The title should be revised to reflect that the defect is in columnar specificity rather than layer targeting.

Supplementary Figures S1 to S6 Combined UMAPs showing the expression of spz-1, -3 (DNT-3),-4 and -5 (DNT-2) and Toll-1, -2 (18w), -6 and -8 (Tollo) in distinct cells over time.

Information about which UMAP corresponds to which time point is missing

MATERIALS AND METHODS Genetics. Please see S1 Table for the list of the stocks used and Table S3 for full genotypes for each experiment. Table S3 is not the full genotypes for each experiment. This information is partly available in the Source Data excel file

Significance

During development, neurons are initially produced in excess. One mechanism by which the final neuronal numbers are refined relies on trophic support, which maintains the survival of necessary neurons, while excess neurons are eliminated.

In the Drosophila optic lobe, a wave of apoptosis occurs during pupation, peaking at a critical period thought to be essential for establishing final neuronal numbers and supporting proper neural circuit formation. However, the mechanisms underlying this developmental process remain poorly understood. In this manuscript, Alshamsi et al. investigate this wave of apoptosis by examining the role of Drosophila neurotrophins (DNTs), which are encoded by spätzle (spz) paralogous genes and signal through Toll receptors to regulate neuronal survival during brain development.

The authors use translation reporters to demonstrate that DNTs and Toll receptors are differentially expressed across various neuronal types innervating all optic lobe neuropils during development. They then focus on DNT-3 and DNT-2, which they show to be necessary for controlling neuronal numbers, likely by maintaining neuronal survival during pupal stages.

Notably, the results reveal a previously uncharacterized interaction between DNT-2 and Toll-2. The findings suggest that DNT-2 acts as a neurotrophic factor produced by medulla intrinsic neurons, binding to the Toll-2 receptor expressed in other neurons innervating the medulla. By examining Toll-2+ L1 neurons, which are postsynaptic in the lamina and presynaptic in the medulla, the authors provide compelling evidence that DNT-2/Toll-2 signaling regulates L1 neuronal numbers.

Interestingly, the authors also show that disruption of DNT-2/Toll-2 signaling affects L1 axonal and dendritic morphologies. However, the extent to which changes in neuronal survival and neuronal morphology are mechanistically or cellularly linked is not addressed. These findings are consistent with previous reports showing that DNTs and Toll receptors regulate neuronal survival in embryonic, larval, and pupal ventral nerve cords, as well as in the adult. Importantly, DNTs and Tolls can also promote cell death, highlighting their dual role in controlling neuronal number and circuit formation.

While the data is for the most part solid, I have concerns regarding the execution, interpretation of certain results and the conclusions drawn. Additionally, references to previous work are often incorrect or incomplete; I provide several examples below along with non-exhaustive suggestions for improvement. Finally, the manuscript would benefit from careful text revision and improvements to figure presentation, for which I also offer non-exhaustive guidelines.

Overall, I would recommend this manuscript undergo revision before its publication and I would be happy to reassess a revised version that addresses the comments above.

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-2023-01861

Corresponding author(s): Manuela, Baccarini

1. General Statements

We were happy to learn that all three reviewers found the paper novel and of interest for a cell biology audience. They especially highlighted the carefully conducted screen, whose results will be integrally published with this paper and will be of use for scientists interested in lysosome biology. The revised manuscript contains key validation experiments (antibody/KO controls, lysosome positioning quantification, live-cell actin dynamics) to strengthen our central conclusions.

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

• The colocalization of endogenous PLEKHG3 and LAMP1 as depicted in figures 3B and 3C (data from fixed cells) is not convincing. PLEKHG3 appears to be present on cortical actin structures as opposed to being colocalized with LAMP1 on lysosomes. And related to this point:

• There is no apparent colocalization of PLEKHG3 and lysotracker in the movie S5.

Answer:

We do not claim that the two structures always colocalize, but that PLEKHG3 is a LAMTOR3 vicinal protein that co-enriches with a subset of peripheral lysosomes at focal adhesions (FAs)/protrusions. The images in Figure 3C are schematic for how PLEKHG3-high/low and LAMP1-high/low regions were defined for quantification. We agree with the Reviewer and with the previous literature that PLEKHG3 main localization is to cortical actin structures, as reaffirmed by the strong cortical actin localization shown in Figure 3F of the original version and in Figure S2C (HEK293T cells) and in Figure S3A in HeLa cells in the revised version. We have clarified the text referring to Figure 3F on page 26, line 11-14 as follows:

“Immunofluorescence experiments showed the reported colocalization of endogenous PLEKHG3 (Figure S2C in HEK293T cells, Figure S3A in HeLa cells) and GFP-PLEKHG3 with cortical actin structures and the partial localization of LAMP1-positive vesicles to these structures in correspondence with vinculin-positive focal adhesions.”

Live imaging in GFP-PLEKHG3-expressing cells (including movie S5, and particularly the stills of the leading edge in Figure 4F) further supports this spatial association without implying obligate colocalization. We added explicit language (p. 27, lines 19-21): “Following a single cell over time, we could observe that __a subset of __lysosomes appears to travel to PLEKHG3 accumulation sites and specifically move into developing protrusions.”

• The authors should also confirm the specificity of the PLEKHG3 antibody in immunofluorescence using control and PLEKHG3 siRNA in untransfected cells that have not been transfected with GFP-PLEKHG3 (as is shown in Fig. S2C). Numerous antibodies recognize the overexpressed protein but do not recognize the same protein at endogenous expression levels.

Answer: To assess the specificity of the antibody for endogenous PLEKHG3 we have used HEK293T cells, which based on the fact that PLEKHG3 is most highly expressed in neuronal cells (https://www.proteinatlas.org/ENSG00000126822-PLEKHG3/tissue#expression_summary) should yield a clearer endogenous signal. The pattern of PLEKHG3-positive bands is similar to that observed in HeLa cells, and only the band around 250 kD is clearly reduced by the siPLEKHG3. The IF images show a selective loss of the PLEKHG3 signal in correspondence of actin filaments close to the plasma membrane, whereas the nuclear signal is preserved, and therefore to be considered non-specific (revised Figure S2B-C).

Extract from revised Figure S2B-C: ____PLEKHG3 KD test in HEK293T cells: B) Western blot of HEK293T cells showing downregulation of PLEKHG3 expression upon siPLEKHG3 treatment compared to siScr. Bar plot shows quantification of PLEKHG3 bands from immunoblot above. Error bars = SEM, n=3. * = p values according to student's t-test. C) Immunofluorescence images of HEK293T cells. siPLEKHG3 shows drop in PLEKHG3 intensity in the periphery of the cell and less colocalization with Phalloidin. Scale bar = 50 µm. Line plots show intensity profiles of Phalloidin (green) and PLEKHG3 (red) along the white lines in the merged inset images. Scale bar = 10 µm.

In addition, we have now generated a PLEKHG3 CRISPR-Cas KO in HeLa cells. The results, shown in revised Figure S2G-I, confirm the specificity of our reagents and the localization of PLEKHG3 seen in HEK293T cells.

Extract from revised Figure S2G-I: G) Immunoblot and quantification of HeLa PLEKHG3 KO cells represents the degree of PLEKHG3 depletion achieved using different guides compared to WT cells transfected with empty vector (EV). The most potent guides (8-9) are boxed in red. H) Immunofluorescence images of WT and PLEKHG3 KO8 cells reveal an overall drop in PLEKHG3 intensity and the specific loss of PLEKHG3 signal at the periphery of the cells. I) Quantification of PLEKHG3 intensity as displayed in H for two KO cell lines compared to WT cells. Dots represent individual data points of each of the three-color coded replicates; diamonds represent the mean of each replicate; black bars represent the mean ± the SEM of three biological replicates; * = p values according to 2-way-ANOVA. Error bars = SEM.

These results establish that the peripheral cortical signal is specific for endogenous PLEKHG3; the nuclear signal is non-specific. Loss of PLEKHG3, however, had no effect on lysosomal distribution, morphometric parameters (see revised Figure S4A-C) or protrusive activity (see revised Figure S6E-F) compared to WT cells.

Extract from revised Figure S4A-C: ____PLEKHG3 KO does not influence lysosomal distribution or cell morphometry: A) __Quantification of lysosomal distribution in WT compared to two KO cell lines. N ≥ 50 cells in three biological replicates. __B) Schematic representation of analysis of cell shape descriptors as referred to in C). Left picture shows the calculated outline in yellow based on which the cell area and circularity are calculated. Right picture shows the minor and major cell axis which, calculated as fraction, result in the aspect ratio of the cell. Scale bar = 50 µm. C) Quantification of cell morphometric parameters Area, Circularity and Aspect ratio. N ≥ 50 cells in three biological replicates. Black dots represent mean of each biological replicate. Statistical analysis according to student’s t-test. Error bars = SEM.

• The claim that "peripheral accumulation of lysosomes inhibits protrusion formation and limits cell motility" should be tested more rigorously using the RAMP method, preferably in living cells. Other approaches, such as overexpression/siRNA of Arl8b and other motor adaptors, such as SKIP/PLEKHM2, can be used to alter lysosome positioning and confirm this central finding of the manuscript. The authors could also consider including additional mechanistic data in order to comprehend how lysosome positioning controls cell motility. For instance, the RAMP approach could be employed to investigate cortical actin dynamics upon repositioning of lysosomes to the peripheral/perinuclear region.

Answer: We have purchased the RAMP system from Addgene and adapted the reporters to express fluorophores compatible with our color setup in the different respective cell lines (HeLa GFP/GFP-PLEKHG3 as well as in HeLa PLEKHG3 KO cells. Unfortunately, we’ve experienced difficulties with imaging due to suboptimal efficiency of the double transfection necessary to introduce the RAMP system into the cell lines. The LAMP1 and the KIF plasmids were co-expressed at very different levels in the cells, leading to the need for high laser power in both channels, which too often resulted in cell death. Additionally, the redistribution of the lysosomes after biotin addition was incomplete and slower than initially expected, which made it impossible to investigate cortical actin dynamics.

To gain some mechanistic insight, we have performed further live cell imaging analyses comparing PLEKHG3 WT vs KO cells and GFP vs GFP-PLEKHG3 cells expressing a combination of BFP-LifeAct (to visualize F-actin) with either control mCherry or mCherry-KIF1A to move lysosomes to the periphery.

• In all experiments, locking lysosomes in the periphery drastically reduces membrane dynamics (protrusion formation and retractions).
• PLEKHG3 remains colocalized with LifeAct under KIF1A (Fig. S6C–D), indicating that the reduced protrusiveness is upstream or independent of PLEKHG3’s cortical localization
• Live-cell BFP-LifeAct imaging revealed that KIF1A-driven peripheral lysosomes reduce protrusion formation/retraction and dampen cortical actin dynamics in both WT and PLEKHG3 KO cells (Fig. S6A–B, E–G; Movies S12–15; S18–21), indicating that these phenomena are independent of PLEKHG3. We believe these data, together with the quantitative lysosome repositioning and FA analyses, substantiate the central finding that forced peripheral lysosome clustering correlates with more adhesive FA states and suppressed protrusive activity. We have clarified scope and limitations accordingly.

Extract from Figure S____1____A-C,E,G: PLEKHG3 localizes to F-actin independently of lysosomal transport but is dispensable for protrusive activity. A) __Stills from live cell imaging (Movies S12-15). Cells stably expressing GFP or GFP-PLEKHG3 were transfected with the indicated mCherry constructs. Yellow arrows = forming protrusions; blue arrows = retracting protrusions. Stills were generated over a period of 2 hrs. Scale bar = 50 µm. __B) __Quantification of protrusions formed and retracted over time in cells from A. Values indicate average number of protrusions formed in a timespan of three hours from a total of ≥ 15 cells per condition. Error bars = SEM. C) Quantification of colocalization by Fijis Coloc2 Plugin (see materials and methods) over a timespan of three hours. Lines represent mean of all cells per condition, and light-color shading represents the SEM. __E) Stills from live cell imaging (Movies S18-21). PLEKHG3 WT and PLEKHG3 KO cells were transfected with the indicated mCherry constructs and incubated with LysoTracker. Yellow arrows = forming protrusions; blue arrows = retracting protrusions. Stills were generated over a period of 2 hrs. Scale bar = 50 µm. G) Quantification of protrusions formed and retracted over time in cells from E. Values indicate average number of protrusions formed in a timespan of one hour from a total of ≥ 15 cells per condition. Error bars = SEM. In B,G, black asterisks denote p values according to Kruskal-Wallis and Bonferroni post-hoc testing, comparing the effect of KIF1A against mCherry or PLEKHG3 WT against KO.__ __

• It is not clear how the authors conclude that Figure 4E graph shows "the LAMP1 signal was stronger in paxillin-labeled FA compared to control regions". The 4E graph shows LAMP1 signal in GFP versus GFP-PLEKHG3 and shows a modest enrichment of LAMP1 in FAs in GFP-PLEKHG3 overexpression. LAMP1 enrichment in FAs is also not obvious in the image shown in Figure 4B.

Answer: We stand corrected – the Figure we referred to was not in the manuscript. It has been inserted now, as a plot next to Figure. Figure 4B (schematic representation of colocalization analysis) was designed to explain how we define focal adhesions (paxillin positive) and adjacent control regions (same size and shape, but paxillin-negative). The actual analysis was missing and has now been inserted. We apologize for this mistake.

We do not claim that PLEKHG3 brings lysosomes to FAs. The enrichment of lysosomes in FA regions of cells expressing GFP-PLEKHG3 compared to GFP-expressing cells shown in 4E, as the Reviewer correctly notes, is marginal and is not highlighted anywhere in the text exactly for this reason.

• In Fig. 2B, there appears to be a labeling error. The lanes 2,4 and 7 appear to be transfected with L3-T-V5 but labeled as GFP-V5-cyto. Here the PLEKHG3 band should be indicated.

• AND -Fig. 2C is an IP experiment as per the manuscript text but it is labeled as pulldown.

Answer: We stand corrected, and the necessary changes have been made in the revised version in Figure 2B.

Reviewer 2

1 - Specificity of PLEKHG3 antibody: In Fig. S2, authors show that PLEKHG3 antibody recognizes 3 bands (above 100 kDa, above 130 kDa and 250 kDa) and all of them are reduced by the silencing of PLEKHG3. Then, in Fig. 2A and C, authors only show the band above 130 kDa, despite implying that the specific band should be "much higher than the 134 kDa calculated from the aminoacid sequence of the protein".

In Fig. 2 B, they show all the bands shown in Fig. S2 and presumably favor that the specific band is the 250 kDa one. Finally, in Fig. 2D, they show all bands and note that the band above 130 kDa is not specific. Therefore, authors need to conclude what is the specific band and always analyze the same one, and, possibly, use a different antibody or purify this one to remove non-specific binding. Without this, the main result of the paper, cannot be substantiated.

Answer: We apologize for this misunderstanding. The antibody recognizes three bands, all reduced by siRNA treatment. These three bands are only resolved in the gels in Figure S2A and B, and in Figure 2B. The reason for this is the high molecular weight of the isoforms, that are resolved in these 8% gels, but collapse into one band in the 15% gels shown in Figure 2A and C. Therefore, the high molecular weight bands are not resolved under these conditions. 8% gels such as the ones in Figure 2B are needed to resolve the high molecular weight bands.

Figure 2D shows an 8% gel, and therefore all bands are visible. The band marked by an arrow is only present in the streptavidin pulldowns but not in the input or in the supernatant and is therefore considered unspecific. This has been clarified in the revised figure legend on page 41. In addition, to assess the specificity of the antibody for endogenous PLEKHG3 we have used HEK293T cells, which based on the fact that PLEKHG3 is most highly expressed in neuronal cells (https://www.proteinatlas.org/ENSG00000126822-PLEKHG3/tissue#expression_summary) should yield a clearer endogenous signal. The results of this experiment are shown in Figure S2B-C of the revised manuscript. The pattern of PLEKHG3-positive bands is similar to that observed in HeLa cells, and only the band around 250 kD is clearly reduced by the siPLEKHG3. The IF images show a selective loss of the PLEKHG3 signal in correspondence of actin filaments close to the plasma membrane, whereas the nuclear signal is preserved, and therefore to be considered non-specific. More importantly, we have now generated a PLEKHG3 CRISPR-Cas KO in HeLa cells. The results, shown in Figure S2G-I confirm the specificity of our reagents and the localization of PLEKHG3 seen in HEK293T cells. Loss of PLEKHG3 however, had no effect on lysosomal distribution or morphometric parameters compared to WT cells (Figure S4A-C).

2 - In page 12, authors state that "These results indicated that PLEKHG3 is a transient interactor, or a proximal, not directly binding protein, of L3" and in page 14 that "... PLEKHG3 is a proximal L3 protein rather than a transient physical interactor". It is not clear at all how did the authors reach such conclusions, nor they have data to conclude this. Indeed, they would have to express the proteins in vitro and test their interaction to conclude about a direct binding. They also do not know what is the stability of the interaction.

Answer: This is also a misunderstanding. We mislabeled Figure 2C as “pulldown”, rather than “IP”, as it is characterized in the text. We revised terminology to “vicinal (proximity-labeled) protein” throughout, avoiding claims on directness. Our basis is: robust L3 TurboID labeling of PLEKHG3; failure to co-immunoprecipitate PLEKHG3 with V5-tagged L3 (Fig. 2C); lack of PLEKHG3 labeling by TMEM192; and unchanged PLEKHG3 FA localization in L3 KO (Fig. S3H–J). Together, these support spatial proximity rather than a stable L3–PLEKHG3 complex. We explicitly state that we did not perform in vitro binding due to the negative co-IP.

Based on these negative data, we did not proceed to test the possibility of complex formation in vitro.

3 - Still in page 12, authors state that "... two different membrane structures, protrusions and ruffles". What do the authors mean exactly by "protrusions", as there are several different ones (e.g., lamellipodia, filopodia, pseudopods)? And how can they distinguish between ruffles and, for example, lamellipodia? They need to use markers and more carefully analyze their morphology to be able to distinguish these. Like this, it is too preliminary.

Answer: It was our intention to indicate with the arrows the trajectories in the figure along which we measured the MFI of LAMP1 and PLEKHG3. Although this is indicated in the figure legend, it had apparently given the impression that the arrows indicated specific membrane structures. Since we are focusing on different types of membrane protrusions rather than ruffles, we replaced the ambiguous terms "ruffles" and "protrusions" with the terms "elongated protrusions" (Figure 3D upper panel) and then compared these with "non" elongated protrusions” (Figure 3D lower panel). Indeed, we note that PLEKHG3 accumulation is possible below and along the plasma membrane, but colocalization with lysosomes occurs preferentially in elongated protrusions. We therefore amended the text on page 26, line 4-9 as follows:

„More specifically, we found that PLEKHG3 colocalized more strongly with LAMP1-positive vesicles in elongated membrane structures (Figure 3D-E). Focal adhesion sites, which anchor the intracellular cortical actin network to the extracellular matrix and are remodeled with the help of late endosomes/lysosomes during protrusion formation and cell motility, can also be found in such elongated membrane protrusions (reviewed in [58,59]).”

5 - It is not clear if in cells KO for PLEKHG3, the overexpression of KIF1A leads to more lysosomes localizing close to the PM, as well as more protrusions and more cell motility, as the authors only compare cell overexpressing GFP or GFP-PLEKHGL3.

Answer: We have now generated a PLEKHG3 KO cell line. In these cells, KIF1A still drives peripheral lysosome clustering and suppresses protrusive activity and actin dynamic (see revised Figure S4A-C displayed below). Baseline lysosome distribution and morphometric parameters are unchanged in KO cells (see revised Figure S6E-F displayed below).

Extract from revised Figure S4A-C: ____PLEKHG3 KO does not influence lysosomal distribution or cell morphometry: A) __Quantification of lysosomal distribution in WT compared to two KO cell lines. N ≥ 50 cells in three biological replicates. __B) Schematic representation of analysis of cell shape descriptors as referred to in C). Left picture shows the calculated outline in yellow based on which the cell area and circularity are calculated. Right picture shows the minor and major cell axis which, calculated as fraction, result in the aspect ratio of the cell. Scale bar = 50 µm. C) Quantification of cell morphometric parameters Area, Circularity and Aspect ratio. N ≥ 50 cells in three biological replicates. Black dots represent mean of each biological replicate. Statistical analysis according to student’s t-test. Error bars = SEM.

Extract from revised Figure S6E,G: ____PLEKHG3 localizes to F-actin independent of lysosomal transport but is neglectable for lysosomal effect on protrusive activity. E) Stills from live cell imaging (Movies S18-21). PLEKHG3 WT and PLEKHG3 KO cells were transfected with the indicated mCherry constructs and incubated with LysoTracker. Yellow arrows = forming protrusions; blue arrows = retracting protrusions. Stills were generated over a period of 2 hrs. Scale bar = 50 µm. G) Quantification of protrusions formed and retracted over time in cells from E. Values indicate average number of protrusions formed in a timespan of one hour from a total of ≥ 15 cells per condition. Error bars = SEM. In F,G, black asterisks denote p values according to Kruskal-Wallis and Bonferroni post-hoc testing, comparing the effect of KIF1A against mCherry or PLEKHG3 WT against KO.

6 - Regarding the statistical analysis, authors assert that it was done using Student's t tests, unless otherwise stated. However, they never refer in figure legends other statistical analysis methods. If so, they cannot use such test, for example, in cases where more than two groups are compared.

Answer: We clarified in Methods that we performed two-group comparisons unless otherwise stated. Where >2 groups are compared, we used appropriate tests with correction (e.g., Kruskal–Wallis with Bonferroni in Fig. S6B, S6G). Figure legends now explicitly state the test used.

__Minor comments: __

1 - In the abstract, authors refer that cytosolic proteins are recruited to platforms on the limiting membrane of lysosomes. What do they mean by "platforms"? Is it microdomains?

Answer: We apologize for this lack of clarity and have now changed the first sentence in the abstract on page 1 to “Lysosomes are key organelles involved in metabolic signaling pathways through their ability to recruit cytosolic molecules to protein platforms bound to the lysosomal membrane”. We refer to protein platforms as multifunctional protein complexes that can recruit and assemble signaling components (e.g., the recruitment of mTORC1 activating proteins by the LAMTOR complex).

2 - In the Introduction, there is a period before the reference at the end of the first paragraph.

Answer: We stand corrected. See changes on page 18, line 9.

3 - In the results, Fig. 1E is mentioned before Fig. 1D and Figure S1F before Fig S1E, which can be confusing.

Answer: Figure S1E on page 6 was mislabeled as Figure 1E and Figure S1K on page 9 was mislabeled as Figure 1K. We stand corrected. See changes on page 21, line 21+23 and page 23, line 5.

4 - All the immunofluorescence images need to be bigger, in general, and have zoom-ins, except Fig. 3A, 4B, 4F, and S2C. Also, in Fig. S1F, the green channel has different intensities and the V5-lyso signal is clearly saturated. Finally, Fig. S1D, S1I and S3F must be enlarged, too.

Answer: We appreciate the Reviewer's suggestion, but enlarging all the immunofluorescence images and including zoom-ins would make the manuscript very crowded and could distract from the main findings. Regarding the expression levels of the baits, as mentioned in the manuscript, we aimed to express them at near-endogenous levels. However, TMEM192 is expressed at higher levels than LAMTOR3 in these cells, which may have resulted in the observed discrepancy. We hope the Reviewer will understand our decision and find the current presentation of the data clear and informative.

5 - In page 9, where it reads "Figure 1K", should read "Figure S1K".

Answer: See answer to minor point 3.

6 - The observation that PLEKHG3 silencing leads to loss of the perinuclear clustering of LAMP1-positive vesicles, and increase in their accumulation at the cell tips, is not referred in the text.

Answer: While this might seem the case in part of the cells shown in the representative image in Figure S2C, population-level analysis (n > 30 cells) did not support a shift in lysosome distribution with PLEKHG3 silencing.

__Figure 1 for Reviewer 2: __Lysosomal distribution in HeLa cells transfected with either siScr or siPLEKHG3. X-axis is relative distance from the nucleus and Y-axis the normalized intensity of the LAMP1 channel. Results are averages of >30 cells from one experiment (only displayed in “final revision” document).

Similar results were obtained using two independent PLEKHG3 KO cell lines, and are shown in Figure S4A

__Extract from revised Figure S4A-C: PLEKHG3 KO does not influence lysosomal distribution or cell morphometry: A) __Quantification of lysosomal distribution in WT compared to two KO cell lines. N ≥ 50 cells in three biological replicates.

7 - Fig. 2C is not referred in the legend.

Answer: We stand corrected and have changed the legend of Figure 2 accordingly on page 41.

8 - Figure S3A and B: authors should show the colocalization of endogenous PLEKHG3 with phalloidin and not only the GFP-tagged protein.

Answer: We thank the Reviewer for this comment and have performed this experiment showing the colocalization of endogenous PLEKHG3 with F-actin structures stained by Phalloidin. Even though the endogenous PLEKHG3 staining in HeLa cells is rather weak, sites where membrane protrusions are formed are clearly marked with PLEKHG3 staining below the plasma membrane. These data confirm the specific colocalization of PLEKHG3 with Phalloidin shown in the revised Figure S3A. See also the extract from Figure S3A below.

Extract from revised Figure S3A: Immunofluorescence images of HeLa cells. A) HeLa cells stained with PLEKHG3 (red) and Phalloidin (green). The nucleus is indicated by DAPI staining (blue). Scale bar = 50 µm. Insets on the right as indicated by white box in image on the left. Scale bar = 10 µm. Line plot corresponds to white line in merged inset.

9 - In page 14, authors refer to Fig. 3G, which does not exist.

Answer: We stand corrected, the sentence on page 14, line 9 (now page 26 line 24 in revised document) refers to Figure S3G.

10 - In page 30 and page 32, different antibodies for LAMP1 and PLEKHG3 are mentioned, but in the figure legends authors do not refer which one they used.

Answer: We tried different PLEKHG3 antibodies but ended up using only one. The other antibody has been excluded from the list on page 32, line 18 (now page 9, lines 4-5 in revised manuscript). We have specified which LAMP1 antibodies were used in which Figure in the Material and Methods on page 6, line 23 and page 7, line 4-5.

11 - In page 33, where it reads "300 µm protein", it should probably read "300 µg protein".

Answer: We stand corrected. See changes on page 10, line 2.

Reviewer 3

A key issue … is that the authors focus solely on peripheral lysosomes as target compartments for PLEKHG3. This is not self-evident, particularly in light of images presented in Figures 2 and 3, where colocalization of PLEKHG3 with perinulcear lysosomes appears very likely. The authors should make differences/similarities they observe between effects on perinuclear versus peripheral lysosomes explicit both with data and in the text, if such differences exist.

Answer: The Reviewer is likely addressing the images in Figure 3, which were obtained by staining endogenous PLEKHG3 and do diffuse staining around the nucleus. This perinuclear haze is resistant to siPLEHG3 (revised Figure S2C) or to PLEKHG3 CRISPR-Cas9-mediated ablation (revised Figure S2H) and is not observed with the GFP-PLEKHG3 fusion protein (revised Figure S2E-F), which gives a less diffuse signal. This is why we are confident about the colocalization of PLEKHG3 with peripheral lysosomes.

Data presented in Figure 6 showing cell motility analysis is interesting and has potential to make the manuscript impactful. Similarly, data in Figure 4F (live cell imaging) looks attractive but is not informative in the absence of relevant genetic perturbations as comparisons. These types of experiments would benefit greatly from PLEKHG3 loss of function analysis, as well as mutational analysis in the over-expression setting.

Answer: We have now generated a PLEKHG3 KO cell line. Loss of PLEKHG3, however, had no effect on lysosomal distribution or morphometric parameters compared to WT cells, and it does not impact the suppression of protrusions/actin dynamics by KIF1A is preserved in KO, indicating PLEKHG3 is not required for this phenotype.

Mutational analysis of PLEKHG3–LAMTOR binding is not feasible in the absence of co-IP or other direct binding evidence (see revised Figure S6E,G displayed in the answer above).

Minor point: 1. Multicolor overlays with one of the channels in white is in my view not reader-friendly. Appreciating colocalization between endosomes/lysosomes, actin and G is very important for this study, and while is typically reserved to show overlay between green and magenta or green (standard for 2 channels), red and blue (standard for 3-channels). I therefore advise the authors to choose a different color combination throughout the figures when presenting microscopy images.

Answer: White as a channel color has been substituted for with red (in the 2- and 3-color images) or with blue (in the 4-color images) throughout the images of the revised manuscript, except for the stills from the videos that have not been changed because no colocalization analysis has been performed in this case.

1. Description of analyses that authors decided not to carry out

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

Reviewer 2

4 - At least Fig. 2F and 3A need quantification. Regarding cell motility, there is no quantification and the authors must perform a quantitative assay (despite stating that "As another measure of cell motility, analysis of the number of forming protrusions and retracting membranes..."). Not only this is not a measure of cell motility, but there the issue of what are "protrusions" referred above. Therefore, authors need to quantify the distance that the cells move and/or perform quantitative motility/migration assays.

Answer: We appreciate the Reviewer’s attention to detail and agree that the quantification of these figures is essential to understand the results. We believe that the Reviewer refers to Figure 3F and Figure 4A, as there is no Figure 2F, and Figure 3A only confirms the localization of endogenous PLEKHG3, as previously reported in (Nguyen et al., PNAS 2016). If our assumption is correct, then the salient aspects of Figure 3F, which is a representative image, are quantified

• in Figure 3C-E (endogenous PLEKHG3 colocalization with LAMP1/lysosomes)
• In Figure 4E and 5F-G (FA with LAMP1/lysosomes).
• Figure 4A is quantified in Figure 4C-E (GFP-PLEKHG3 colocalization with FAs, labeled with paxillin in this experiment, and LAMP1 colocalization with FAs). In response to the Reviewer's comment regarding the absence of quantification for cell movement/migration in our study, we apologize for any confusion that may have arisen from our use of the term "cell motility." We have clarified usage to mean membrane remodeling dynamics integral to migration rather than net displacement. To avoid overclaiming, we removed statements that could imply directed migration and focused on protrusion/retraction metrics and shape changes. In this context, our statement that lysosomal subcellular localization plays a role in cell motility remains valid. The relationship between membrane protrusive activity and motility is evident from our observations in cells overexpressing KIF1A-mCherry, where both membrane remodeling/protrusive activity and movement are significantly impaired compared to control cells (refer to Movie S7 vs. S6 and S10 vs. S9).

To address the Reviewer's concern, we have clarified our definition of motility in the introduction by stating on page 19, line 23 – page 20, line 2: "We demonstrate that PLEKHG3 colocalizes with lysosomes at focal adhesion and protrusion sites, and that the localization and function of this protein – and consequently, overall cell motility – are fundamentally dependent on lysosomal dynamics." This revision ensures that our results are accurately described and minimizes any potential confusion. Additionally, we have removed the statement on page 23, line 1 of the original manuscript. We apologize for any confusion our original wording may have caused and appreciate the opportunity to clarify our intentions.

Reviewer 3

1. The mechanism of PLEKHG3 action on lysosomes/late endosomes is underdeveloped in my view. In the absence of for instance mutational analyses to examine what drives the interaction of PLEKHG3 with LAMTOR3, as well as delineation of at least some molecular consequences of this binding, the study remains incomplete.

Answer: We are grateful for the Reviewer's feedback and concur that gaining insight into the mechanistic details of PLEKHG3's interaction with LAMTOR3 would be beneficial. We now consistently refer to PLEKHG3 as a L3 vicinal protein based on TurboID and lack of co-IP. Because L3 KO does not alter PLEKHG3 FA localization and we find no evidence for a stable complex, mutational binding analyses lack a clear readout and are beyond the scope of this revision. We emphasize the conceptual advance—lysosome positioning gates PLEKHG3 cortical enrichment at FAs while peripheral lysosome clustering correlates with more adhesive, less protrusive behavior—and explicitly flag mechanistic questions (e.g., integrin turnover, IQGAP, Rac1 pools) as future work.

We hope that the Reviewer will bear with us on this point, considering the novelty of our findings, which illuminate the interplay between lysosomes and actin dynamics as well as the role of PLEKHG3 in regulating cell protrusions—findings not previously reported in the literature.

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 Ettelt et al describes the identification of PLEKHG3 as a collaborator of the LAMTOR complex on lysosomes using proximity-based biotinylation. The biotinylation screen is well executed and controlled. The authors choose to follow up on PLEKHG3, a protein involved in actin dynamics, which they refer to as understudied (I let the validity of the latter statement to be evaluated by the editor). Generally speaking, the data are of good quality, and the manuscript is clear and well written. However, much of the evidence on the role of PLEKHG3 on lysosomes is suggestive at best and further investigation into its mechanisms of action is warranted. Some important points to address prior to publication are detailed below.

Major Points:

1. The mechanism of PLEKHG3 action on lysosomes/late endosomes is underdeveloped in my view. In the absence of for instance mutational analyses to examine what drives the interaction of PLEKHG3 with LAMTOR3, as well as delineation of at least some molecular consequences of this binding, the study remains incomplete.
2. A key issue possibly (but not necessarily) related to the point above is that the authors focus solely on peripheral lysosomes as target compartments for PLEKHG3. This is not self-evident, particularly in light of images presented in Figures 2 and 3, where colocalization of PLEKHG3 with perinulcear lysosomes appears very likely. The authors should make differences/similarities they observe between effects on perinuclear versus peripheral lysosomes explicit both with data and in the text, if such differences exist.
3. Data presented in Figure 6 showing cell motility analysis is interesting and has potential to make the manuscript impactful. Similarly, data in Figure 4F (live cell imaging) looks attractive but is not informative in the absence of relevant genetic perturbations as comparisons. These types of experiments would benefit greatly from PLEKHG3 loss of function analysis, as well as mutational analysis in the over-expression setting.

Minor point

1. Multicolor overlays with one of the channels in white is in my view not reader-friendly. Appreciating colocalization between endosomes/lysosomes, actin and G is very important for this study, and while is typically reserved to show overlay between green and magenta or green (standard for 2 channels), red and blue (standard for 3-channels). I therefore advise the authors to choose a different color combination throughout the figures when presenting microscopy images.

Significance

In principle, I consider this study to be of interest to the community of cell biologists working on the endolysosomal system and/or the actin cytoskeleton and its relationship to intracellular membranes. However, the authors find themselves in a rather crowded field. I feel that developing the mechanism of action of PLEKHG3 on lysosomes beyond this first submission could help with boosting the impact of the study. There is clearly something interesting going on, but what that is exactly, remains unclear in my view.

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 use proximity-dependent labelling and mass spectrometry to identify cytoplasmic proteins that interact with lysosomes. They show that PLEKHG3 interacts with the LAMTOR complex; that PLEKH3 accumulates in focal adhesion sites, where it colocalizes with peripheral lysosomes; and that the increased translocation of lysosomes to the periphery leads to less "protrusions", as well as rounder cells and less motile cells.

Major comments: While the study is generally carefully performed and thorough, there are major shortcomings that affect the conclusions taken, namely the specificity of the PLEKHG3 antibody, the identification of "protrusions" and ruffles, several quantifications missing, and the data used to conclude about cell motility. There are also conclusions for which there is no concrete or solid evidence.

Specific issues:

1. Specificity of PLEKHG3 antibody: In Fig. S2, authors show that PLEKHG3 antibody recognizes 3 bands (above 100 kDa, above 130 kDa and 250 kDa) and all of them are reduced by the silencing of PLEKH3. Then, in Fig. 2A and C, authors only show the band above 130 kDa, despite implying that the specific band should be "much higher than the 134 kDa calculated from the aminoacid sequence of the protein". In Fig. 2 B, they show all the bands shown in Fig. S2 and presumably favor that the specific and is the 250 kDa one. Finally, in Fig. 2D, they show all bands and note that the band above 130 kDa is not specific. Therefore, authors need to conclude what is the specific band and always analyze the same one, and, possibly, use a different antibody or purify this one to remove non-specific binding. Without this, the main result of the paper, cannot be substantiated.
2. In page 12, authors state that "These results indicated that PLEKHG3 is a transient interactor, or a proximal, not directly binding protein, of L3" and in page 14 that "... PLEKHG3 is a proximal L3 protein rather than a transient physical interactor". It is not clear at all how did the authors reach such conclusions, nor they have data to conclude this. Indeed, they would have to express the proteins in vitro and test their interaction to conclude about a direct binding. They also do not know what is the stability of the interaction.
3. Still in page 12, authors state that "... two different membrane structures, protrusions and ruffles". What do the authors mean exactly by "protrusions", as there are several different ones (e.g., lamellipodia, filopodia, pseudopods)? And how can they distinguish between ruffles and, for example, lamellipodia? They need to use markers and more carefully analyze their morphology to be able to distinguish these. Like this, it is too preliminary.
4. At least Fig. 2F and 3A need quantification. Regarding cell motility, there is no quantification and the authors must perform a quantitative assay (despite stating that "As another measure of cell motility, analysis of the number of forming protrusions and retracting membranes..."). Not only this is not a measure of cell motility, but there the issue of what are "protrusions" referred above. Therefore, authors need to quantify the distance that the cells move and/or perform quantitative motility/migration assays.
5. It is not clear if in cells KO for PLEKHG3, the overexpression of KIF1A leads to more lysosomes localizing close to the PM, as well as more protrusions and more cell motility, as the authors only compare cell overexpressing GFP or GFP-PLEKHGL3.
6. Regarding the statistical analysis, authors assert that it was done using Student's t tests, unless otherwise stated. However, they never refer in figure legends other statistical analysis methods. If so, they cannot use such test, for example, in cases where more than two groups are compared.

Minor comments:

1. In the abstract, authors refer that cytosolic proteins are recruited to platforms on the limiting membrane of lysosomes. What do they mean by "platforms"? Is it microdomains?
2. In the Introduction, there is a period before the reference at the end of the first paragraph.
3. In the results, Fig. 1E is mentioned before Fig. 1D and Figure S1F before Fig S1E, which can be confusing.
4. All the immunofluorescence images need to be bigger, in general, and have zoom-ins, except Fig. 3A, 4B, 4F, and S2C. Also, in Fig. S1F, the green channel has different intensities and the V5-lyso signal is clearly saturated. Finally, Fig. S1D, S1I and S3F must be enlarged, too.
5. In page 9, where it reads "Figure 1K", should read "Figure S1K".
6. The observation that PLEKHG3 silencing leads to loss of the perinuclear clustering of LAMP1-positive vesicles, and increase in their accumulation at the cell tips, is not referred in the text.
7. Fig. 2C is not referred in the legend.
8. Figure S3A and B: authors should show the colocalization of endogenous PLEKHG3 with phalloidin and not only the GFP-tagged protein.
9. In page 14, authors refer to Fig. 3G, which does not exist.
10. In page 30 and page 32, different antibodies for LAMP1 and PLEKHG3 are mentioned, but in the figure legends authors do not refer which one they used.
11. In page 33, where it reads "300 µm protein", it should probably read "300 µg protein".

Significance

The study provides evidence that lysosome positioning can affect cortical actin cytoskeleton dynamics, as well as cell shape and motility. Experiments are in general thorough and data subjected to quantification. However, there are fundamental conclusions that are preliminary at this stage, as some of the data is not yet solid enough. Therefore, it needs to be further strengthened to be considered for publication. In general, it reads well but the amount of abbreviations (e.g. in the case of the constructs) makes it somehow difficult to follow. The study will be interesting for the cell biology, membrane trafficking and cytoskeleton dynamics communities.

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 Ettelt et al., describes identification of Rho guanine nucleotide exchange factor- PLEKHG3 as one of the positive hits from a TurboID proximity-dependent labeling screen using LAMTOR3 (one of the subunits of the pentameric LAMTOR complex/Ragulator) as a bait protein. The authors find that PLEKHG3 colocalizes with lysosomes at focal adhesions and that peripheral clustering of lysosomes promotes PLEKHG3 localization near the plasma membrane, and also inhibits protrusion formation and cell motility. The experiments, particularly the Turbo ID proximity-dependent labeling screen, are well-executed, and the imaging data is aptly quantified. The manuscript explores an exciting question of how lysosome positioning regulates cortical actin dynamics and thereby cell motility.

Major comments:

• The colocalization of endogenous PLEKHG3 and LAMP1 as depicted in figures 3B and 3C (data from fixed cells) is not convincing. PLEKHG3 appears to be present on cortical actin structures as opposed to being colocalized with LAMP1 on lysosomes. The authors should also confirm the specificity of the PLEKHG3 antibody in immunofluorescence using control and PLEKHG3 siRNA in untransfected cells that have not been transfected with GFP-PLEKHG3 (as is shown in Fig. S2C). Numerous antibodies recognize the overexpressed protein but do not recognize the same protein at endogenous expression levels.

Moreover, do the authors observe colocalization between GFP-PLEKHG3 and lysotracker in living cells? There is no apparent colocalization of PLEKHG3 and lysotracker in the movie S5. - The authors observe that GFP-PLEKHG3 is concentrated at the cell's periphery when KIF1A is overexpressed, whereas RUFY3 overexpression results in more cytosolic staining. To bolster their conclusion that a change in lysosomal positioning alters the subcellular localization of PLEKHG3, it is preferable to employ inducible techniques, such as the recently described "reversible association with motor proteins" (RAMP) (PMID: 31100061). The method is a rapid and reversible method for altering organelle positioning. It is still unknown whether PLEKHG3 is associated with lysosomes and mechanism of how positioning of lysosomes affects PLEKHG3 localization. - Similarly to the preceding point, the claim that "peripheral accumulation of lysosomes inhibits protrusion formation and limits cell motility" should be tested more rigorously using the RAMP method, preferably in living cells. Other approaches, such as overexpression/siRNA of Arl8b and other motor adaptors, such as SKIP/PLEKHM2, can be used to alter lysosome positioning and confirm this central findings of the manuscript. The authors could also consider including additional mechanistic data in order to comprehend how lysosome positioning controls cell motility. For instance, the RAMP approach could be employed to investigate cortical actin dynamics upon repositioning of lysosomes to the peripheral/perinuclear region. - It is not clear how the authors conclude that Figure 4E graph shows "the LAMP1 signal was stronger in paxillin-labeled FA compared to control regions". The 4E graph shows LAMP1 signal in GFP versus GFP-PLEKHG3 and shows a modest enrichment of LAMP1 in FAs in GFP-PLEKHG3 overexpression. LAMP1 enrichment in FAs is also not obvious in the image shown in Figure 4B. - In Fig. 2B, there appears to be a labeling error. The lanes 2,4 and 7 appear to be transfected with L3-T-V5 but labeled as GFP-V5-cyto. Here the PLEKHG3 band should be indicated. - Fig. 2C is an IP experiment as per the manuscript text but it is labeled as pulldown.

Significance

Using a TurboID proximity-dependent labelling screen, the authors identified an interesting subset of actin-remodeling proteins that interact with the lysosomal protein LAMTOR3. The authors further characterized one of these proteins, PLEKGH3, and found that lysosome positioning regulates PLEKGH3 localization, as well as plasma membrane protrusion formation and cell motility. This study suggests that lysosome peripheral accumulation could regulate cortical actin remodelling and consequently cell migration by regulating PLEKGH3 localization (although this is not tested in the manuscript). This study adds to the previous findings that microtubule-based transport of late endosomes regulate delivery of late endosomal LAMTOR proteins to the vicinity of focal adhesions, which in turn, regulate focal adhesion dynamics. The mechanism of how lysosomes can influence actin remodeling will be important in the context of cancer cell migration. My area of expertise is lysosome fusion and motility and I have limited expertise in regulation of actin dynamics and how Rho family members regulate actin remodeling.

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 #2

Evidence, reproducibility and clarity

Summary:

In this manuscript, the authors investigate the role of Transcription Termination Factor 2 (TTF2) in the regulation of mitotic transcription. Using siRNA-mediated knockdown in two distinct human cell types combined with nascent RNA labeling (EU pulse), the authors identify an unexpected role for TTF2 in the timing of RNA Polymerase I (Pol I) reactivation following mitosis. The study suggests that this temporal misregulation may have downstream consequences for nucleolar morphology and function in interphase. The manuscript is well-written, and the figures are of high quality and clearly presented.

Major comments:

1. A primary limitation of the current study is that it does not deeply explore the underlying mechanism of the observed phenomenon. To strengthen the claims, the following points should be addressed:

1a. Directionality of Phenotype: In Page 9, the authors conclude that TTF2 depletion is linked to abnormal nucleolar organization during interphase. It remains unclear if this is a direct result of mitotic misregulation or an independent interphase effect. To distinguish between these possibilities, I suggest the following experiment: perform a mitotic shake-off early in the siRNA treatment (~24h), collect mitotic cells, and allow them to re-enter G1 to image for nascent RNAs and nucleolin. This would clarify if the mitotic defect precedes and causes the interphase morphology changes. Alternatively, the authors should state that their current study cannot distinguish between these two possibilities.

1b. Secondary Effects: The long duration of siRNA treatment (48h) raises the possibility that TTF2 knockdown misregulates the expression of other Pol I regulatory factors, leading to secondary effects. This limitation should be explicitly acknowledged in the Discussion. 2. The term "significant" is used throughout the manuscript without accompanying statistical testing.

2a. Please provide statistical analyses (e.g., p-values) for the average plots in Figures 1-3 to substantiate the findings.

2b. Where statistics are not performed, the language should be softened to "notable" or "observed increase" rather than "significant." 3. siRNA knockdowns are generally supported by quantification. Please provide the percentage reduction of the target protein by quantifying the blots provided in the supplemental figures. 4. To rule out the possibility that the increased nucleolin signal observed after TTF2 KD is simply due to higher protein abundance, the authors should perform a western blot to confirm that total nucleolin protein levels remain unchanged upon TTF2 depletion.

Minor comments:

1. The abstract and discussion refer to the role of TTF2 as a "conserved" process. As the study only tested human cell lines, "conserved" is technically inaccurate (as it implies evolutionary comparison). I recommend using "general" or "cell-type independent."
2. While the Methods section is detailed, the Results section would benefit from brief descriptions of the treatments to improve flow.

Example revision (Page 4): "...we treated two distinct cell lines with control and TTF2-specific siRNAs for 48 hours, followed by a 30-minute EU pulse to label nascent RNAs. Click chemistry and Hoechst labeling enabled 2-color imaging of mitotic chromosomes and nascent RNA..." 3. The data generally agree across both cell types; however, the presence of clustered signals in HeLa metaphase chromosomes is a notable divergence. It would be beneficial to include speculation in the Discussion on whether this represents a failure to silence Pol I transcription or an even earlier reactivation, and what this implies about a cancer cell line.

Significance

General assessment:

The study is strong in its use of two different cell systems, providing confidence that the observed effects are not cell-line-specific. The figures are beautifully presented and the writing is clear. The primary limitation is the lack of mechanistic depth regarding how TTF2 specifically interfaces with the Pol I machinery compared to its known roles with Pol II.

Advance:

This work reports a previously unrecognized role for TTF2 in the temporal control of Pol I reactivation. While TTF2 is well-known for its role in terminating transcription and facilitating Pol II release during mitosis, its specific influence on the nucleolar transcription cycle provides a new perspective on how cells transition out of the mitotic state.

Audience:

This research will be of interest to researchers in the fields of gene regulation, the cell cycle, and nucleolar biology. Because it touches on the fundamental process of how transcriptional machinery is reset after cell division, it has implications for the broader basic research community interested in epigenetic memory and cellular identity.

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 paper by the Oliveira lab a new perspective on TTF2 function during mitosis is proposed. Known for its role to terminate transcription at mitotic onset, this paper further shows an exciting involvement of TTF2 to schedule timely rDNA transcription at mitotic exit. Moreover, this role is shown to have a clear importance in the structuration of nucleoli, since TTF2 depletion is associated with premature partial assembly of nucleoli on the mitotic chromosomes and, subsequently, to fragmented nucleoli in interphase. These conclusions, which are well supported by imaging data, are original, interesting and, in fact, largely unexpected.

The paper is very simple in its execution, based on siRNA depletion of TTF2 and monitoring of transcription by imaging using EU incorporation and rRNA-FISH, as well as nucleoli morphology and dynamics using immunostainings. Yet, it is well executed and has no major caveats. However, the authors should consider the following:

1. The Teves lab has shown that TBP is a key factor maintaining its binding at rDNA loci during mitosis, enabling a prompt reactivation of rRNA production in interphase (Kwan et al. RNA 2024). This paper should be discussed on the light of current findings.
2. In relation to the previous comment, I would strongly recommend the authors to analyse TBP-depleted cells, ideally using the line generated in the Teves lab, to address whether delayed rDNA transcription after mitosis leads to delayed nucleoli structuration. This assay would allow them to further confirm their model.
3. In addition, it would be important to test if in the absence of mitotic TBP, the depletion of TTF2 does also lead to mitotic transcription.

Significance

Strengths: originality of the observation and simplicity of the experimental setup

Limitations: exclusively based on imaging data

Advance: completely unanticipated observation

Audience: general readers interested in gene regulation

My expertise: gene regulation through mitosis

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-2026-03407

Corresponding author(s): Laura Cantini, Julio Saez-Rodriguez

[The "revision plan" should delineate the revisions that authors intend to carry out in response to the points raised by the referees. It also provides the authors with the opportunity to explain their view of the paper and of the referee reports.

• *

The document is important for the editors of affiliate journals when they make a first decision on the transferred manuscript. It will also be useful to readers of the reprint and help them to obtain a balanced view of the paper.

• *

If you wish to submit a full revision, please use our "Full Revision" template. It is important to use the appropriate template to clearly inform the editors of your intentions.]

1. General Statements [optional]

This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

We thank both reviewers for their thorough and constructive evaluation of our manuscript.

Reviewer 1 highlighted that the manuscript would benefit from 1) a stronger positioning of ReCoN within the existing literature on multicellular modelling and network exploration, 2) a justification of our methodological choices, including the use of Random Walk with Restart (RWR), 3) the choice of input datasets for GRN inference and an assessment of the robustness of ReCoN's predictions to noise in these networks, 4) a more systematic exploration of ReCoN's parameter space (restart probability, layer transition probabilities, filtering thresholds).

Reviewer 2 raised concerns about 1) the generalisability of the α parameter value (by default, 0.8) across independent datasets, 2) the expected contribution of the indirect effect in prediction performances, 3) the robustness of GRN across datasets and systems, and 4) the need for more quantitative validation in the spatial/microenvironment showcase. They also pointed out an unsupported claim regarding gene knockout prediction in the abstract.

Several clarifications on figures, methods, and writing were also requested by both reviewers.

As the main addition to the manuscript, we propose a new showcase based on the recently published Human Cytokine Dictionary (Oesinghaus et al., 2025). This showcase will simultaneously address several reviewer concerns by allowing us to 1) test the robustness and performance of α = 0.8 in an independent dataset, 2) evaluate the impact of different GRN inference methods (HuMMuS, SCENIC+, CellOracle, GRNBoost2) and noise on ReCoN's predictions..

We will conduct a systematic parameter exploration on the Heart Atlas showcase, covering restart probability and inter-layer transition probabilities. We will additionally strengthen the validation of the microenvironment showcase by providing additional comparison to matched single-cell fibroblast data.

Regarding the manuscript, we will substantially expand the discussion to better contextualise ReCoN within existing multicellular modelling approaches and the methods to justify our methodological choices (RWR/MultiXrank, dataset selection). We will remove the unsupported gene knockout claim from the abstract and reframe it as a future direction. In addition, we will clarify the distinction between ReCoN variants and rename them for clarity in the results section 1.2., improve figure legends. Finally, we will also work on the tool's documentation, including new tutorials on using spatial data and on running ReCoN with scRNA-seq-only GRN inference.

We believe these revisions will substantially strengthen the manuscript and address the reviewers' concerns regarding method's robustness, generalisation, and contextualisation.

2. Description of the planned revisions

Reviewers' comments are in blue

Authors' answers are in black

Proposed text modifications are in green

Reviewer #1

R1.1. This is a very well-written paper; the methods used are adequate, and the use cases are relevant and broad, exploiting state-of-the-art datasets and tools.

The author's claims are mostly justified. The authors could make an effort to more explicitly cite other efforts in similar directions. The claim 'We envision ReCoN as an extension to prior multicellular modelling, offering an interesting compromise between prediction of cell type responses and understanding of their molecular coordination.' is very general and could be better substantiated. In fact, the authors do not really give examples of alternative approaches to study systems of interacting cells, other than mechanistic agent-based models, which are clearly very different.

Response:

We thank the reviewer for pointing out the lack of contextualisation for ReCoN in this closing discussion.

We wanted to remind that ReCoN builds notably on multicellular factor decomposition methods. We also want to emphasise the interest in completing cell communication methods that describe the big picture in multicellular interactions.

• *

We proposed to *explicitly state these two points with such rephrasing: *

• *

Network-based representations of multicellular systems have been an active field for many years, from early conceptual cytokine networks (Frankenstein, Alon, and Cohen 2006) to curated ligand-receptor cascades of hematopoietic tissue (Kirouac et al. 2010, Qiao et al. 2014). In parallel, and from bulk RNA-seq, the consideration of tissue specificities in GRN inference has been another way to consider the importance of the context in molecular mechanisms reconstruction (Sonawane et al. 2017). Single-cell analysis allowed decomposing tissue composition and quantifying gene expression, opening the possibility of scaling the inference of these networks and the inference of multicellular mechanisms in general, to large sets of molecules. Several methods have been developed to recover multicellularity. A first direction extends ligand-receptor interaction inference into the receiver cell response through curated signalling cascades, yielding ligand to target cascades (Browaeys, Saelens, and Saeys 2020, Jin et al. 2021, Zhang et al. 2021, Yan et al. 2025). A second direction leverages spatial context through explainable multi-view models that decompose marker variation in both intra- and intercellular contributions (Arnol et al. 2019, Tanevski et al. 2022), without considering the mediating cascades. Finally, the more recent family of multicellular factor decomposition methods focuses on the coordinated aspect of cellular programs rather than on the mechanisms. ReCoN's methodology proposes a network-based approach based on single-cell data and the philosophy of this last group of methods. Indeed, ReCoN aims to retrieve links between molecular drivers and such coordinated multicellular programs by bridging and exploring CCC inference and GRN modelling (Badia-i-Mompel et al. 2023) within large and coherent heterogeneous multilayer network.

Arnol D, Schapiro D, Bodenmiller B et al. Modeling Cell-Cell Interactions from Spatial Molecular Data with Spatial Variance Component Analysis. Cell Rep 2019;29(1):202-211.e6. https://doi.org/10.1016/j.celrep.2019.08.077.

Badia-i-Mompel P, Casals-Franch R, Wessels L et al. Comparison and evaluation of methods to infer gene regulatory networks from multimodal single-cell data. Preprint, bioRxiv, 21 Dec. 2024, 2024.12.20.629764. https://doi.org/10.1101/2024.12.20.629764.

Badia-i-Mompel P, Wessels L, Müller-Dott S et al. Gene regulatory network inference in the era of single-cell multi-omics. Nat Rev Genet 2023;24(11):739-54. https://doi.org/10.1038/s41576-023-00618-5.

Browaeys R, Saelens W, Saeys Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods 2020;17(2):159-62. https://doi.org/10.1038/s41592-019-0667-5.

Frankenstein Z, Alon U, Cohen IR. The immune-body cytokine network defines a social architecture of cell interactions. Biol Direct 2006;1(1):32. https://doi.org/10.1186/1745-6150-1-32.

Jin S, Guerrero-Juarez CF, Zhang L et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun 2021;12(1):1088. https://doi.org/10.1038/s41467-021-21246-9.

Kirouac DC, Ito C, Csaszar E et al. Dynamic interaction networks in a hierarchically organized tissue. Mol Syst Biol 2010;6(1):MSB201071. https://doi.org/10.1038/msb.2010.71.

Oesinghaus L, Becker S, Vornholz L et al. A single-cell cytokine dictionary of human peripheral blood. Preprint, bioRxiv, 15 Dec. 2025, 2025.12.12.693897. https://doi.org/10.64898/2025.12.12.693897.

Qiao W, Wang W, Laurenti E et al. Intercellular network structure and regulatory motifs in the human hematopoietic system. Mol Syst Biol 2014;10(7):MSB145141. https://doi.org/10.15252/msb.20145141.

Radig J, Droit R, Doncevic D et al. Tracking biological hallucinations in single-cell perturbation predictions using scArchon, a comprehensive benchmarking platform. Preprint, bioRxiv, 27 June 2025, 2025.06.23.661046. https://doi.org/10.1101/2025.06.23.661046.

Sonawane AR, Platig J, Fagny M et al. Understanding Tissue-Specific Gene Regulation. Cell Rep 2017;21(4):1077-88. https://doi.org/10.1016/j.celrep.2017.10.001.

Tanevski J, Flores ROR, Gabor A et al. Explainable multiview framework for dissecting spatial relationships from highly multiplexed data. Genome Biol 2022;23(1):97. https://doi.org/10.1186/s13059-022-02663-5.

Yan L, Cheng J, Nie Q et al. Dissecting multilayer cell-cell communications with signaling feedback loops from spatial transcriptomics data. Genome Res published online 12 May 2025. https://doi.org/10.1101/gr.279857.124.

Zhang Y, Liu T, Hu X et al. CellCall: integrating paired ligand-receptor and transcription factor activities for cell-cell communication. Nucleic Acids Res 2021;49(15):8520-34. https://doi.org/10.1093/nar/gkab638.

R1.2. Moreover, the exploration of the multilayer networks with RWR is a very reasonable choice but could there be other approaches? I think the authors could discuss this issue to briefly support their choice of this method.

Response:

It is a very relevant comment, as this choice has not been discussed in the paper; we propose extending the method section about ReCoN's networks exploration with a justification about this choice.

• *

There is currently a limited set of network exploration methods that have been implemented for multilayer networks. It includes notably pymnet (Nurmi et al., 2024), natively adapted to heterogenous multilayer networks, and multinet (Bagavathi et al., 2019) and muxviz (De Domenico et al., 2015), initially developed for multiplexed networks (e.g. social network where the same set of nodes is present in each layer) but adaptable to more complex multilayer networks. However, to our knowledge, only MultiXrank proposes a robust measurement of proximity between each pair of nodes.

Indeed, pymnet does not propose implementation for pairwise distance, similarly for muxViz, which focuses on community and motif detection. Multi-net does propose pairwise distance based on shortest paths, but implements it only for nodes of the same multiplex (e.g. in our network, it would only be two genes, or two receptors, respectively). https://www.rdocumentation.org/packages/multinet/versions/4.3.2/topics/multinet.distance

• *

We provide the additional justification for choosing RWR and MultiXrank over a reimplementation of another method or an extension of another method.

• *

• The total complexity of the RWR is O(δm) - when the number of nodes is negligible compared to the number of edges, with m the number of edges and δ the number of iterations in the walk (Baptista et al., 2022 - Supp Notes 2.A; Jin W. et al, 2019). This linear increase with the number of edges is particularly interesting for large networks, such as ReCoN ones that can contain several million* edges. The number of iteration δ and the computational time increases inversely to the restart probability, which is an important factor to keep this probability high. *

• *

• *MultiXrank is particularly interesting for its flexibility as it allows to easily attribute different weights to the different layers and to precise the direction of the exploration easily. *

• *

• It also produces deterministic results by prolonging exploration until convergence.

• *

• Additionally, in the context of ReCoN, the indirect effect of each cell is run independently. We previously extended the implementation of multiXrank for running RWR in parallel in a previous work (Trimbour et al., 2024), making it already adapted for optimising ReCoN's explorations.

• *

For all these reasons MultiXRank implementation seemed to be the best choice for robust and efficient exploration of ReCoN's HMLN.

• *

Bagavathi, A., Krishnan, S. (2019). Multi-Net: A Scalable Multiplex Network Embedding Framework. In: Aiello, L., Cherifi, C., Cherifi, H., Lambiotte, R., Lió, P., Rocha, L. (eds) Complex Networks and Their Applications VII. COMPLEX NETWORKS 2018. Studies in Computational Intelligence, vol 813. Springer, Cham. https://doi.org/10.1007/978-3-030-05414-4_10

Manlio De Domenico, Mason A. Porter, Alex Arenas, MuxViz: a tool for multilayer analysis and visualization of networks, Journal of Complex Networks, Volume 3, Issue 2, June 2015, Pages 159-176, https://doi.org/10.1093/comnet/cnu038

Nurmi et al., (2024). pymnet: A Python Library for Multilayer Networks. Journal of Open Source Software, 9(99), 6930, https://doi.org/10.21105/joss.06930

Jin, Woojeong, Jinhong Jung, and U. Kang. "Supervised and extended restart in random walks for ranking and link prediction in networks." PloS one 14.3 (2019): e0213857

R1.3. Generally the discussion should provide the reader the context in the existing literature in which the work can be set, detailing its impact. I think this could be improved.

Response:

• *

We hope that the correction on the context proposed for comment R1.1 offers a first clarification on the context in the literature.

• *

We also propose to extend the description of ReCoN's impact with the following sentences in the discussion: "Unlike purely data-driven approaches, ReCoN contextualizes prior knowledge balancing both robustness through literature data, and specificity through new measurements. This mechanistic approach opens new possibilities for understanding how cellular coordination shapes tissue-level responses and for designing targeted molecular interventions."

• *

R1.4. Regarding the choice of datasets, it is clear that the method is quite demanding, requiring single cell and different omics to build the model, in addition to the expression dataset that is used as a use case. This inevitably leads to using a mix of datasets.

For example in the mouse experiments the gene regulatory network was inferred from both a lymph node scRNA-seq dataset and a splenic scATAC-seq dataset, presumably due to the lack of multiome data in this setting. However the cell-cell communication network was inferred from the control case of the Immune Dictionary. Why can't the authors use the control data also for inferring GRNs?

Is atac-seq really necessary in the inference of the GRN? What is the impact of the fact that lymph node and spleen samples might be different?

:

• *

Is it a very *interesting comment, and we propose to add both 1) an explanation about our dataset choice to generate the GRN as a Supplementary text, and 2) a new experiment about the effect of GRNs built from multi-omics and scRNA-seq alone. *

• *

• Dataset choice

• *

We decided to infer a GRN using multiomics data, as these methods seem to perform better and are becoming the state of the art (Badia-i-Mompel et al. 2023, Trimbour, Deutschmann, and Cantini 2024, Yuan and Duren 2025).

As scATAC-seq data was not produced for the Mouse Immune dictionary, we tried to find an external dataset, used HuMMuS, the method we previously developed, as it is also based on RWR and performs well on unpaired data.

• *

scATAC-seq

Our first criteria was to match the mouse model used in the immune dictionary dataset, which reduced importantly the number of multicellular immune cell datasets available. We extended our research to a splenic dataset, as spleen is itself classified as a high specialised lymphatic structure, (check) and contains notably the same cell types than classical lymph nodes.

• *

scRNA-seq

While we could technically use the control mice of the Immune Dictionary single-cel RNA-seq data with the spleen scATAC-seq data, the Immune Dictionary only provides 100 or less cells for each cell types per stimulation, which would results in a low number of cells. As GRN quality seems to depend a lot on the number of cell used, we favoured choosing a larger dataset.

• *

Our choice to use single-cell multiomics methods was driven by the novelty of these methods over scRNA-seq based ones, the performance improvement that they seemed to offer in several benchmarkings, and the will of developing a pipeline integrating the most complete data available for contextualization (Badia-i-Mompel et al. 2024).

• *

• GRN impact over the Human Immune Dictionary

• *

While it does not relate directly to this showcase, we will also add a new dataset analysis, detailed in the the comment R1.12. In the Human Cytokine Dictionary showcase,, we propose exploring the effect of choosing different GRNs, built from external multi-omics data or from the control scRNA-seq data of the dataset itself. We hope it can partially help users to decide in general wether to use external datasets of higher quality or sample-specific datasets.

• *

Finally, we propose to add in the documentation of the tool, a section showing how to use ReCoN with only scRNA-seq for the GRN inference, and the performance of different GRNs for the Human Cytokine Dictionary dataset directly in the paper.

• *

R1.5. The code is very clear, we were able to install and run it and it is quite well-documented. However, a few more details should be given in the text regarding how the evaluation of the performance is carried out.

For example: If I understand correctly, when predicting the impact of cytokine perturbations the ReCoN predictions of genes impacted are compared to differentially expressed genes identified through traditional DEG analysis. What is compared is the ranking of these genes from ReCoN with the ranking provided by DEseq2. There is no description of how this comparison of ranking gives rise to AUROC values. Also, is it just the ranking that is predicted or can they also estimate how well they can predict the effect size?

Response:

• *

We are thankful for pointing out the unclear technical details. DEG results were binarised, to obtain the list of differentially genes using the thresholds indicated in the section 4.4.4. We considered a gene as perturbed in each cytokine treatment if the comparison of control and treated cells had a t-test p-value below 0.1 and if the log-fold change was above 1.

• *

The second, and more general point of the reviewers, ReCoN scores should be considered to provide ranking on the possible regulations, but cannot be considered proportional to the effect size. As they are represent a likelihood more than a score, the binarisation should be the most appropriate transformation for the validation

• *

*Moreover, as the scores can be seen as the probability to end up the exploration on each node, they are always summing to one. This also prevents interpreting the scores as the amplitude of change. As an illustration example: if a receptor regulates three genes identically, they would (hopefully) all be having a score of (1 - R)/3, R being the restart probability in ReCoN, whether their expression doubles or is multiplied by 10. *

• *

While it can legitimately be seen as a downside, we believe it is similar in practice to most methods inferring GRN methods in practice, where trying to predict the true amplitude of gene perturbations usually results in very low performances (Badia-i-Mompel et al. 2024).

• *

We propose changes related to this comment.

• *

• We would modify the section 4.4.4. of the method with the following paragraph to explicit that it consists in a binary selection: "For each cytokine-cell type pair, differentially expressed genes were binarised: genes passing the significance thresholds (FDR P-val 1) were labelled as positives, and all remaining genes as negatives. ReCoN scores were then used to rank all genes, and AUROC values were computed from this ranking against the binary labels."

• *

• We will also include a section "ReCoN scores interpretation" on the documentation website, as score interpretation precisions will be particularly useful for users.

  R1.6. When describing the use cases, I think a bit more detail would help.

For example 'To identify the cell-type-specific genes associated with HF, we used the MOFAcell scores of the multicellular factor 1 (MCP1) reported in ReHeat236' I supposed the explanation is on the dataset but for the sake of clarity it would be good to expand this sentence to give at least an idea of the approach.

Response:

• *

We completely agree that more explanations should be provided, to avoid for the reader having to switching between articles to understand the concepts behind this showcase. As suggested by the reviewer, we propose a general description of the approach with the short paragraph, and to remove the term "loading":

• *

"In the ReHeat2 study, the first multicellular factor (MCP1) was associated with heart failure. We used the gene loadings of MCP1 as a proxy for the cell-type-specific transcriptomic changes associated with heart failure, ranking genes by their absolute loading values."

• *

We also propose to complete the method section: "MOFAcell is a multicellular factor analysis method that decomposes multi-sample single-cell data into latent factors representing coordinated gene expression patterns across cell types. Each factor is characterised by cell-type-specific gene scores, reflecting their individual contribution to the coordinated program. In this showcase, we use the first multicellular program (MCP1), as it was associated with heart failure"

R1.7. Regarding the calculation of the R matrix from the NichNet matrices L and G, I gather that the R matrix is calculated once and is thus fully data-independent and available just like the L and G matrices from NichNet. This was not very clear in the tutorials.

Response:

• *

We are very thankful for the reviewers' involvement in testing the tools itself and its documentation. First, we propose a new website page explaining the pre-computed resources available for receptor - gene links, and added a descriptive paragraph in the tutorial themselves.

*Second, we notice a typo in the equation, where it should actually be L = R * G with the current definition. We corrected it in the next version, and precised that R is fully data independent and solely inferred from prior knowledge. *

R1.8. Also, this might just be a typo in the tutorial: 'The default α = 0.8 gives more weight to direct effects, which has been empirically validated. You can adjust this based on your biological question." I believe the manuscript says alpha>0.5 refers to indirect effects dominating.

Response:

• *

We corrected the saying in the tutorials. Indeed, a high alpha represents a stronger indirect effect. Additionally, a similar typo was in the first equation of the paper, we are correcting it too.

R1.9. Same for the pre-processing of the spatial data for the third use case, a little more details on how this was done would help the users and readers.

Response:

• *

We propose adding a specific section about the spatial pre-processing and analysis in the methods.

We are also adding a tutorial on spatial data. Since spatial data processing is computationally intensive without GPUs, we will also provide the data already processed, in order to allow anyone to test this tutorial too.

• *

R1.10. I don't see issues with the statistical power of the analysis.

Rather, I think the authors should provide some examination of the parameter space for their model. Whereas ana analysis of the impact of the Alpha parameter is provided, I believe there are several more parameters that have a crucial impact and choices for their values should be discussed.

For example 'In the GRN reconstruction only the links with a score above 1.5e-7 were retained in ReCoN's gene regulatory layer. How was this chosen?

We have identified the following parameters that are somehow justified but could be explored to have a better feel for how they impact the results

Restart probability: How often the walker goes back to the starting seed/molecule

Layer transition probability: How often the walker stays in the same layer - different cell? - different layers? Gamma

Node transition within a layer: How often one jumps to a different layer

Response:

This is a very valid point raised by the reviewer about parameters explorations.

• *

We focused on exploring the alpha (direct/indirect effect) parameter, as its value was the incertitude when designing the model.

• *

We would like to address this comment by adding new explorations for the restart probability and the transition probability between layers. The probability to transition between specific nodes inside a layer directly depends itself on 1) the restart probability, 2) the transition probabilities, and 3) the weights of the edges, that are determined before and independently to ReCoN's exploration.

• *

The Heart Atlas showcase allows to evaluate each set of parameters in around 10 min instead of 10h for the Immune Dictionary. We thus propose to evaluate restart probability and layer transition probabilities on the data of this showcase.

• *

• We would explore the restart probability of 0.1 * N, with N between 1 and 9.

• *

• For transitions probabilities we propose varying GRN, receptor, and cell communication importance with the following configurations: - Staying in CCC probabilities (- not jumping to receptor layer) among (0.1, 0.3, 0.5, 0.7, 0.9), staying in receptor layer (- not jumping to GRN) of (0.25, 0.5, 0.75), staying in GRN layer (- not jumping to CCC) of (0.25, 0.5, 0.75). It would result in 9 intracellular variations combined with 5 intercellular variations.

• *

We envision an evaluation by measuring the correlation between the results of the different configurations, and the time before convergence of the results, as it could potentially increase drastically when decreasing the restart probability. If correlations below 0.9 are observed between some results, we will compare their absolute performances.

• *

We would include the figures related to these explorations in the supplementary data. We would highlight the main findings in the method section dedicated to the random walk with restart. Finally, we would briefly describe the parameter exploration design in the first section of the results, for curious readers who would like to verify parameter choice before reading the showcases.

• *

R1.11. Weighting parameters: How much weight for direct or indirect effect to account for the combined effect - alpha - this is the only one that is explicitly explored.

Response:

We are very thankful for this comment, and we decided to modify our tutorial guidelines to make this choice more intuitive and general.

• *

Indeed, 1.5e-7 would hardly make sense for most methods, which would not produce such low scores. We now propose to select the first 2 million connections of GRNs, in order to keep a complete or a large portion of the network if other methods than HuMMuS are applied.

• *

In our case, 1.5e-7 was empirically determined from the distribution of HuMMuS scores, to keep the 2 million top connections as HuMMuS networks are generally almost fully connected, which is a particularity for classical GRN inference methods, and keeping it entirely would make exploration time much longer.

• *

R1.12. Finally, this might be considered OPTIONAL but would greatly improve the work in our opinion:

The method crucially depends on the networks that are used in the different layers and to connect layers and cell types. As we know, biological data is noisy and incomplete (FP and FN) at each level and in each datatype. It would be really useful to estimate what is the robustness of the results to this noise. Particularly, from personal experience, we think the GRNs reconstructed from data are often almost fully connected and it is exceedingly difficult to validate them in specific contexts. This means that some 'errors' are likely to be present.

Since several methods exist for inferring GRNs one could simply compare the results using different methods for this part of the network.

A related point involves the characteristics of the RWR algorithm, that will be quite impacted by the presence of hubs in these networks (either in single layers or across several) that is likely to impact the exploration. If proteins that are hub are effectively important, that is not a problem, but in some layers, for example, the receptor-receptor layer that presumably will contain PPIs, there might be biases in hubs being just better studied proteins, and these hubs might have an 'unjustified' weight in the walks.

One potential approach to assess the robustness of the method to these issues could be an empirical one that just randomly perturbs the networks in ReCoN to see to what extent similar predictions are achieved.

*Response: *

• *

We are thankful for this relevant comment on GRN and prediction stability, and would like to take it as an opportunity to support the hypothesis that different GRN methods can be used in ReCoN.

• *

When developing our previous HMLN-based tool, HuMMuS (Trimbour et al. 2024 - Supp Figure 6), we observed that its multilayer structure provided more robust results than individual layers. We would like to reproduce such an analysis, verifying that ReCoN results have less variability than the GRN layers individually.

We propose to integrate a new showcase on the Human Cytokine Dictionary (Oesinghaus et al. 2025), trying to predict cytokine downstream effects similarly to the Mouse Immune Dictionary showcase.

This showcase would be useful to confirm the contribution of the indirect effect and test the impact of different GRN on the results.

We would generate different GRN with several other GRNs methods: SCENIC+, CellOracle, and GRNBoost2 - the latest using only the scRNA-seq of the control samples in the Human Cytokine Dictionary.

• *

The GRN methods produce generally output with very low overlap (Badia-i-Mompel et al. 2024)*. *

*If we observe high correlations between the ReCoN predictions associated with the different GRNS, it would provide already a validation of ReCoN's robustness to GRN noise. *

If lower correlations between ReCoN's predictions are obtained, we will add a specific permutation experience over the HuMMuS GRN, creating different level of artificial noise and assessing more precisely the robustness of ReCoN to GRN stochasticity.

• *

Regarding PPI hub justification, our *applications did not use receptor PPI and are not affected by bias at this level in the showcases. This bias could specifically be present in the receptor-gene links, as we derive it from the ligand-gene connections of Nichenet which was itself partially based on prior knowledge. It is thus possible that some receptor are reached more often due to this bias and not a stronger effect. It seems however, hard to control in this context, as ReCoN currently relies on this prior knowledge. Currently, we hope that the combination of personalised, literature-agnostic GRN with literature-based receptor - gene can provide an interesting trade-off. In future development, we could imagine a receptor-gene network based solely on perturbations, but it would require controlling also the bias of ligand - receptor binding couples, which limits even the use of ligand-based experience. *

We propose adding a short point in the discussion about hub effects from RWR-based methods.

• *

R1.13. Please add page numbers.

*Response: *

• *

We will add the page numbers.

• *

R1.14. Figures are nice and clear.

Some specific minor points are listed here below.

Define hMLN on first appearance fig1 caption (no page numbers..

2nd appearance heterogeneous multilayer structure (HMLN) ...

Response:

• *

We updated the legend of the figure to include the definition of the acronym, as it arrives before first text occurrence. (Or define at both positions ?)

R1.15. Bi_j not so clear to what it refers when first mentioned

Response:

• *

*Bi_j represents a weight that can be attributed to favour some cell-to-cell transitions. It is usually not necessary to use them.

*

*It is of interest notably to model 1) known spatial patterns in situ and hypothesis/design where cell types favour some connections. *

• *

E.g.: for modelling the skin, a user might notably want to increase connections between epidermic and dermic cells, and between dermic and hypodermic cells.

• *

We propose a new explanation of Bi_j to both explain it's meaning in the modelling, and illustrates situations for using it: "The coefficient B_{i,j} modulates the influence of cell type i on cell type j in the indirect effect computation. By default, all B_{i,j} are set to one, weighting each cell type's contribution equally per cell. However, it can be adjusted to encode additional biological knowledge, such as spatial proximity between cell types or known cooperation patterns. For instance, when modelling the skin, a user might increase B_{i,j} between epidermal and dermal cells, and between dermal and hypodermal cells, to reflect their spatial organisation."

R1.16. personalized interaction specificity. - maybe better word than personalised (contextualised?)

Response:

• *

We agree that contextualised explicits better the meaning behind this model. Personalised might notably lead to expect patient-specific data, which is not the case here.

• *

We propose to rephrase all the model names to : Receptor-matrix, ReCoN-no-CCC, ReCoN-no-context, ReCoN-complete.

R1.17. ReCoN-genetic and ReCoN, ( generic?)

Response:

• *

We will correct this typo.

R1.18. responses. It is expected to observe common behaviors in-between cell-type, that the GRN

and the generic CCC network already contribute captures.

• not very clear

Response:

• *

We aimed here to provide an explanation to the already good performance of the "ReCoN-no-context" (or its name updated according to comment R1.16), which could be surprising as no cell-type specific information is used. The explanation proposed is the good prediction of several properties shared by all immune cell types, such as similar metabolic pathways, despite their specific roles. If we adopt a quantitative view on their transcriptome like in this showcase, it can be expected that the cell type responses are relatively well predicted through the common properties only.

• *

As this is a very relevant comment, and that several comments pre-submission we received were also related to this result, we would like to keep an explanatory sentence.

• *

R1.19. Figure 2b the icon of cells with double arrows might suggest phenotype shift when instead this is just communication

Response:

(left side) We are very thankful for paying attention to the details of the paper and fully agree with this analysis. We propose to represent ligand emission instead of arrows, reusing the convention of the Figure 1.

R1.20. eTACs explain acronym and what they are

Response:

• *

We update the first occurrence of eTACS to extrathymic Aire-expressing cells (eTACS).

R1.21. Due to very few genes being differentially

expressed, only cDC1 was conserved and evaluated for IL22,

Not so clear

Response:

• *

As we are commenting on IL22 stimulation results, we reorganised the sentence to make it less convoluted: "For IL22 stimulation, only cDC1 presented enough genes being differentially expressed."

R1.22. In this showcase (not very clear, use case?)

Response:

• *

We perceive "use case" as describing a type of use for the method, while a show case is a specific example of a use case. We thus find showcase more appropriate here. We will however go over all use of the word, to be sure it is only used for the precise examples we provided, and not to describe "use cases".

R1.23. different fibroblast specializations - maybe phenotypes?

Response:

• *

• • It is a very good suggestion, as specialisation would involve functional aspects (that we can't really be sure of), and a chronological evolution*
• Phenotype generally includes numerous properties, such as morphology, that we cannot validate here. We think the use of phenotype might be stronger than specialisation here. To simplify, phenotype can work, to be more precise: transcriptomic specialisation? I am honestly not sure of the best change here.

R1.24. Figure 4b

1. b) Schematic view of the deconvolution process and cell type-specific count inference from the spatial niches.

Not so clear what the heatmap shows, rows and columns

Spots heatmap : label niche on rectangles in cols

And each col is a spot

Rows are cell types or cells?

In the cell types x spot

Response:

This figure can indeed benefit strongly from legend modifications. On both matrix, lines represent the genes, while columns represent the spot / individual cells deconvoluted per spots

• *

• We would annotate the niche legend (here the colour surroundings) by a symbolic drawing instead of writing it on the matrix

• *

Legend "genes" on the first matrix

• *

Write deconvolution ON the figure directly

R1.25. Cell2location. Add reference, maybe explain basic functionality?

Response:

• *

Cell2location was not referenced in the results section, and was only referenced in the section 4.6.2 of the methods, as the 72th citation. We corrected this oversight, and propose 1) a brief explanation of deconvolution right before, 2) a brief explanation of Cell2location particularity in inferring individual cell profiles - which is not common in spatial deconvolution.

R1.26. reconstructing different patients, tissues, and microenvironments to predict

context-specific molecular treatments.

Unclear

fibrosis in different - at

molecular levels

Response:

• *

We will modify this section title according to the reviewer's citation and the different reformulation.

R1.27. Figure 5d myeloid and endothelial colour code inversed from 5 BC

Response:

• *

The legends are individually correct, but there is no reason to not make them coherent across panels. We will update the legend of the panel 5.d..

• *

R1.28. 5d indicate important pathways in organe should not change the colour of the nodes (purple=common, blue or green specific). Use border colour maybe?

Response:

• *

We had forgotten to precise the colour code of this panel, where the choice of orange highlighted here the gene set related to molecular pathways instead of functional annotations. As the name already explicits pathway, we now think that the orange background is redundant informations and may create some confusion. We thus would like to update Wnt and TNFA pathways backgrounds to ___ (more enriched in cell type), and purple (significantly enriched in all cell types).

R1.29. 5e is not a venn diagram

1. e) Venn diagram showing the overlap between transcription factors (TFs) predicted by ReCoN (green) and those previously

implicated in fibrosis (orange) or cardiac diseases (violet). Only the top 10 TFs were annotated from literature

sources; full sizes of fibrosis- and cardiac disease-related receptor sets can therefore not be represented.

1. f) also not a venn diagram e/f now in supp

the "NABA ECM collagens" gene set. Nodes are

grouped by molecular type (e.g., transcription factors, receptors, ligands), and links represent the weighted,

direct regulatory interactions present in the ReCoN-constructed

Response:

• *

As the diagrams do not indicate the total number of receptor/TF that are in the literature, it cannot be Venn diagrams. We updated the legend to :Venn diagram showing the Overlapp between [...]

• *

As we reorganised the paper, these plots are now only in supplementary; we removed the duplicate occurrence in the figure 5 legend.

R1.30. Why Sankey plot? Normally sankey plot represents flow (of regions changing from 1 state to another) but here this is just a weighted network?

No communication from firbos back to other cell types? No communication between ventricular/myeloid/lymphoid?

Response:

• *

We are thankful for this useful feedback which helped us realising interesting details were missing from the paragraph.

• *

*This is only intended for visualising regulatory cascade, so users have to decide on one receiving cell, a set of target genes, and sending cells. It includes a specific subset of regulatory cells, and only their interactions with the target cells. Here, we illustrated the regulation of some ECM genes produced by fibroblast. *

• *

Sankey Diagram might indeed not be the clearest representation, as we are not modelling the all diffusion, and not a flow per se. We propose to replace by another representation that we hope will be more intuitive for biologists (and more aesthetic), such as illustrated below:

R1.31. as a extension to - an

underrepresented in the current. - current framework?

Response:

• *

framework works perfectly to fill the missing word in the sentence

• *

R1.32. However, it can't represent more - cannot

Borrowing representation from hypergraphs, which introduces

The network exploration implementation of ReCoN also present some limitations.

limitations. While random walks

with restarts offer a stable and fast exploration workflow for multilayer networks, it

currently only considers positive weights to predict regulation strengths. It involves that the

nature of the regulation, as activation or inhibition, has to be identified a posteriori.

• check concordance/grammar

Response:

• *

We will update the raised grammatical errors

• *

R1.33. Only the nodes that are included in one of the layers are present in the

final results, ignoring the ones present only in bipartites.

Unclear

Response:

• *

Layers and bipartites are treated differently by the algorithm, and layer presence is necessary to appear in the results.

• *

In practice, it just means that receptors/ligands not paired in the CCC, or genes not regulated by any TF in the GRN, won't appear.

• *

We propose clarifying with this second explanation

• *

"In practice, a node must have at least one connection in its layer to appear in the final results. It thus means that receptors or ligands absent from the CCC network and genes not targeted by any transcription factor in the GRN will not receive a score from the random walk exploration."

• *

R1.34. a scATAC - an

• *

Barsi et al is published https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013188

Response:

• *

We updated the reference with the published article.

R1.35. effects, allowing for modulating in a second

time their contribution. - word order

Response:

• *

We propose to formulate "allowing in a second time to modulate their contribution"

R1.36. others. However, it is possible to adjust the Beta coefficient to

represent it based on the available information for each dataset.

Represent- adjust?

Response:

• *

We agree with the reviewer's suggestion to use adjust.

R1.37. We use the latter to compare the different models. - what is the latter?

Response:

• *

The latter referred to the 25 cytokines of the Immune Dictionary which had at least one connection in the inferred cell communication network with CellPhoneDB. We propose clarifying this formulation to "..."

R1.38. It resulted in the scRNA-seq in 1,789 cells with 13,167

genes, and for the scATAC-seq in 3,759 cells with 254,545 regions.

Check english

Response:

• *

We propose replacing this sentence by the following: "It resulted in a scRNAseq dataset of 1,789 cells with 13,167 genes, and a scATACseq dataset of 3,759 cells with 254,545 regions."

R1.39. GRETA pipeline.- reference

Response:

• *

We added the citation to the paper of the GRETA pipeline in the section 4.5 of the methods: "Badia-i-Mompel et al., 2026"

R1.40. We kept all the cells whose annotations through unsupervised clustering,

followed by marker gene annotations, through scANVI were coherent.

Word order

Response:

• *

We propose the following reformulation to correct the sentence: "We kept all cells whose annotations were coherent between unsupervised clustering with marker-gene labelling and scANVI-based label transfer"

R1.41. In parallel, pairs of ligands and receptors with both associated with scores above

an absolute gene loading of 0.1 were considered potential driver interactions in HF.

Unclear

Response:

• *

In the MOFAcell results, factors correspond to linear combination of genes that explain a large part of the data variance; the contribution of each gene is called loading. We chose the factor that classified the best patient with and without fibrosis, and kept all the top genes, all of those with a score above 0.1.

• *

We propose reformulating this sentence as the word "loading" could overcomplicate here for most readers: "To identify the ligand and receptors driving heart failure, we considered all of those with an absolute contribution to the multicellular factor of 0.1."

R1.42. gseapy Python - reference?

Response:

• *

The gseapy package was indeed not cited, we now include the citation : "Zhuoqing Fang, Xinyuan Liu, Gary Peltz, GSEApy: a comprehensive package for performing gene set enrichment analysis in Python, Bioinformatics, 2022;, btac757, https://doi.org/10.1093/bioinformatics/btac757"

R1.43. and to calculate average for each spatial context the average cell type expression.

Unclear

Response:

• *

we propose to reformulate the sentence to: "These cell-type-spot profiles were used later for each spatial context to create a specific cell-cell communication networks and to calculate cell type average expressions."

R1.44. We only used the loadings of all cell

types but the fibroblasts to consider the effect of the sole environment.

Unclear

Response:

• *

we propose to use "APART from the fibroblast" to clarify the sentence and "to ONLY consider the environment effect".

R1.45. We realised a downstream - performed

Response:

• *

We fully agree with the reviewer's suggestion.

R1.46. The profiles inferred by ReCoN were first very correlated in all three contexts. - unclear

Response:

• *

The sentence was missing clarity and deserved being rephrased. We propose: "When looking at the absolute scores of ReCoN in all three contexts, results were initially highly correlated. To focus on context-specific differences, enrichments were performed using the log-ratio of each context profile over the mean of the other profiles."

• *

R1.47. Potentially the closest results are models that can predict the effect of perturbations on cell line cultures. Several approaches in the literature employ either transformers or optimal transport to predict the effect of perturbations in single cell datasets. One of the main issues is an underlying necessary assumption that the perturbation effect will be larger than the heterogeneity (in cell lines for example), which becomes increasingly difficult when considering in-vivo experiments. ReCoN obviously goes beyond this by considering explicitly the presence of different cell types but distinctions of cell types are sometimes quite arbitrary and potentially application of ReCoN to some of the in-vitro culture datasets, even on cell lines, could be a way to test its performance and benchmark it against other methods.

The main bottleneck in the application of this framework to 'personalisation' of therapies, mentioned even in the abstract as a potential future goal for such an approach, will be the lack of data. This approach requires single cell level descriptions of the system at hand, plus additional datasets to build the model structure. To a certain extent, public data of related tissues/contexts can be used, but it will be necessary to test the dependence of performance on coherence of the input data to develop sufficient trust to use it for new predictions, especially in a medical field.

• *

We thank the reviewer for these reflections, which raise several distinct points that we would like to add in the discussion.

Cell line perturbation is indeed a close and active field of research, with notably numerous models based on optimal transport and VAE and relevant benchmarks(Radig et al. 2025)*. In our view, ReCoN tries to take a complementary angle, by both focusing on the environment effect and using a network-driven approach providing explainability. *

These perturbation methods are typically benchmarked on single cell line screenings, where cell-cell communication is highly limited or absent by design, while ReCoN is specifically designed to exploit multiple cell types interactions. Furthermore, ReCoN relies on a network that aims to provide only explainable hypotheses and molecular cascades. They also typically learn from different data, as ReCoN only uses single-cell data and best perturbation prediction methods learn from a subset of perturbation experiments.

Exploring the performance of ReCoN in perturbation predictions would require designing extensive comparisons with the state-of-the-art taking into account all these nuances which we believe goes outside of the scope of the present study. It however still raises a fundamental question for the development of the next methods and the need to assess whether the perturbation effect is actually larger than the heterogeneity, and we propose to extend the discussion to cover these aspects.

Secondly, this comment raised a point about cell type definition, which can be a hard task and sometimes a wrong description of cells heterogeneity. We note that even if ReCoN relies on grouping cells in some way, it does not impose any particular cell type ontology: users can define their own cell types or cell states, since the CCC layer is typically inferred from single-cell RNA-seq alone and does not require canonical cell-type annotations. This flexibility allows ReCoN to accommodate finer or coarser groupings depending on the biological question. We do not propose a framework to take into account diversity in other ways than homogeneous clusters of cells, but we think that it constitutes an interesting future development of ReCoN or new multicellular modelling methods.

Lastly, we fully agree that an important limitation for ReCoN's use is data availability and generation, which was also a limitation when identifying datasets for the manuscript's applications. We hope that the development of open source atlases will make it easier to leverage tissue-specific prior knowledge and increase potential application, prediction performances, and trust in ReCoN results.

In conclusion, we propose to state in the discussion two new points:

*1) extending multicellular perturbations (including gene knock-out) to conditions where cell types cannot be defined prior to the analysis, or are more to consider across a spectrum, will be an interesting future direction. *

2) there is new a need for broad benchmarks covering both multicellular and single-cell line tasks to evaluate the trade-off between accounting for cell heterogeneity and overall prediction accuracy.

Radig, J., Droit, R., Doncevic, D. et al. scArchon: a scalable benchmarking framework for assessing single-cell perturbation models. Genome Biol 27, 162 (2026). https://doi.org/10.1186/s13059-026-04104-z

R1.48. The authors could comment on how their method compares to others that do not require single cell level information. Despite clear differences, it might be important to show the advantage of using this more complex approach that requires data that is less available. Given the ease with which bulk profiles can be constructed from single cell data, it might be possible to compare the approaches directly. For example, see

1. Wang, S. Patkar, J.S. Lee, E.M. Gertz, W. Robinson, F. Schischlik, D.R. Crawford, A.A. Schäffer, E. Ruppin Deconvolving Clinically Relevant Cellular Immune Cross-talk from Bulk Gene Expression Using CODEFACS and LIRICS Stratifies Patients with Melanoma to Anti-PD-1 Therapy

Mike van Santvoort, Óscar Lapuente-Santana, Maria Zopoglou, Constantin Zackl, Francesca Finotello, Pim van der Hoorn, Federica Eduati,

Mathematically mapping the network of cells in the tumor microenvironment,

Cell Reports Methods 2025

We propose to extend the discussion with additional methods, notably from before single-cell technology developments. We did not plan to include this two specific methods, as to our knowledge, they don't provide output directly comparable to ReCoN's purpose.

• The first work proposes to deconvolute the bulk RNA-seq profile into cell-type-specific expression profiles. It is an interesting reference, as it could allow applying ReCoN even to bulk RNA-seq, but they do not provide comparable results, as their final task corresponds to inferring the ligand-receptor interactions, without providing downstream molecular mechanisms.
• The second method proposed in this paper, RaCInG builds cell-to-cell networks for individual patients. They do not explore the molecular interactions inside the cells themselves, which could be used to build personalised ReCoN's model but seem to be more a prior to recent CCC than ReCoN itself.

• *

• *

Reviewer #2

R2.1. It is not clear how well it performs in independent validations. Authors showed that it can predict the effect of cytokine perturbations in the immune dictionary by selecting an optimal alpha. Authors should validate that using the same alpha value of 0.8, it is possible to accurately predict the effect of cytokine perturbations in independent datasets. This is particularly concerning for cytokine-cell type pairs where the optimal alpha is not known. Therefore, the potential utility of Recon to estimate the effect of multicellular perturbations is not well established.

• *

Response:

• *

*The reviewers raised a very relevant point by pointing out that the alpha coefficient might vary between datasets. *

• *

The value of 0.8 was chosen because it produced the best results in two independent datasets, the immune dictionary and the heart failure showcases. We could here observe some cross-dictionary reproducibility. To complete these findings, we will also verify that 0.8 provides the best performance in a new showcase: the Human Cytokine Dictionary (Oesinghaus et al. 2025)

• *

We tried to contrast this choice by opening on the need to confirm the importance of the indirect effect. We propose to add a sentence explicitly commenting on the impact of these new findings on the alpha coefficient and its robustness value.

• *

It is also accurate to say that ReCoN cannot currently estimate the alpha parameter autonomously. We proposed this default value as it worked on both datasets, but it is possible that no default value could fit them all. The value of alpha is currently a default value, but users are completely free in the current implementation of ReCoN to modify its value depending on their needs

If it was not the case, one option could be to fit its value using similar prior perturbations, when such data is available. For example, perturbing one or a few cytokines, a user could choose the value that explained the best the gene expression responses.

• *

R2.2. Authors claimed that optimal alpha value of 0.8 implies the dominance of indirect effect. But in contrast to this claim, the performance across cytokine-celltype pair only increased from 0.72 to 0.76, which seem to imply that indirect effects do not add much.

*Response: *

• *

The range of performance improvement is an interesting point to discuss for us, as it roughly doubles the computational time and consequently a trade-off between resource usage and this improvement.

• *

While the average improvement from combining the direct and indirect effects observed on the first showcase was around 5%, it reached more than 10% in some cell types. We consider that it still corresponds to an interesting improvement for the current task. Indeed, it here "only" incorporates the coordination of immune cells to a cytokine stimulation, which should not necessarily change their profile drastically compared to isolated exposition.

R2.3. How does the cell-type specific effects prediction perform by just considering the intracellular layers? The authors constructed multiple variants of ReCoN to estimate unicellular and multicellular effects. How is the variant ReCoN-grn different from full ReCoN where gamma is set to zero.

*Response: *

• *

We are thankful for this comment, which will help to restructure the section 2.2.

• *

As the ReCoN-GRN differs from the full ReCoN model, even with a gamma value of 0, as the latest include ligand-to-receptor weights. However, the ReCoN-GRN would correspond to the ReCoN-generic with an alpha of 0, which does not weight ligand-to-receptor links.

• *

We propose to clarify this detail in the section 2.2.2 by adding after the introduction of the ReCoN-generic model the sentence: "Note that ReCoN-grn corresponds to the ReCoN-generic model with alpha set to zero, where no indirect effects are considered. It differs from the full ReCoN model with alpha set to zero, which still includes ligand-to-receptor weights through the receptor-gene bipartite network."

R2.4. In section 2.2, authors assert that if matching datasets are not available, GRN layer can be extracted from other datasets. How well does the GRN layer from one system generalizes to the other system in terms of perturbation prediction?

*Response: *

• *

It is, of course, a complex question, as it probably strongly depends on the studied system. However, we believe while it is important to consider similar systems, using the same samples for the cell-communication and the GRN layer is not necessary.

• *

The first showcase that we propose explores exactly this case. We built the GRN from two unpaired datasets, and the cell communication from a third one. It provided convincing performances, justifying our earlier claim. It is additionally something done in most methods contextualising prior knowledge, which usually comes from other samples and sometimes even other organs (Browaeys, Saelens, and Saeys 2020, Jin et al. 2021, Badia-i-Mompel et al. 2023).

• *

To provide additional insights, we will run the new Human Cytokine Dictionary showcase using both 1) multiomics methods on external PBMC datasets, and 2) a single-cell RNA-seq only method on the Human Dictionary directly. We will then be able to show performances using both data and corresponding methods.

• *

To justify more clearly our claim according to reviewer's comment, we propose highlighting in the showcase itself this justification: ".... this showcase highlights the possibility to combine networks obtained from distinct datasets...".

Related to combining datasets, we propose to clarify the reasons behind our choices for the Immune Dictionary showcase with the additional supplementary text proposed in response to the comment R1.4.

• *

Badia-i-Mompel P, Wessels L, Müller-Dott S et al. Gene regulatory network inference in the era of single-cell multi-omics. Nat Rev Genet 2023;24(11):739-54. https://doi.org/10.1038/s41576-023-00618-5.

Browaeys R, Saelens W, Saeys Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods 2020;17(2):159-62. https://doi.org/10.1038/s41592-019-0667-5.

Jin S, Guerrero-Juarez CF, Zhang L et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun 2021;12(1):1088. https://doi.org/10.1038/s41467-021-21246-9.

R2.5. In the abstract, authors claimed that ReCoN can predict the effect of gene knockouts. But authors did not show any application or validation to support this claim.

Response:

• *

We indeed had no showcase that could explicitly measure the performance of ReCoN directly for gene knockout, while the possible application was introduced in the abstract.

* We believe that ReCoN could be used in the future to infer such perturbations, but we fully agree that this claim cannot be presented without justification.

We propose to remove the introduction of gene-knockout there, and to introduce it in the discussion opening instead, specifying that it will require specific experience and constitutes a possible future extension of the work.*

R2.6. The communication between cells might be dependent on their spatial proximity. Is it possible to construct the CCC layer by incorporating the context-matched spatial data? How would that affect the performance of multicellular response prediction?

Response:

• *

*This is a very interesting comment as numerous methods using spatial transcriptomic data have been published recently. *

• *

In the current formulation, the beta coefficient Bi_j modulates the impact of the cell type i on the cell type j. If the spatial transcriptomic data can inform on the proximity between cell types, and its overall impact on their communication, users could enforce more communication between some.

• *

However, as ReCoN is a cell-type centric model, adding spatial information can only be done at a general scale, or by modelling independently spatial regions such as presented in the Microenvironments heart infarction showcase. It means that ReCoN cannot beneficiate from the potential of spatial transcriptomic as much as models representing the tissue structure.

R2.7. In the fibroblast application in Fig 4d, based on the cardiac cell types expression in region type, they are predicting fibroblast gene expression. Wouldn't the most direct benchmarking be comparison with observed fibroblast expression from the ST (after deconvolution perhaps)?

Response:

• *

This was a helpful comment to guide the restructuration of the microenvironment heart infarction showcase, as we believe the whole showcase objective was not formulated clearly enough.

• *

We aim at modelling the impact of the environment on the transcriptome. As the complete transcriptome of a cell results from numerous interacting variables, we believe that comparing the correlation between ReCoN's scores and the transcriptome would not evaluate the prediction of the environment impact.

• *

For this reason, we wanted to compare the results to the specific differences from the microenvironment. We focused on gene set enrichment that seemed less noisy for such a comparative experiment, in particular from Visium10X data that has a particularly high dropout rate.

• *

We propose to strengthen the validation by providing molecular insights into the three groups of cells studied.

The spatial data themselves are bulk, adding a layer of noise over the small number of genes captured by Visium. Instead of a correlation with the deconvoluted spots, we have equivalent single-cell RNA-seq fibroblast data annotated in the same study, which matches the three modelled niches. We propose to conduct a differential expression here and try to compute a correlation between these groups and ReCoN scores, providing a quantitative analysis.

If the correlation was low because of the noise in the data (notably leading to the permutation of individual gene orders even if overall biological signals and gene set orders are conserved), we will additionally do a pathway enrichment over this data, enriching also the qualitative validation.

R2.8. Section 2.6 Besides the cytokine section, it is difficult to assess the added value of this approach. Likely there is a lot of valuable findings here but difficult to say because the assessment is very qualitative.

Response:

• *

One of the challenges around this work was to find relevant dataset to evaluate ReCoN. We tried to complete the direct quantitative evaluation from the Immune Dictionary with another quantitive evaluation from the heart atlas multicellular programs, despite a much less direct validation.

• *

We hope that the production of new perturbation experiments over multicellular datasets, especially cell-type targeted perturbations, will provide more opportunities to validate the different findings and claim from our current manuscript.

• *

On a similar note, no method seemed proposing similar predictions to be compared to. It led to the use of Nichenet score and the current decomposition of the ReCoN model in the section 2.2.1 to evaluate the contribution of the model.

R2.9. The article is dense and writing should be reorganized for better readability.

Minor issues -

No p-values in figures.

*Response: *

• *

We agree that integrating values directly in the panels would make the reading of the figure easier. We would like to introduce the p-values in the panels 2d, 2e, 2f, 2g. We had forgot to indicate in the legend of the panel 4.d that all bold scores were associated with a p-value *

R2.10. Typo - ReCoN-genetic should be - ReCoN-generic.

• *

Response:

• *

We are thankful for noticing the typo and corrected it in the new version.

• *

R2.11. Authors may consider adding figures to describe their results on balance between direct and indirect effects in section 2.2.2.

• *

Response:

• *

Depending on the new findings on the indirect effect iterations, we propose adding an additional panel on their combination or a supplementary figure.

• *

R2.12. Redundancy in the following two lines -

o While these approaches effectively describe what tissue-wide programs are coordinated, they generally offer limited insight into the molecular mechanisms that establish or regulate these programs.

o Despite their ability to identify coordinated tissue-wide programs, multicellular program analyses typically offer limited insight into the underlying molecular mechanisms that orchestrate these programs.

• *

Response:

• *

We propose in the version of the manuscript to remove the first sentence. In our opinion, starting the next paragraph by this clarification seems more helpful to guide the reader than having it at the end of the previous one.

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

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

• *

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

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

R2.13. The direct and indirect effects are treated in two separate steps. In reality of course these effects are operating simultaneously. I wonder if this could be better modelled by iterating through the two steps. It might be worthwhile

trying to see if that improves the performance.

We thank the reviewer for this interesting idea, and propose to add a supplementary text to present the result of this discussion to the readers.

• *

The direct effect is supposed to be measurable from the first iteration only, as we try to represent the effect of direct receptor binding. Regarding the indirect effect, iterations could be done to model the indirect effect, which could represent more distant effect in time.

• *

On an algorithmic note, the indirect effect already allow several "iterations" of this effect, as each random walk can loop between all cell types until restart. However, it does not allow to control the weight of the different successive transition. In practice, with a high restart probability, an extreme weight is given to the first "iteration" over the second, as there is three layers to cross to explore the next cell.

• *

First, we propose clarifying this section of the manuscript, to explain the depth of the indirect effect explorations.

• *

Biologically, it is highly possible that these iterations have an important role to explain the complete reaction of the cells. However, we believe that it hits a major limitation of our modelling, and RWR based exploration in general, as it goes against the enforcement of restarts.

• *

We aim to represent pairwise measurements, representing the impact of one node on another. But random walks without restart are not naturally well fitted to this problem, as they naturally converge to a stationary distribution ((László, Lov, and Erdos 1996)). In the case of ReCoN, it means that each gene and receptor, if we pushed the exploration indefinitely, would have the same probability to end up on each node of the system.

• *

The restart mitigates this impact and enforces the impacts of the seeds by ensuring that the walkers stay close to the seed. (Tong, Faloutsos, and Pan 2006). By iterating successively from the new distribution obtained from the RWR, we would go against this important probability and progressively converge toward the stationary distribution from classical random walks.

• *

So we completely share the opinion of the reviewer that the iterative nature of the indirect effect should be explored too, but we don't believe that ReCoN can model them accurately. We hope that new exploration methods will be able to decipher the importance of these iterations, once additional arguments have been gathered to justify the global interest of considering the indirect effect.

• *

Bibliography:

• *

László L, Lov L, Erdos O. Random Walks on Graphs: A Survey. 1 Jan. 1996:1-46.

• *

Tong H, Faloutsos C, Pan J yu. Fast Random Walk with Restart and Its Applications. Sixth Int Conf Data Min ICDM06 Dec. 2006:613-22. https://doi.org/10.1109/ICDM.2006.70.

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 - This is an interesting paper where the authors predict the multicellular response to the molecular perturbations. The idea is somewhat novel and offers a conceptual enhancement by modelling the multicellular response as collective outcome of cell intrinsic gene regulatory changes coupled with cell-cell communication by using a simple network diffusion-based approach. We have a few comments to help strengthen the work.

• It is not clear how well it performs in independent validations. Authors showed that it can predict the effect of cytokine perturbations in the immune dictionary by selecting an optimal alpha. Authors should validate that using the same alpha value of 0.8, it is possible to accurately predict the effect of cytokine perturbations in independent datasets. This is particularly concerning for cytokine-cell type pairs where the optimal alpha is not known. Therefore, the potential utility of Recon to estimate the effect of multicellular perturbations is not well established.
• Authors claimed that optimal alpha value of 0.8 implies the dominance of indirect effect. But in contrast to this claim, the performance across cytokine-celltype pair only increased from 0.72 to 0.76, which seem to imply that indirect effects do not add much.
• How does the cell-type specific effects prediction perform by just considering the intracellular layers? The authors constructed multiple variants of ReCoN to estimate unicellular and multicellular effects. How is the variant ReCoN-grn different from full ReCoN where gamma is set to zero.
• In section 2.2, authors assert that if matching datasets are not available, GRN layer can be extracted from other datasets. How well does the GRN layer from one system generalizes to the other system in terms of perturbation prediction?
• In the abstract, authors claimed that ReCoN can predict the effect of gene knockouts. But authors did not show any application or validation to support this claim.
• The communication between cells might be dependent on their spatial proximity. Is it possible to construct the CCC layer by incorporating the context matched spatial data? How would that affect the performance of multicellular response prediction?
• The direct and indirect effects are treated in two separate steps. In reality of course these effects are operating simultaneously. I wonder if this could be better modelled by iterating through the two steps. It might be worthwhile trying to see if that improves the performance.
• In the fibroblast application in Fig 4d, based on the cardiac cell types expression in region type, they are predicting fibroblast gene expression. Wouldn't the most direct benchmarking be comparison with observed fibroblast expression from the ST (after deconvolution perhaps)?
• Section 2.6 Besides the cytokine section, it is difficult to assess the added value of this approach. Likely there is a lot of valuable findings here but difficult to say because the assessment is very qualitative.
• The article is dense and writing should be reorganized for better readability.

Minor issues

• No p-values in figures.
• Typo - ReCoN-genetic should be - ReCoN-generic.
• Authors may consider adding figures to describe their results on balance between direct and indirect effects in section 2.2.2.
• Redundancy in the following two lines -
  • While these approaches effectively describe what tissue-wide programs are coordinated, they generally offer limited insight into the molecular mechanisms that establish or regulate these programs.
  • Despite their ability to identify coordinated tissue-wide programs, multicellular program analyses typically offer limited insight into the underlying molecular mechanisms that orchestrate these programs.

Significance

This is an interesting paper where the authors predict the multicellular response to the molecular perturbations. The idea is somewhat novel and offers a conceptual enhancement by modelling the multicellular response as collective outcome of cell intrinsic gene regulatory changes coupled with cell-cell communication by using a simple network diffusion-based approach.

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 propose an approach to model complex regulatory processes in tissue or cell collections in specific environments taking into account intra- cellular regulatory processes at multiple levels and inter-cellular communication, importantly offering a chance to estimate the importance of indirect effects of perturbations on one cell type via processes in other cell types. Increasingly more complete models allow testing the impact of each component and of integrating data as context-specific information versus general prior knowledge. 3 main use cases are provided exploiting public datases: prediction of the effect of specific in-vivo cytokine perturbations on mouse lymph node tissues Healthy and disease myocardium in a heart failure multiome dataset Myocardial infarction spatial transcriptomics to identify how different cellular neighbourhoods are related to fibroblast phenotype and fibrosis The main framework is an extension of their previous HuMMus framework to investigate multilayer networks of regulation within a single cell type to also consider inter-cellular interactions, thus including i) tf-target GRN, ii) receptor a receptor layer based on PPI, and cell-cell communication based on LR interactions. These complex networks are then explored within the framework of Random Walk with Restart, which allows to establish 'interaction weights' between different nodes in the network, based on repeated simulations of spreading on the network that thus produce scores of proximity between network nodes, across possible paths. In this study first RWR that only allow intra-cell type walks are performed to calculate direct interaction of perturbation on node states, then RWRs across layers are also enabled, to calculate the importance of inter-cell interactions (via coeff gamma). The importance of each cell type is given by another coeff B that can either correspond to cell type proportions or spatial proximity of cell pairs and finally the scores of within and inter-cell interactions are weighted with a coefficient alpha.

The central contribution that allows coupling of intra with inter-cellular interactions is the establishment of receptor-gene links. Instead of inferring it from data, they propose to express the receptor-gene matrix as: R = L ⋅ G taking ligand-receptor (L) and ligand-gene (G) adjacency matrices from NicheNet and using NNLS to compute R.

Generally, for all these cases, comparison between performance in inferring the effect of perturbation or the upstream regulators or downstream targets are provided with assessment of AUROC/AUPRC values.

• Are the claims and the conclusions supported by the data or do they require additional experiments or analyses to support them?

This is a very well-written paper, the methods used are adequate and the use cases are relevant and broad, exploiting state of the art datasets and tools.

The author's claims are mostly justified. The authors could make an effort to more explicitly cite other efforts in similar directions. The claim 'We envision ReCoN as a extension to prior multicellular modelling, offering an interesting compromise between prediction of cell type responses and understanding of their molecular coordination.' is very general and could be better substantiated. In fact, the authors do not really give examples of alternative approaches to study systems of interacting cells, other than mechanistic agent based models, that clearly are very different. Moreover, the exploration of the multilayer networks with RWR is a very reasonable choice but could there be other approaches? I think the authors could discuss this issue to briefly support their choice of this method.

Generally the discussion should provide the reader the context in the existing literature in which the work can be set, detailing its impact. I think this could be improved.

Regarding the choice of datasets, it is clear that the method is quite demanding, requiring single cell and different omics to build the model, in addition to the expression dataset that is used as a use case. This inevitably leads to using a mix of datasets. For example in the mouse experiments the gene regulatory network was inferred from both a lymph node scRNA-seq dataset and a splenic scATAC-seq dataset, presumably due to the lack of multiome data in this setting. However the cell-cell communication network was inferred from the control case of the Immune Dictionary. Why can't the authors use the control data also for inferring GRNs? Is atac-seq really necessary in the inference of the GRN? What is the impact of the fact that lymph node and spleen samples might be different?

'

• Please request additional experiments only if they are essential for the conclusions. Alternatively, ask the authors to qualify their claims as preliminary or speculative, or to remove them altogether.

• If you have constructive further reaching suggestions that could significantly improve the study but would open new lines of investigations, please label them as "OPTIONAL".

• Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated time investment for substantial experiments.

• Are the data and the methods presented in such a way that they can be reproduced? The code is very clear, we were able to install and run it and it is quite well-documented. However, a few more details should be given in the text regarding how the evaluation of the performance is carried out. For example: If I understand correctly, when predicting the impact of cytokine perturbations the ReCoN predictions of genes impacted are compared to differentially expressed genes identified through traditional DEG analysis. What is compared is the ranking of these genes from ReCoN with the ranking provided by DEseq2. There is no description of how this comparison of ranking gives rise to AUROC values. Also, is it just the ranking that is predicted or can they also estimate how well they can predict the effect size?

When describing the use cases, I think a bit more detail would help. For example 'To identify the cell-type-specific genes associated with HF, we used the MOFAcell scores of the multicellular factor 1 (MCP1) reported in ReHeat236' I supposed the explanation is on the dataset but for the sake of clarity it would be good to expand this sentence to give at least an idea of the approach.

Regarding the calculation of the R matrix from the NichNet matrices L and G, I gather that the R matrix is calculated once and is thus fully data-independent and available just like the L and G matrices from NichNet. This was not very clear in the tutorials.

Also, this might just be a typo in the tutorial: 'The default α = 0.8 gives more weight to direct effects, which has been empirically validated. You can adjust this based on your biological question." I believe the manuscript says alpha>0.5 refers to indirect effects dominating.

Same for the pre-processing of the spatial data for the third use case, a little more details on how this was done would help the users and readers.

• Are the experiments adequately replicated and statistical analysis adequate? I don't see issues with the statistical power of the analysis. Rather, I think the authors should provide some examination of the parameter space for their model. Whereas ana analysis of the impact of the Alpha parameter is provided, I believe there are several more parameters that have a crucial impact and choices for their values should be discussed.

For example 'In the GRN reconstruction only the links with a score above 1.5e-7 were retained in ReCoN's gene regulatory layer. How was this chosen?

We have identified the following parameters that are somehow justified but could be explored to have a better feel for how they impact the results

Restart probability: How often the walker goes back to the starting seed/molecule Layer transition probability: How often the walker stays in the same layer - different cell? - different layers? Gamma Node transition within a layer: How often one jumps to a different layer Weighting parameters: How much weight for direct or indirect effect to account for the combined effect - alpha - this is the only one that is explicitly explored.

Finally, this might be considered OPTIONAL but would greatly improve the work in our opinion: The method crucially depends on the networks that are used in the different layers and to connect layers and cell types. As we know, biological data is noisy and incomplete (FP and FN) at each level and in each datatype. It would be really useful to estimate what is the robustness of the results to this noise. Particularly, from personal experience, we think the GRNs reconstructed from data are often almost fully connected and it is exceedingly difficult to validate them in specific contexts. This means that some 'errors' are likely to be present. Since several methods exist for inferring GRNs one could simply compare the results using different methods for this part of the network. A related point involves the characteristics of the RWR algorithm, that will be quite impacted by the presence of hubs in these networks (either in single layers or across several) that is likely to impact the exploration. If proteins that are hub are effectively important, that is not a problem, but in some layers, for example, the receptor-receptor layer that presumably will contain PPIs, there might be biases in hubs being just better studied proteins, and these hubs might have an 'unjustified' weight in the walks. One potential approach to assess the robustness of the method to these issues could be an empirical one that just randomly perturbs the networks in ReCoN to see to what extent similar predictions are achieved.

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?

Please add page numbers. Figures are nice and clear. Some specific minor points are listed here below.

Define hMLN on first appearance fig1 caption (no page numbers..;) 2nd appearance heterogeneous multilayer structure (HMLN) ... Bi_j not so clear to what it refers when first mentioned personalized interaction specificity. - maybe better word than personalised (contextualised?) ReCoN-genetic and ReCoN, ( generic?) responses. It is expected to observe common behaviors in-between cell-type, that the GRN and the generic CCC network already contribute captures. - not very clear

Figure 2b the icon of cells with double arrows might suggest phenotype shift when instead this is just communication eTACs explain acronym and what they are Due to very few genes being differentially expressed, only cDC1 was conserved and evaluated for IL22, Not so clear In this showcase (not very clear, use case?) different fibroblast specializations - maybe phenotypes?

Figure 4b b) Schematic view of the deconvolution process and cell type-specific count inference from the spatial niches. Not so clear what the heatmap shows, rows and columns Spots heatmap : label niche on rectangles in cols And each col is a spot Rows are cell types or cells? In the cell types x spot

Cell2location. Add reference, maybe explain basic functionality?

reconstructing different patients, tissues, and microenvironments to predict context-specific molecular treatments. Unclear fibrosis in different - at molecular levels

Figure 5d myeloid and endothelial colour code inversed from 5 BC 5d indicate important pathways in organe should not change the colour of the nodes (purple=common, blue or green specific). Use border colour maybe? 5e is not a venn diagram e) Venn diagram showing the overlap between transcription factors (TFs) predicted by ReCoN (green) and those previously implicated in fibrosis (orange) or cardiac diseases (violet). Only the top 10 TFs were annotated from literature sources; full sizes of fibrosis- and cardiac disease-related receptor sets can therefore not be represented. f) also not a venn diagram e/f now in supp the "NABA ECM collagens" gene set. Nodes are grouped by molecular type (e.g., transcription factors, receptors, ligands), and links represent the weighted, direct regulatory interactions present in the ReCoN-constructed

Why Sankey plot? Normally sankey plot represents flow (of regions changing from 1 state to another) but here this is just a weighted network? No communication from firbos back to other cell types? No communication between ventricular/myeloid/lymphoid?

as a extension to - an underrepresented in the current. - current framework? However, it can't represent more - cannot Borrowing representation from hypergraphs, which introduces The network exploration implementation of ReCoN also present some limitations. limitations. While random walks with restarts offer a stable and fast exploration workflow for multilayer networks, it currently only considers positive weights to predict regulation strengths. It involves that the nature of the regulation, as activation or inhibition, has to be identified a posteriori.

• check concordance/grammar

Only the nodes that are included in one of the layers are present in the final results, ignoring the ones present only in bipartites. Unclear a scATAC - an Barsi et al is published https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013188 effects, allowing for modulating in a second time their contribution. - word order

others. However, it is possible to adjust the Beta coefficient to represent it based on the available information for each dataset. Represent- adjust?

We use the latter to compare the different models. - what is the latter?

It resulted in the scRNA-seq in 1,789 cells with 13,167 genes, and for the scATAC-seq in 3,759 cells with 254,545 regions. Check english GRETA pipeline.- reference

We kept all the cells whose annotations through unsupervised clustering, followed by marker gene annotations, through scANVI were coherent. Word order In parallel, pairs of ligands and receptors with both associated with scores above an absolute gene loading of 0.1 were considered potential driver interactions in HF. Unclear gseapy Python - reference?

and to calculate average for each spatial context the average cell type expression. Unclear

We only used the loadings of all cell types but the fibroblasts to consider the effect of the sole environment. Unclear We realised a downstream - performed

The profiles inferred by ReCoN were first very correlated in all three contexts. - unclear

Significance

Provide contextual information to readers (editors and researchers) about the novelty of the study, its value for the field and the communities that might be interested.

This is a very timely paper, dealing with an important gap in the literature. It is not an entirely new framework, but it integrates different existing approaches to solve a complex issue in a creative way. To my knowledge, it is the first attempt to consider and formalise regulation processes involving both intra- and inter-cellular interactions. The results support the importance of distinguishing the different paths that can relate the impact of a perturbation to specific genes/functions in different cells and their overall ecosystem.

General assessment: provide a summary of the strengths and limitations of the study. What are the strongest and most important aspects? What aspects of the study should be improved or could be developed?

The tool offers a combination of approaches, providing a coherent framework. The code is well documented and functional. The use cases are quite compelling. Sadly, the only type of validation possible involves confirmation of known facts from the literature, which makes it hard to evaluate the full impact of some of the predictions. I think the details of how the method works and especially how the performance was evaluated could be expanded and an assessment of how different parameters and choices impact the results would also be very helpful. An effort to compare the presented variations of the method to some other approach would be very welcome, but I am finding it hard to identify what an alternative approach could be comparable.

Advance: compare the study to the closest related results in the literature or highlight results reported for the first time to your knowledge; does the study extend the knowledge in the field and in which way? Describe the nature of the advance and the resulting insights (for example: conceptual, technical, clinical, mechanistic, functional,...).

Potentially the closest results are models that can predict the effect of perturbations on cell line cultures. Several approaches in the literature employ either transformers or optimal transport to predict the effect of perturbations in single cell datasets. One of the main issues is an underlying necessary assumption that the perturbation effect will be larger than the heterogeneity (in cell lines for example), which becomes increasingly difficult when considering in-vivo experiments. ReCoN obviously goes beyond this by considering explicitly the presence of different cell types but distinctions of cell types are sometimes quite arbitrary and potentially application of ReCoN to some of the in-vitro culture datasets, even on cell lines, could be a way to test its performance and benchmark it against other methods. The main bottleneck in the application of this framework to 'personalisation' of therapies, mentioned even in the abstract as a potential future goal for such an approach, will be the lack of data. This approach requires single cell level descriptions of the system at hand, plus additional datasets to build the model structure. To a certain extent, public data of related tissues/contexts can be used, but it will be necessary to test the dependence of performance on coherence of the input data to develop sufficient trust to use it for new predictions, especially in a medical field.

The authors could comment on how their method compares to others that do not require single cell level information. Despite clear differences, it might be important to show the advantage of using this more complex approach that requires data that is less available. Given the ease with which bulk profiles can be constructed from single cell data, it might be possible to compare the approaches directly. For example, see K. Wang, S. Patkar, J.S. Lee, E.M. Gertz, W. Robinson, F. Schischlik, D.R. Crawford, A.A. Schäffer, E. Ruppin Deconvolving Clinically Relevant Cellular Immune Cross-talk from Bulk Gene Expression Using CODEFACS and LIRICS Stratifies Patients with Melanoma to Anti-PD-1 Therapy

Mike van Santvoort, Óscar Lapuente-Santana, Maria Zopoglou, Constantin Zackl, Francesca Finotello, Pim van der Hoorn, Federica Eduati, Mathematically mapping the network of cells in the tumor microenvironment, Cell Reports Methods 2025

Audience: describe the type of audience ("specialized", "broad", "basic research", "translational/clinical", etc...) that will be interested or influenced by this research; how will this research be used by others; will it be of interest beyond the specific field?

Broad interest to biomedical researchers and also biologists in other fields. While the method allows advances in basic research on biological process regulation, a clear clinical application can be envisaged in immuno-oncology for example/ immunology and even general molecular medicine.

Please define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

I am a computational biologist with expertise in network models, regulatory networks, agent-based models and especially familiar with the tumour microenvironment and processes therein. I can more or less appreciate the meaningfulness of the biological findings related to the mouse lymphnode example. I am much less of an expert on heart tissue modeling, heart failure, fibrosis etc, required to fully comprehend the impact of the second and third use cases.

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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)): __ In this manuscript, the authors describe the discovery of a molecular regulator of the immune transcriptional program, which is activated by intestinal distension upon bacterial colonization of the C. elegans intestine. Taking advantage of the fact that inhibition of aex-5 is known to cause intestinal distension and a C-type lectin gene clec-60 as a marker for the immune response to intestinal distension (clec-60p::gfp), the authors performed a forward genetic screen for suppressors of the immune response activation. Of the two mutants isolated, they focused on the stronger suppressor, which corresponded to a cysteine-type DUB, the Ubiquitin Specific Peptidase-14 (usp-14). Through rescue experiments, phenocopy analyses, and quantitative RT-PCR, they validated usp-14 as the causal gene and initiated characterization of its role in immune response activation. To this end, the authors investigated the tissue of action, identifying the intestine as the tissue in which usp-14 mediates the regulation of the immune response. Through transcriptomic analyses, they found that the signalling pathway likely regulated by usp-14 in response to intestinal distension is the Wnt pathway, as they have observed reduction in the transcriptional level of some of the Wnt pathway components in usp-4(tm1481), in response to infection with S. aureus. Additionally, transcriptomic data indicate that usp-14 plays a role in immunity regulation even in the absence of infection. Based on these findings, the authors propose that usp-14 has a dual role in immune regulation: one in surveillance immunity, preventing overactivation of immune responses, and another as a mediator of pathogen-induced responses, such as those triggered by P. aeruginosa or S. aureus. The experiments are rigorous and the results robust; however, some points would benefit from further investigation or clarification. __Response: We thank the reviewer for an excellent summary of our work and for the valuable feedback.

Comment: The expression domain of usp-14 appears to be quite expanded based on single cell RNAseq data (e.g. PMID: 28818938) therefore it is likely that the transgenes used for expression analysis are lacking key regulatory information. Alternative methods like smFISH would be more appropriate to characterise the spatiotemporal pattern of usp-14 expression in more detail. Response: We thank the reviewer for this valuable suggestion. In the original version of the manuscript, we used a 714 bp region upstream of the usp-14 start codon to generate the transcriptional reporter. In the revised manuscript, we reconstructed the reporter using a longer 1924 bp upstream promoter region together with a portion of exon 1. Using this updated reporter, we observed substantially broader expression of usp-14, particularly during the early larval stages. These results are described on page 6, lines 147-152: “We next examined the spatiotemporal expression pattern of usp-14 in C. elegans. To this end, we generated transgenic worms expressing GFP under the control of the usp-14 promoter (usp-14p::gfp). During early larval development, usp-14 was broadly expressed across multiple tissues (Figure 3A). However, in L4 larvae and adult animals, expression became more restricted and was predominantly observed in the intestine and a subset of neuronal cells. Notably, both intestinal and neuronal expression persisted throughout development (Figure 3A).”

Comment: __The mutation mapped in usp-14(jsn19) is a missense mutation (E122K) that suppresses the immune response to a degree comparable to the usp-14(tm1481) deletion allele. However, the authors do not show the functional domains in Fig. 1E potentially affected by this missense mutation. __Response: We have now updated Figure 1E to include the functional domains of USP-14 and mapped both the usp-14(jsn19) missense allele and the usp-14(tm1481) deletion allele onto the protein schematic.

Comment: __How USP-14 regulates Wnt and how Wnt signalling relates to activation of immune responses is not fully supported. Are the Wnt components mentioned in the study induced specifically in the intestine upon infection and does USP-14 act in the intestine in the context of this regulation? How do the authors interpret that both Wnt ligands and receptors are induced ? Does Wnt signalling appear as a GO term in the transcriptomic analysis? The authors can include Wnt signalling components in the analysis of the transcriptomic results. __Response: We thank the reviewer for these insightful comments. Previous studies have shown that the Wnt pathway components examined in our study are induced in the intestine upon infection and function within the intestine to regulate host defense against bacterial pathogens (PMID: 29768179; PMID: 36323254).

We did not observe significant enrichment of Wnt signaling terms in the GO analysis of our transcriptomic dataset. We believe this is likely due to the stringent thresholds used for differential expression analysis (fold change > 2 and p At present, the precise mechanism by which USP-14 regulates Wnt pathway components remains unclear. One possibility is that USP-14 influences Wnt signaling indirectly through additional substrates or interacting proteins that regulate transcriptional outputs. We have now clarified this point in the Discussion (page 12, lines 340–345): “These observations raise the possibility that additional USP-14 substrates or interacting proteins modulate transcriptional outputs downstream of intestinal distension. Future studies aimed at identifying the direct substrates of USP-14 and defining how USP-14 interfaces with neuronal ACC-4 signaling and other distension-responsive pathways will provide important mechanistic insight into how intestinal distension is coupled to innate immune activation.”

Regarding the simultaneous induction of Wnt ligands and receptors, we interpret this as a potential amplification or reinforcement mechanism that enhances Wnt/β-catenin signaling during infection-induced intestinal distension. However, further studies will be required to determine the mechanistic significance of this coordinated transcriptional regulation.

Comment: __Overall, in most of the figures, the micrographs are in general quite dark and exhibit poor contrast between signal and background, particularly in Fig. 1, panels B and J, and Fig. 2, panels B and F (upper rows). Even though these panels are intended to show absence of response, the outlines of the worms are difficult to discern. __Response: We thank the reviewer for the feedback. We have now improved the image presentation throughout the manuscript by either increasing the intensity or adding dotted outlines to more clearly indicate worm positions.

Comment: __In Figure S3, panels A and B, the pmk-1(km25); usp-14(tm1481) animals subjected to aex-5 RNAi show some level of fluorescence/response induction comparable to pmk-1(km25) alone. This observation is not discussed in the text. __Response: We have now discussed this observation in the text. These results are described on page 9, lines 240-244: “Although pmk-1(km25);usp-14(tm1481) worms displayed relatively higher GFP levels than usp-14(tm1481) single mutants upon aex-5 RNAi treatment, this effect likely reflects the elevated basal GFP expression observed in pmk-1(km25) mutants (Figure S4B). Importantly, pmk-1(km25);usp-14(tm1481) animals still exhibited significantly lower GFP levels than pmk-1(km25) single mutants.”

Reviewer #1 (Significance (Required)): __ __Comment: __The work is interesting because it expands some previous work in the field demonstrating immune response induction as a consequence of intestinal distension even in the absence of bacterial infection. This is known to be mediated by the neuronal acetylcholine receptor ACC-4, which signals to the intestine where it regulates immune genes via the Wnt pathway. However, how USP-14 relates to ACC-4 is currently unclear and whether USP-14 function is really required in the intestine to control Wnt signalling is not demonstrated. The authors should include a model to describe how their findings relate to the previous literature and how USP-14 may link mechanistically to Wnt signalling pathway activation. __Response: We thank the reviewer for this insightful comment. We agree that the relationship between USP-14, ACC-4, and Wnt signaling requires further clarification. As suggested by the reviewer, we have now included a model summarizing the current understanding of intestinal distension-induced immune activation and integrating our findings with previous literature (Figure 6H).

Comment: __It remains also unclear whether usp-14 is the only deubiquitinase involved in intestinal distension-induced signalling via the Wnt pathway, or whether other paralog usp genes might also contribute to regulation of immune-responsive transcription. Notably, several mammalian deubiquitinases have established roles in cancer suppression and inflammatory response and innate immunity in other systems so this would increase the potential significance of the work. __Response: We thank the reviewer for this valuable suggestion. To systematically examine whether additional DUBs contribute to intestinal distension-induced immune activation, we performed an RNAi screen targeting all DUBs available in the Ahringer RNAi library using the aex-5(sa23);clec-60p::gfp reporter strain. Among the DUBs tested, knockdown of usp-14 produced the strongest suppression of clec-60p::gfp expression. Although knockdown of usp-5 also partially suppressed GFP induction, usp-5 RNAi did not affect survival during P. aeruginosa infection, suggesting that usp-5 is not required for host defense under these conditions. Together, these findings identify USP-14 as the major DUB required for intestinal distension-induced immune activation in our experimental system. These results are now included in Figure 1G, H, and Figure S2.

Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ Summary C. elegans are soil-dwelling nematodes that feed on bacteria and fungi and thus must be able to distinguish between innocuous and pathogenic species of microbes to survive. Though they lack adaptive immunity, these animals have an ancient version of an innate immune system that has no circulating sentinel or phagocytic cells yet can still mount a response to pathogen exposure. A consequence of the mode of infection of some ingested bacterial pathogens is intestinal distension which by itself, even in the absence of pathogens, is sufficient to trigger the expression of genes encoding immune effectors, including proteins that are bactericidal. The complete mechanistic scheme connecting intestinal distension to the expression of immunity genes has not been resolved, motivating the authors to perform a forward genetic screen for additional components of this pathway. One mutant that the authors isolated was usp-14, encoding an evolutionarily conserved deubiqutinating enzyme. Functional analysis revealed that usp-14 confers protection from microbial pathogens and that the intestine is its primary site of action for its role in host defense. The authors' data indicate that while USP-14 regulates the expression of innate immunity genes that are induced by intestinal distension, surprisingly it functions independently of several canonical innate immune signaling pathways, including the pmk-1/p38 MAPK pathway. Instead, USP-14 appears to act through Wnt signaling to regulate immune effectors by upregulating the expression of several components of that pathway, including the C. elegans ß-catenin ortholog bar-1. This places usp-14 within a gut-brain axis previously shown to control the C. elegans innate immune response through acetylcholine-mediated activation of Wnt signaling. The authors' findings provide new mechanistic insight to this pathway and add to the understanding of ubiqutination as an immune regulatory module. __Response: We thank the reviewer for providing an excellent summary of our work.

Major comments __1. There are three types of experiments in which the authors use the same set of controls across several different figure panels, as stated in the legend to Figure 2. First, when quantifying GFP levels of clec-60::gfp in RNAi-treated animals, the authors use the same clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls for Fig. 1K, 2C, and 2G. For infection assays with S. aureus NCTC8325, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2E are the same as the ones used in Fig. 1M. Similarly, for infection assays with P. aeruginosa PA14, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2I is the same as was used for Fig 1I. In each case, if the authors in fact collected all of the data for each strain that they studied at the same time but then chose to parse larger datasets into separate figure panels to make it more clear to the reader, then this approach is valid but the authors need to explicitly state that this is what they did. However, if the data pertaining to the control strains were collected at a different time or if it comes from a separate biological replicate, then re-using data from the controls is not appropriate because it would not accurately reflect the specific conditions of the experiment to which the data are being compared. If this is indeed the scenario, then the authors will need to repeat these experiments and include the appropriate control in each iteration. __Response: While preparing the manuscript, these experiments were performed simultaneously. Therefore, all panels that share controls have results from experiments performed simultaneously and represent the same biological replicate. We have added this additional information in the relevant figure legends.

Comment: __2. From the legends describing figure panels that include data pertaining to clec-60p::gfp expression levels as assessed by fluorescence microscopy it seems that, in general, the authors measured GFP fluorescence in about 30 animals to produce quantitative data. How many biological replicates of these types of experiments were carried out? This is not explicitly stated in the section describing fluorescence imaging in the Methods section. Following the description of their methodology regarding statistical analysis of survival curves from microbial infection assays, however, the authors state that, "[a]ll experiments were performed independently at least three times unless otherwise noted." Does this statement apply to microscopy or only to experiments involving infection assays? If the data reporting quantitation of GFP signal is based on only 30 animals, then additional biological replicates are necessary, along with appropriate statistical analyses. __Response: The quantified GFP fluorescence data are derived from three independent biological replicates. In each experiment, we typically imaged and quantified approximately 10 worms per condition, yielding a total of ~30 worms analyzed per genotype or treatment across all replicates (except Figure S1B, where we had two independent replicates). We have added the number of experiments in the figure legends for these data.

Comment: __3. The authors have made all of the RNASeq data publicly available on the Sequence Read Archive, and they include data from several pairwise comparisons for differential gene expression analysis in their supplemental files. One of the most important facts to come out of the authors' Gene Ontology analyses of their RNASeq data is that the genes that are upregulated in a usp-14-dependent manner upon intestinal distension are enriched for those whose products play a role in innate immunity/host defense. The authors should say more about these genes. Are there any commonalities between them with regard to function? Are any of them targets of transcription factors that are known to function in C. elegans innate immunity? If so, this could provide clues as to what the substrates of USP-14 might be. Importantly, the specific identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense should be revealed in a supplemental file, and designated as being dependent on or independent of usp-14 for their expression during intestinal distension. __Response: We thank the reviewer for this insightful suggestion. We have now expanded the Results section to describe the functional categories enriched among the USP-14-dependent intestinal distension-induced immune genes, including C-type lectins, ShK toxin domain-containing proteins, and lysozymes (page 7, lines 193-195).

In addition, we compared our transcriptomic dataset with previously published transcription factor-regulated gene sets using WormExp analysis and identified a substantial overlap with genes regulated by the GATA transcription factor ELT-2. These new analyses are described on page 7, lines 196-206: “To identify transcription factors potentially involved in intestinal distension-induced immune activation, we performed transcription factor enrichment analysis using WormExp on genes upregulated in N2 worms following aex-5 RNAi treatment. This analysis revealed a substantial overlap between aex-5 RNAi-induced genes and genes regulated by the GATA transcription factor ELT-2 (Figure S3D). We next examined whether USP-14-dependent immune genes overlapped with ELT-2-dependent immunity genes induced by intestinal distension. To this end, we identified innate immune genes common to both ELT-2-regulated gene sets and aex-5 RNAi-induced genes. Strikingly, these ELT-2-dependent intestinal distension-induced immune genes showed substantial overlap with USP-14-dependent immune genes (Figure S3E and Table S5), suggesting that USP-14 may regulate distension-induced immunity, at least in part, through ELT-2-dependent transcriptional programs.”

Finally, we have created a new table (Table S5) that specifies the identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense, for USP-14-dependent and independent genes.

Comment: __4. The authors' data suggest that in response to bacterial infection USP-14 upregulates the expression of bar-1, along with other components of the Wnt signaling pathway, which in turn upregulates innate immunity genes. This could be further substantiated by directly demonstrating that there are USP-14-regulated innate immunity genes whose induced expression in the presence of microbial pathogens also requires bar-1. Along those lines, an initial test would be to assess clec-60p::gfp expression in bar-1 animals versus bar-1;usp-14 double mutants, similar to the experiment whose results are reported in Fig. S4. If generating the bar-1;usp-14 double mutant is not feasible, then RNAi could be used to knockdown bar-1 expression in clec-60p::gfp;usp-14(tm1481) animals. To expand this analysis, the expression of the six innate immunity genes shown to be regulated upon intestinal distension in usp-14-dependent manner could be measured in the presence and absence of intestinal distension or microbial infection in bar-1 and bar-1;usp-14 animals by qRT-PCR. At a minimum, the authors should conduct a bioinformatics analysis to compare the USP-14-regulated innate immunity genes identified in their RNAseq studies to lists of known BAR-1 transcriptional targets to look for potential overlap. __Response: We agree that extending these analyses to qRT-PCR experiments examining additional immune genes would be informative. However, both bar-1 mutants and bar-1 RNAi-treated worms exhibited severe developmental and physiological defects, including sick and dead animals during development, likely reflecting the pleiotropic developmental roles of BAR-1. Although fluorescence imaging and survival assays could be performed by selectively transferring surviving adults, we were concerned that bulk collection of worms for qRT-PCR analyses would introduce confounding effects arising from developmental defects and reduced viability.

To further address the reviewer’s suggestion, we carried out a comparative analysis between USP-14-dependent intestinal distension-induced immune genes and previously identified BAR-1-dependent immune genes. Although transcriptome-wide datasets for BAR-1-dependent pathogen-induced immune genes are not currently available, an earlier study identified seven immune response genes regulated by BAR-1 during infection (PMID: 18981407). We found that six of these genes overlap with the USP-14-dependent intestinal distension-induced immune genes identified in our study. These analyses have now been added to the Results section and included in Table S5.

Comment: __5. While in their Discussion section the authors mention evolutionarily conserved roles for protein ubiquitination as means of immunomodulation, there are few if any comments regarding ubiqutination as a regulatory scheme in C. elegans innate immunity or how their findings enhance our understanding of this phenomenon. Ubiquitination affects C. elegans immunity at multiple levels, from avoidance behavior to gene regulation, and it seems appropriate for the authors to address this in order to more fully contextualize their findings. __Response: We thank the reviewer for the suggestion. We have now added a new paragraph to the Discussion that places our findings in the context of the existing literature on ubiquitination, deubiquitination, and innate immunity in C. elegans. The discussion is added on pages 10-11, lines 295-308: “Although ubiquitin-mediated signaling has emerged as a central regulator of innate immunity across metazoans (Jiang & Chen, 2011; Mello-Vieira & Dikic, 2026), the contribution of DUBs to host defense in C. elegans remains poorly understood. Previous studies in C. elegans have shown that ubiquitin-dependent processes regulate diverse aspects of immunity, including immune surveillance, xenophagy, and pathogen tolerance (Garcia-Sanchez et al, 2021). Perturbations in proteasome function have also been shown to activate surveillance immunity (Ghosh & Singh, 2026; Troemel et al, 2026), highlighting the importance of ubiquitin-associated pathways in sensing pathogen-induced cellular damage. However, most prior studies have focused on ubiquitin ligases, proteasome-associated pathways, or global ubiquitin signaling rather than on specific DUBs directly regulating antibacterial immune responses. To our knowledge, our study provides the first direct evidence that a specific DUB regulates antibacterial innate immunity in C. elegans. Thus, our findings establish USP-14 as a previously unrecognized regulator of host defense and identify deubiquitination as an important regulatory layer in intestinal distension-mediated immunity.”

__Minor comments __1. In the Results section, the authors state that "[k]nockdown of cec-10 led to only a marginal decrease in survival during P. aeruginosa infection" (lines 92 and 93) and that cec-10 "has minimal impact on C. elegans survival during infection" (lines 93 and 94). However, as reported in Supplemental Table 5 the magnitude of the calculated difference in mean survival time between animals treated with RNAi targeting cec-10 and untreated control animals (-20% to -24% and statistically significant in 3/3 replicates) closely approximates the difference in mean survival between usp-14 mutants and controls (-19% to -28% and statistically significant in 3/3 replicates), which the authors clearly find to be significant. If by this metric usp-14 is important for host defense, then so too is cec-10. In light of this, the authors should use different language to describe the impact of cec-10 knockdown on the susceptibility of C. elegans to microbial infection and the potential role of cec-10 in immunity.

Response: We chose not to pursue cec-10 further primarily because it lacks a clear human homolog and because the mutant exhibited reduced expression of the co-injection marker, raising the possibility of broader transgene-related effects. We have modified the text on page 4, lines 92-96: “Knockdown of cec-10 resulted in a significant reduction in survival during P. aeruginosa infection (Figure S1C). However, we did not pursue cec-10 further for two reasons: (i) cec-10(jsn20) mutants exhibited a modest but significant reduction in the myo-2p::mCherry co-injection marker (Figure 1D), raising the possibility of broader transgene-related defects, and (ii) cec-10 lacks a clear human homolog.”

Comment: __2. All of the micrographs in Fig. 1B appear very dark. The GFP expression in the control animals appears dim, making it difficult for the reader to compare the signal in those animals to the GFP expression levels in the mutants. I recommend adjusting the brightness level in an equivalent manner across all of the micrographs to account for this. __Response: We have increased the brightness of all the images, as suggested by the reviewer.

__Comment: __3. Fig. 1E depicts a gene structure diagram for usp-14 with the position of the point mutation in the jsn19 allele isolated in the authors' forward genetic screen indicated by the amino acid substitution symbol drawn over the second exon. Instead of mixing gene- and protein-level information about the jsn19 allele, I recommend replacing the gene structure diagram with a domain structure diagram of the USP-14 protein that depicts the conserved C19 peptidase and ubiquitin-like domains. The relative position of the E122K substitution should still be noted. __Response: __We have now updated Figure 1E to include the functional domains of USP-14 and mapped both the usp-14(jsn19) missense allele and the usp-14(tm1481) deletion allele onto the protein schematic.

Comment: __4. Since all of the information in Fig. 1F appears elsewhere in the text, I recommend eliminating this panel. __Response: We have removed it.

Comment: __5. Regarding the RNAseq analysis, the authors state that 1241 genes are upregulated upon aex-5 knockdown (line 162). The authors then ask which of these genes are regulated by usp-14 in the context of intestinal distension and find that 633 are upregulated a usp-14-dependent manner when aex-5 is targeted by RNAi and that 595 are upregulated even in the absence of usp-14 (Fig. 3D). This accounts for 1228 genes in total, not 1241. Can the authors explain this discrepancy? __Response: We thank the reviewer for carefully noting this discrepancy. The difference arises from the criteria used to classify genes into the categories shown in Figure 5D (previously Figure 3D). Specifically, genes uniquely upregulated in usp-14(tm1481) worms were defined as genes that were either exclusively induced in usp-14(tm1481) worms or expressed at levels more than 2-fold higher in usp-14(tm1481) worms compared to N2 worms. During this classification, 13 genes that were initially identified as upregulated in N2 worms following aex-5 RNAi were found to be expressed at levels more than 2-fold higher in usp-14(tm1481) worms than in N2 worms (Table S4). These genes were therefore reassigned to the “usp-14(tm1481)-specific” category in the Venn diagram. Consequently, the total number of genes represented in the Venn diagram becomes 1228 instead of 1241. To clarify this point, we have now added an explanation to the figure legend.

Comment: __6. For the sake of clarity, in the legend to Fig. 3D I recommend expanding the description of the categories of genes depicted in the Venn diagram by using the same language as in the first worksheet of Supplemental Table 4. __Response: We thank the reviewer for the suggestion. We have now added these details to the legend of Figure 5D (previously Figure 3D). The legend reads: “(D) Venn diagram showing the overlap between genes upregulated upon aex-5 RNAi in N2 and usp-14(tm1481) worms. The GO analyses for the biological processes of unique and common genes are shown. USP-14-dependent genes were defined as genes that were either exclusively upregulated in N2 worms or expressed at levels greater than 2-fold higher in N2 worms than in usp-14(tm1481) worms. USP-14-independent genes were defined as genes upregulated in both N2 and usp-14(tm1481) worms with expression differences of less than 2-fold between the two strains. Genes uniquely upregulated in usp-14(tm1481) worms were defined as genes that were either exclusively induced in usp-14(tm1481) worms or expressed at levels greater than 2-fold higher in usp-14(tm1481) worms than in N2 worms. Thirteen genes classified as upregulated in N2 worms were more than 2-fold higher in usp-14(tm1481) worms than in N2 worms (Table S4) and were therefore included in the usp-14(tm1481)-specific category.”

Comment: __7. In Fig. 4B, the authors' annotation indicates that there is a statistically significant difference (**, p __Comment: __8. In Fig. S5, the shade of blue used to represent the data from the nhr-49(nr2041);usp-14(tm1481);clec60p::gfp animals in panel E is different from that used to represent data from the same animals in panel B. This breaks the pattern of all of the other panels of this figure in which the data pertaining to a given phenotype are depicted in the same color. Also, in the symbol key in panel E there is an extra semi-colon before clec-60p::gfp that should be eliminated in the second genotype notation. __Response: We thank the reviewer for carefully examining the figure and for bringing these issues to our attention. We have made the changes.

Comment: __9. The authors' data show that USP-14 regulates bar-1 expression, and in the Discussion section they mention that in mammals beta-catenin is a substrate of USP14. Can the authors comment on the possibility of/evidence for BAR-1 autoregulation in C. elegans and the prospect of it being facilitated by USP-14? This could be a minor point to add to the Discussion. __Response: In both contexts, USP-14 appears to stabilize BAR-1 by regulating it at either the transcriptional or post-translational level. However, it is currently unknown whether BAR-1 regulates USP-14 expression and thereby participates in an autoregulatory mechanism. Nevertheless, we have added to the Discussion that USP14 may regulate the Wnt pathway through both transcriptional and post-translational mechanisms, depending on the biological context. __Reviewer #2 (Significance (Required)): __ The study described in this manuscript ties in to the findings from two prior genetic screens carried out in C. elegans that aimed to identify immune regulators (Ren et al., Cell Reports, 2022 and Labed et al., Immunity, 2018). Though their strategies differed, both of these previous studies uncovered a role for acetylcholine receptors in modulating the response to ingested microbial pathogens, especially when infection is associated with intestinal distension, indicating that a neuron-to-gut axis controls innate immunity in C. elegans. Labed and colleagues were the first to show that activation of this pathway results in the upregulation of genes encoding Wnt signaling pathway components, including the worm ortholog of beta-catenin called bar-1, which are necessary for the expression of immune effectors in the intestine. The Labed study also revealed that protein ubiquitination could contribute to regulating host defense gene induction because knockdown of lin-23, the substate binding subunit of a ubiquitin ligase complex that mediates BAR-1 degradation, results in constitutive expression of clec-60p::gfp, the same transcription reporter used by Ghosh and Singh as a readout for the expression of innate immunity genes. In their screen that revisits the Ren et al. approach, Ghosh and Singh find that another protein implicated in regulating protein stability via ubiquitination status, USP-14, also controls the expression of innate immunity genes in response to intestinal distension. Interestingly, their data indicate that it does so by upregulating bar-1. This discovery therefore adds an element of mechanistic detail regarding the regulation of Wnt signaling in immunity. While the Labed data suggest that ubiquitination may regulate BAR-1 at the post-translational level, Ghosh and Singhs' results indicate a second layer of regulation of bar-1 at the transcriptional level that also appears to involve ubiquitination. In this case, USP-14 is predicted to modulate the ubiquitination status of a yet-to-be-identified substrate that directly or indirectly governs bar-1 expression. The authors' findings thus bring the field closer to having a complete picture of the Ach-Wnt pathway in C. elegans. As they point out in the Discussion section of their manuscript, ubiquitination is an evolutionarily conserved yet complex means of tuning the immune system. The work described here helps to shed light on this important immune regulatory mode and could have implications for aspects of epithelial immunity that are in common to both invertebrates and vertebrates.

Response: We thank the reviewer for providing such a thoughtful overview of the field and for placing our findings in the context of previous studies on intestinal distension-induced immunity in C. elegans. We also sincerely appreciate the reviewer’s constructive feedback and insightful comments, which have helped us improve the quality and clarity of the manuscript.

My research interest and specific area of expertise pertains to evolutionarily conserved genetic pathways that control healthspan through affecting cellular resilience later in life. Using C. elegans as a surrogate for aging humans, my group studies age-dependent changes in the activity of regulatory modules that protect older animals from the molecular damage associated with intrinsic and extrinsic sources of cellular stress, with a particular emphasis on microbial infection and oxidative stress.

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

C. elegans are soil-dwelling nematodes that feed on bacteria and fungi and thus must be able to distinguish between innocuous and pathogenic species of microbes to survive. Though they lack adaptive immunity, these animals have an ancient version of an innate immune system that has no circulating sentinel or phagocytic cells yet can still mount a response to pathogen exposure. A consequence of the mode of infection of some ingested bacterial pathogens is intestinal distension which by itself, even in the absence of pathogens, is sufficient to trigger the expression of genes encoding immune effectors, including proteins that are bactericidal. The complete mechanistic scheme connecting intestinal distension to the expression of immunity genes has not been resolved, motivating the authors to perform a forward genetic screen for additional components of this pathway. One mutant that the authors isolated was usp-14, encoding an evolutionarily conserved deubiqutinating enzyme. Functional analysis revealed that usp-14 confers protection from microbial pathogens and that the intestine is its primary site of action for its role in host defense. The authors' data indicate that while USP-14 regulates the expression of innate immunity genes that are induced by intestinal distension, surprisingly it functions independently of several canonical innate immune signaling pathways, including the pmk-1/p38 MAPK pathway. Instead, USP-14 appears to act through Wnt signaling to regulate immune effectors by upregulating the expression of several components of that pathway, including the C. elegans ß-catenin ortholog bar-1. This places usp-14 within a gut-brain axis previously shown to control the C. elegans innate immune response through acetylcholine-mediated activation of Wnt signaling. The authors' findings provide new mechanistic insight to this pathway and add to the understanding of ubiqutination as an immune regulatory module.

Major comments

1. There are three types of experiments in which the authors use the same set of controls across several different figure panels, as stated in the legend to Figure 2. First, when quantifying GFP levels of clec-60::gfp in RNAi-treated animals, the authors use the same clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls for Fig. 1K, 2C, and 2G. For infection assays with S. aureus NCTC8325, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2E are the same as the ones used in Fig. 1M. Similarly, for infection assays with P. aeruginosa PA14, the survival plots for the clec-60p::gfp and usp-14(jsn19);clec-60p::gfp controls shown in Fig. 2I is the same as was used for Fig 1I. In each case, if the authors in fact collected all of the data for each strain that they studied at the same time but then chose to parse larger datasets into separate figure panels to make it more clear to the reader, then this approach is valid but the authors need to explicitly state that this is what they did. However, if the data pertaining to the control strains were collected at a different time or if it comes from a separate biological replicate, then re-using data from the controls is not appropriate because it would not accurately reflect the specific conditions of the experiment to which the data are being compared. If this is indeed the scenario, then the authors will need to repeat these experiments and include the appropriate control in each iteration.
2. From the legends describing figure panels that include data pertaining to clec-60p::gfp expression levels as assessed by fluorescence microscopy it seems that, in general, the authors measured GFP fluorescence in about 30 animals to produce quantitative data. How many biological replicates of these types of experiments were carried out? This is not explicitly stated in the section describing fluorescence imaging in the Methods section. Following the description of their methodology regarding statistical analysis of survival curves from microbial infection assays, however, the authors state that, "[a]ll experiments were performed independently at least three times unless otherwise noted." Does this statement apply to microscopy or only to experiments involving infection assays? If the data reporting quantitation of GFP signal is based on only 30 animals, then additional biological replicates are necessary, along with appropriate statistical analyses.
3. The authors have made all of the RNASeq data publicly available on the Sequence Read Archive, and they include data from several pairwise comparisons for differential gene expression analysis in their supplemental files. One of the most important facts to come out of the authors' Gene Ontology analyses of their RNASeq data is that the genes that are upregulated in a usp-14-dependent manner upon intestinal distension are enriched for those whose products play a role in innate immunity/host defense. The authors should say more about these genes. Are there any commonalities between them with regard to function? Are any of them targets of transcription factors that are known to function in C. elegans innate immunity? If so, this could provide clues as to what the substrates of USP-14 might be. Importantly, the specific identity of the genes assigned in the GO analyses to biological processes pertaining to innate immunity and host defense should be revealed in a supplemental file, and designated as being dependent on or independent of usp-14 for their expression during intestinal distension.
4. The authors' data suggest that in response to bacterial infection USP-14 upregulates the expression of bar-1, along with other components of the Wnt signaling pathway, which in turn upregulates innate immunity genes. This could be further substantiated by directly demonstrating that there are USP-14-regulated innate immunity genes whose induced expression in the presence of microbial pathogens also requires bar-1. Along those lines, an initial test would be to assess clec-60p::gfp expression in bar-1 animals versus bar-1;usp-14 double mutants, similar to the experiment whose results are reported in Fig. S4. If generating the bar-1;usp-14 double mutant is not feasible, then RNAi could be used to knockdown bar-1 expression in clec-60p::gfp;usp-14(tm1481) animals. To expand this analysis, the expression of the six innate immunity genes shown to be regulated upon intestinal distension in usp-14-dependent manner could be measured in the presence and absence of intestinal distension or microbial infection in bar-1 and bar-1;usp-14 animals by qRT-PCR. At a minimum, the authors should conduct a bioinformatics analysis to compare the USP-14-regulated innate immunity genes identified in their RNAseq studies to lists of known BAR-1 transcriptional targets to look for potential overlap.
5. While in their Discussion section the authors mention evolutionarily conserved roles for protein ubiquitination as means of immunomodulation, there are few if any comments regarding ubiqutination as a regulatory scheme in C. elegans innate immunity or how their findings enhance our understanding of this phenomenon. Ubiquitination affects C. elegans immunity at multiple levels, from avoidance behavior to gene regulation, and it seems appropriate for the authors to address this in order to more fully contextualize their findings.

Minor comments

1. In the Results section, the authors state that "[k]nockdown of cec-10 led to only a marginal decrease in survival during P. aeruginosa infection" (lines 92 and 93) and that cec-10 "has minimal impact on C. elegans survival during infection" (lines 93 and 94). However, as reported in Supplemental Table 5 the magnitude of the calculated difference in mean survival time between animals treated with RNAi targeting cec-10 and untreated control animals (-20% to -24% and statistically significant in 3/3 replicates) closely approximates the difference in mean survival between usp-14 mutants and controls (-19% to -28% and statistically significant in 3/3 replicates), which the authors clearly find to be significant. If by this metric usp-14 is important for host defense, then so too is cec-10. In light of this, the authors should use different language to describe the impact of cec-10 knockdown on the susceptibility of C. elegans to microbial infection and the potential role of cec-10 in immunity.
2. All of the micrographs in Fig. 1B appear very dark. The GFP expression in the control animals appears dim, making it difficult for the reader to compare the signal in those animals to the GFP expression levels in the mutants. I recommend adjusting the brightness level in an equivalent manner across all of the micrographs to account for this.
3. Fig. 1E depicts a gene structure diagram for usp-14 with the position of the point mutation in the jsn19 allele isolated in the authors' forward genetic screen indicated by the amino acid substitution symbol drawn over the second exon. Instead of mixing gene- and protein-level information about the jsn19 allele, I recommend replacing the gene structure diagram with a domain structure diagram of the USP-14 protein that depicts the conserved C19 peptidase and ubiquitin-like domains. The relative position of the E122K substitution should still be noted.
4. Since all of the information in Fig. 1F appears elsewhere in the text, I recommend eliminating this panel.
5. Regarding the RNAseq analysis, the authors state that 1241 genes are upregulated upon aex-5 knockdown (line 162). The authors then ask which of these genes are regulated by usp-14 in the context of intestinal distension and find that 633 are upregulated a usp-14-dependent manner when aex-5 is targeted by RNAi and that 595 are upregulated even in the absence of usp-14 (Fig. 3D). This accounts for 1228 genes in total, not 1241. Can the authors explain this discrepancy?
6. For the sake of clarity, in the legend to Fig. 3D I recommend expanding the description of the categories of genes depicted in the Venn diagram by using the same language as in the first worksheet of Supplemental Table 4.
7. In Fig. 4B, the authors' annotation indicates that there is a statistically significant difference (**, p<0.01) in the fluorescence signal from clec-60p::gfp in usp-14(jsn19);aex-5(sa23);clec-60p::gfp_EV versus usp-14(jsn19);aex-5(sa23);clec-60p::gfp_bar-1 animals. This is likely a typographical error that should be changed to "ns" to indicate no significant difference in the fluorescence signal between these two groups, which is consistent with what the data show and with the authors' description of these data in the text (lines 211-214).
8. In Fig. S5, the shade of blue used to represent the data from the nhr-49(nr2041);usp-14(tm1481);clec60p::gfp animals in panel E is different from that used to represent data from the same animals in panel B. This breaks the pattern of all of the other panels of this figure in which the data pertaining to a given phenotype are depicted in the same color. Also, in the symbol key in panel E there is an extra semi-colon before clec-60p::gfp that should be eliminated in the second genotype notation.
9. The authors' data show that USP-14 regulates bar-1 expression, and in the Discussion section they mention that in mammals beta-catenin is a substrate of USP14. Can the authors comment on the possibility of/evidence for BAR-1 autoregulation in C. elegans and the prospect of it being facilitated by USP-14? This could be a minor point to add to the Discussion.

Significance

The study described in this manuscript ties in to the findings from two prior genetic screens carried out in C. elegans that aimed to identify immune regulators (Ren et al., Cell Reports, 2022 and Labed et al., Immunity, 2018). Though their strategies differed, both of these previous studies uncovered a role for acetylcholine receptors in modulating the response to ingested microbial pathogens, especially when infection is associated with intestinal distension, indicating that a neuron-to-gut axis controls innate immunity in C. elegans. Labed and colleagues were the first to show that activation of this pathway results in the upregulation of genes encoding Wnt signaling pathway components, including the worm ortholog of beta-catenin called bar-1, which are necessary for the expression of immune effectors in the intestine. The Labed study also revealed that protein ubiquitination could contribute to regulating host defense gene induction because knockdown of lin-23, the substate binding subunit of a ubiquitin ligase complex that mediates BAR-1 degradation, results in constitutive expression of clec-60p::gfp, the same transcription reporter used by Ghosh and Singh as a readout for the expression of innate immunity genes. In their screen that revisits the Ren et al. approach, Ghosh and Singh find that another protein implicated in regulating protein stability via ubiquitination status, USP-14, also controls the expression of innate immunity genes in response to intestinal distension. Interestingly, their data indicate that it does so by upregulating bar-1. This discovery therefore adds an element of mechanistic detail regarding the regulation of Wnt signaling in immunity. While the Labed data suggest that ubiquitination may regulate BAR-1 at the post-translational level, Ghosh and Singhs' results indicate a second layer of regulation of bar-1 at the transcriptional level that also appears to involve ubiquitination. In this case, USP-14 is predicted to modulate the ubiquitination status of a yet-to-be-identified substrate that directly or indirectly governs bar-1 expression. The authors' findings thus bring the field closer to having a complete picture of the Ach-Wnt pathway in C. elegans. As they point out in the Discussion section of their manuscript, ubiquitination is an evolutionarily conserved yet complex means of tuning the immune system. The work described here helps to shed light on this important immune regulatory mode and could have implications for aspects of epithelial immunity that are in common to both invertebrates and vertebrates.

My research interest and specific area of expertise pertains to evolutionarily conserved genetic pathways that control healthspan through affecting cellular resilience later in life. Using C. elegans as a surrogate for aging humans, my group studies age-dependent changes in the activity of regulatory modules that protect older animals from the molecular damage associated with intrinsic and extrinsic sources of cellular stress, with a particular emphasis on microbial infection and oxidative stress.

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 manuscript, the authors describe the discovery of a molecular regulator of the immune transcriptional program, which is activated by intestinal distension upon bacterial colonization of the C. elegans intestine. Taking advantage of the fact that inhibition of aex-5 is known to cause intestinal distension and a C-type lectin gene clec-60 as a marker for the immune response to intestinal distension (clec-60p::gfp), the authors performed a forward genetic screen for suppressors of the immune response activation. Of the two mutants isolated, they focused on the stronger suppressor, which corresponded to a cysteine-type DUB, the Ubiquitin Specific Peptidase-14 (usp-14). Through rescue experiments, phenocopy analyses, and quantitative RT-PCR, they validated usp-14 as the causal gene and initiated characterization of its role in immune response activation. To this end, the authors investigated the tissue of action, identifying the intestine as the tissue in which usp-14 mediates the regulation of the immune response. Through transcriptomic analyses, they found that the signalling pathway likely regulated by usp-14 in response to intestinal distension is the Wnt pathway, as they have observed reduction in the transcriptional level of some of the Wnt pathway components in usp-4(tm1481), in response to infection with S. aureus. Additionally, transcriptomic data indicate that usp-14 plays a role in immunity regulation even in the absence of infection. Based on these findings, the authors propose that usp-14 has a dual role in immune regulation: one in surveillance immunity, preventing overactivation of immune responses, and another as a mediator of pathogen-induced responses, such as those triggered by P. aeruginosa or S. aureus. The experiments are rigorous and the results robust; however, some points would benefit from further investigation or clarification.

The expression domain of usp-14 appears to be quite expanded based on single cell RNAseq data (e.g. PMID: 28818938) therefore it is likely that the transgenes used for expression analysis are lacking key regulatory information. Alternative methods like smFISH would be more appropriate to characterise the spatiotemporal pattern of usp-14 expression in more detail.

The mutation mapped in usp-14(jsn19) is a missense mutation (E122K) that suppresses the immune response to a degree comparable to the usp-14(tm1481) deletion allele. However, the authors do not show the functional domains in Fig. 1E potentially affected by this missense mutation.

How USP-14 regulates Wnt and how Wnt signalling relates to activation of immune responses is not fully supported. Are the Wnt components mentioned in the study induced specifically in the intestine upon infection and does USP-14 act in the intestine in the context of this regulation? How do the authors interpret that both Wnt ligands and receptors are induced ? Does Wnt signalling appear as a GO term in the transcriptomic analysis? The authors can include Wnt signalling components in the analysis of the transcriptomic results.

Overall, in most of the figures, the micrographs are in general quite dark and exhibit poor contrast between signal and background, particularly in Fig. 1, panels B and J, and Fig. 2, panels B and F (upper rows). Even though these panels are intended to show absence of response, the outlines of the worms are difficult to discern.

In Figure S3, panels A and B, the pmk-1(km25); usp-14(tm1481) animals subjected to aex-5 RNAi show some level of fluorescence/response induction comparable to pmk-1(km25) alone. This observation is not discussed in the text.

Significance

The work is interesting because it expands some previous work in the field demonstrating immune response induction as a consequence of intestinal distension even in the absence of bacterial infection. This is known to be mediated by the neuronal acetylcholine receptor ACC-4, which signals to the intestine where it regulates immune genes via the Wnt pathway. However, how USP-14 relates to ACC-4 is currently unclear and whether USP-14 function is really required in the intestine to control Wnt signalling is not demonstrated. The authors should include a model to describe how their findings relate to the previous literature and how USP-14 may link mechanistically to Wnt signalling pathway activation.

It remains also unclear whether usp-14 is the only deubiquitinase involved in intestinal distension-induced signalling via the Wnt pathway, or whether other paralog usp genes might also contribute to regulation of immune-responsive transcription. Notably, several mammalian deubiquitinases have established roles in cancer suppression and inflammatory response and innate immunity in other systems so this would increase the potential significance of the work.

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

Please see the uploaded "Response to Reviewers" PDF file which contains figures.

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 the manuscript titled "Multiple Molecular Pathways to Longevity: Opposing Gene Expression Programs Define Distinct Aging Strategies", the authors investigated diverse genetic pathways that contribute to lifespan extension in Caenorhabditis elegans and aimed to identify shared and distinct molecular mechanisms among various longevity mutants. Through comprehensive RNA sequencing of different longevity mutants representing seven distinct pathways, the authors showed that these mutants cluster into three primary groups based on their gene expression profiles. This transcriptomic analysis revealed that while some longevity genes are commonly regulated across multiple pathways, others exhibit opposing expression patterns, suggesting that distinct molecular strategies can lead to increased lifespan. Specifically, they identified a set of 196 genes that are consistently upregulated in most longevity mutants, many of which are involved in innate immunity and stress defense. By performing RNAi-based screening, the authors further validated the functional roles of several candidates, including C08F11.7, ugt-62, and K05C4.9, supporting their contributions to longevity and stress resistance. The authors conclude that longevity is mediated through multiple molecular pathways and provide a public online tool to study these complex transcriptomic landscapes.

Major comments

1. While the authors identified a set of 196 upregulated genes, the rationale for narrowing these down to the three final candidates (C08F11.7, ugt-62, and K05C4.9) is not clearly described. The authors show that genetic inhibition of several genes, including DC2.5, C05B5.5, T07C4.5, and W03B1.7, decreases lifespan in both nuo-6 mutants and wild-type animals. However, the authors did not describe why these additional validated candidates, which also showed significant effects on longevity, were not pursued for further characterization. The authors should explicitly state the criteria used to prioritize these three genes over the other validated genes.
2. The authors conclude that longevity can be mediated by multiple molecular pathways. However, it remains unclear whether these distinct strategies can operate simultaneously or are mutually exclusive. The authors need to test whether lifespan extension in a Group 1 mutant is further enhanced or suppressed by the knockdown of a key Group 2-specific genes. These experiments would help determine these pathways act additively, antagonistically, or as partially redundant survival programs.
3. The authors provide interesting data on overexpression of the three candidate genes. However, whereas C08F11.7 clearly demonstrates both necessity and sufficiency for lifespan extension, overexpression of ugt-62 and K05C4.9 does not independently extend lifespan. To strengthen the manuscript, the authors should expand the discussion of these divergent results and clarify possible explanations.
4. Key citations are missing and the authors should add multiple citations including the following ones. Please cite the following paper and discuss the authors' finding with respect to the related work (Lee et al PMID: 40814218). Add citations in the sentence describing changes in the transcriptome of C. elegans associated with age (Lee et al., PMID: 38508494). Furthermore, please cite papers describing the overviews of survival assay using C. elegans (Kwon et al., PMID: 40436148, Hwang et al., PMID: 40436147).

Minor comments

1. To improve readability, please provide the full names for all abbreviations at their first appearance in the manuscript.
2. Please ensure that the labels in the figures match the text exactly. For instance, if different promoters are used for generating overexpression animals, it may be helpful to indicate the specific promoter in the figure panel or legend for clarity.
3. For all lifespan and stress resistance assays, please include the total number of animals (n) and the number of independent biological replicates (N) in the figure legends to confirm statistical reliability.
4. Please clearly specify the exact developmental stage of the animals used for the survival assays in the Materials and Methods section.

Referees cross-commenting

I also agree with reviewer #1's comments and recommend revision to further improve the manuscript.

Significance

This study provides a systematic, side-by-side transcriptomic comparison of nine genetically distinct long-lived C. elegans mutants, revealing that lifespan extension arises from both shared and opposing gene expression programs. By identifying three distinct longevity groups and demonstrating that key pathways can be modulated in opposite directions to achieve long life, the work challenges the notion of a single universal transcriptional signature of aging. Importantly, functional validation shows that select commonly regulated genes can directly modulate lifespan and stress resistance, highlighting actionable molecular targets for promoting healthy aging.

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

This manuscript by Rudich ZD et al. systematically profiled the transcriptomic changes in nine long-lived C. elegans mutants and presented a careful and informative comparative analysis of these aging-related changes. In addition to these valuable datasets and bioinformatics analyses, the authors performed a large-scale RNAi screen to assess the role of the differentially expressed genes (DEGs) in these mutants and identify several potential targets to promote healthy aging. Moreover, the authors have provided a user-friendly website to examine genes of interest in those longevity mutants from their datasets.

Major comments

The conclusions of this manuscript are generally well supported. The study is also technically sound. Yet, I still have a few concerns that should be carefully addressed.

1. Although I myself believe that the datasets in this study should be more consistent and comprehensive, the authors should perform a data mining analysis of previously reported transcriptomic changes of these mutants or similar mutants in the same longevity pathway and compare the reported changes with their findings to highlight the necessity and advances of this study.
2. This manuscript does not perform any regulon or transcription factor (TF) analyses. TFs are the drivers of the transcriptomic changes and multiple conserved TFs (e.g., daf-16) have already been identified in these pathways. Therefore, it is necessary to examine and compare the regulons/TFs in these new datasets by bioinformatics. Such analyses can: a) provide more information of the driving force of these transcriptomic changes; b) show the role of these known longevity TFs; c) propose new TFs driving longevity; d) support the findings of 'longevity strategies' and 'longevity groups' from the perspective of TFs.
3. osm-5 and daf-2 are categorized into two different groups in this study. Since the longevity of cilia (-) mutants is through daf-16, the same master TF driving daf-2 longevity, please perform further analyses or discussion to clarify this issue.
4. This manuscript focused on genes whose RNAi suppressed the mutants longevity. Please also use bioinformatics to analyze the functions of those whose RNAi extends the mutants longevity, because these genes could tell the health price these mutants pay and help improve ageing interventions by reducing side effects.
5. (OPTIONAL) I strongly suggest a comprehensive comparison of these transcriptomic changes in long-lived mutants with published age-related transcriptomic changes in wild type worms. This comparison will significantly All the suggested analyses are pure bioinformatics and should be realistic to finish in several months.

Minor comments

I also have a few minor comments on data presentation: 1. Please further clarify the analysis of DEGs correlated with lifespan extension in Fig. 2 by a depiction. In Fig. 2C and D, please label data dots from different strains with different colors. 2. In Fig. 3 and S20, please label the percentage of overlapping genes on top of each bars.

Referees cross-commenting

I agree with Reviewer #2's comments and would suggest giving the authors enough time to revise their manuscript.

Significance

Compared to previous transcriptomic analyses of these mutants in differen reports, this study minimized the technical variations and benefitted from the advances in RNA-Seq technology and bioinformatics tools. Therefore, it should provide a more consistent and comprehensive view of the molecular mechanisms underlying the longevity of these mutants. The datasets in this manuscript are valuable to other researchers in the biology of aging.

Meanwhile, since these mutants have been extensively studied, the advance of this study in unknown mechanisms remains limited. Therefore, I would recommend its publication as a 'Resource' article after addressing my concerns.

(I am an expert in the biology of ageing, using C. elegans and mouse as major models.)

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 reviewers for their constructive evaluation of our manuscript. We are pleased by the overwhelmingly positive consensus regarding the quality and significance of our data. In particular, the reviewers highlighted that this is a "nice, clean study with interesting data" and noted that our in vivo functional genetic findings in the Drosophila wing are "clearly a strength" that "moves the paper beyond cell-culture correlations" to provide a "simple, straightforward take-home message".

The principal critique across the reports concerns the extent of direct mechanistic evidence linking Groucho (Gro) to regulation of the early elongation checkpoint. Several reviewers suggested additional genomic experiments, including RNA-seq, PRO-seq, or Pol II ChIP approaches, to further examine transcription and pausing behaviour. However, we would like to flag up that genomic datasets addressing these questions across multiple Drosophila cell lines have already been published previously, including work from our own group and others.

The primary objective of the current study is therefore not to replicate these existing genomic analyses, but rather to build directly upon them. We identify a consistent genomic association between Gro and pausing/elongation factors across cell types. Importantly, we extend these findings beyond genomic correlations through in vivo genetic analysis in the developing Drosophila wing.

1. Description of the planned revisions

• *

• *

Reviewer 1

The figures and text could lay out the logic of the genetic interactions for non-Drosophila readers. For example, the comparison of single and double copies of Gro-RNAi to combinatorial knockdowns, when it is additive, and when it is interpreted as synergistic.

The statistical analyses presented in Figure 5C, including Fisher’s exact tests comparing phenotype distributions between genotypes, were intended to address the distinction between additive and synergistic genetic interactions. However, we agree that the presentation of these comparisons could potentially be made clearer for readers less familiar with Drosophila genetic interaction assays. We would therefore be open to revising the presentation of Figure 5 and the accompanying explanatory text following editorial guidance and with consideration of the intended readership of the eventual journal.

The statistical analysis of the phenotype distributions should be shown more clearly (Fig. 5B).

Figure 5B is intended to present the distribution of observed phenotypic classes and does not include statistical comparisons. A similar analysis has been published for experiments looking at the phenotypes of moderate Groucho overexpression in the wing in the presence of HDAC inhibitors (Winkler et al., 2010 doi.10.1371/journal.pone.0010166). Statistical analyses of the genetic interaction experiments are presented separately in 5C. We therefore believe the current presentation of Figure 5B is appropriate for illustrating phenotype frequencies rather than statistical inference, but we will consider moving this panel to the Supplementary material.

Minor comments

-Figure 5 would gain clarity if the phenotype classes/panel letters were shown more clearly on the images. -The legends of the wing figures should be expanded, especially for readers outside the Drosophila field. -"in vivo" should be italicised consistently.

We agree that clearer labelling of phenotype classes, panel annotations and expanded figure legends could improve the accessibility of Figure 5, particularly for readers less familiar with Drosophila wing phenotypes and genetic interaction assays. We would therefore be open to revising the presentation of this figure and its accompanying legends in a future revised version.

We thank the reviewer for noting the typographical inconsistency of italics for in vivo. This will be corrected during manuscript revision and proofing.

__Reviewer #2 __

Reviewer #2 (Significance (Required)):

I think this is nice little paper providing a simple, straightforward take-home message. It does not conceptually shake the world, and the evidence consists of (nice) correlations, with no direct proof put forward for the conclusions. I am not a Drosophila geneticist but probably rather an 'expert' on basic transcription mechanisms. I think the data in the paper are of high quality, if limited in scope, and that the conclusions are supported by the results, but I do not think the results or conclusions will have a big audience. Having said that, I found it interesting to learn about this group of repressors and their likely mode of action.

On the other hand, it is worth emphasizing that proteins such as NELF and CDK9 would arguably be expected to be found at very many genes, as promoter-proximal pausing does exist at a plethora of genes, also genes that are house-keeping genes, ie not regulated by cell type or stimuli. So, lots of genes with pausing are not regulated by modulation of pausing. So, basically, the fact that knockdown of the repressor Groucho and loss of pausing is additive does not in my opinion necessarily mean that Groucho works by stabilizing pausing. Although it is admittedly a reasonably assumption, Groucho could also work by repressing transcription initiation; the genetic outcomes of 'double relief' would be the same, ie higher transcription levels. I think a brief comment to this effect might be appropriate, especially in the absence of (difficult to obtain) direct evidence that the transcription initiation step is not affected by Groucho.

While we agree that the current study does not directly exclude possible effects of Groucho on transcription initiation, previously published work has already provided evidence arguing against repression by Groucho occurring primarily through inhibition of transcription initiation or prevention of pre-initiation complex assembly. Groucho-bound transcriptional start sites were previously shown to retain RNAP II occupancy, active chromatin features, and detectable basal transcriptional activity despite repression (Kaul et al., 2014).

To acknowledge this possibility and explain why it is unlikely, we will add the sentence “While effects on transcription initiation cannot be completely excluded, previous work argues against Gro repressing transcription primarily through inhibition of transcription initiation. Gro-bound promoters remain accessible, overlap RNAP II occupancy, and retain active chromatin features and basal transcriptional activity” to the start of the third paragraph of the Discussion.

Reviewer #3

The methods section is lacking details on how ChIP-seq was performed in the BG3 cell line. The methods section does a good job of indicating how the data were processed. Information on the antibodies and conditions used is critical, as is whether spike-in controls were used.

The generation of the ChIP-seq data from BG3 cells has already been published. __We will add the line “The production of ChIP-seq datasets for Gro binding in Kc167, S2R+ and BG3 cells has been described elsewhere (Kaul, Schuster and Jennings, 2014; Bar-Cohen et al., 2023)” in the Analysis of ChIP-seq data subsection of the Methods. __

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

• *

__Reviewer #1 __ Major comments 1. The main weakness is the lack of a mechanistic link between Gro and the early elongation checkpoint. This is really the main point for this reviewer. The manuscript builds an interesting model, and the data support a functional connection between Gro and pausing-related factors, but the mechanistic link is absent. At present, the paper relies on co-localisation of ChIP peaks and genetic interaction in vivo. This is interesting and supportive, but with several possible interpretations. The title and some parts of the text are thus a bit stronger than what is directly demonstrated. Two possibilities could be proposed: either tone down the mechanistic claim or strengthen it experimentally. A more direct assay of pause release or productive elongation after Gro depletion at endogenous targets would be highly valuable. For example, Gro-KD followed by Pol II Ser2-P ChIP, or promoter vs. gene body analysis on Gro-bound genes, ideally comparing genes with Gro at TSS vs. not-TSS, would greatly support the proposed model. If the assay is established, this seems feasible in about 4 months.

We thank the reviewer for this thoughtful comment. We agree that the current study does not directly measure genome-wide RNAP II pause release following Gro depletion. However, several key observations linking Gro with promoter-proximal pausing have already been published and are summarised in the Introduction. Previous work demonstrated that Gro occupancy correlates with paused genes and that depletion of Gro reduces RNAP II pausing and increases elongating RNAP II at the endogenous E(spl)mbeta-HLH locus, an established target gene of Groucho-mediated repression (Kaul et al., 2014; doi.10.1371/journal.pgen.1004595). We also note that several of the experiments proposed by the reviewer have already been addressed in previous work. Specifically, Kaul et al. (2014) demonstrated that Gro depletion increases elongating RNAP II (Ser2-P) at the endogenous E(spl)mbeta-HLH locus while total promoter-associated RNAP II occupancy remains largely unchanged. Promoter versus gene body analyses in that study further supported a role for Gro in regulating progression through the early elongation checkpoint rather than transcription initiation.

The aim of the current manuscript was therefore to build upon these earlier mechanistic and genomic observations by asking whether the relationship between Gro and pausing-associated factors extends across multiple cell types and whether it has functional significance in vivo. By integrating comparative genomic analyses with sensitised developmental genetic assays in the wing, we provide evidence that Gro functionally interacts with multiple regulators of the early elongation checkpoint during development.

The bioinformatic part could be strengthened on "distinct TF repertoires" between cell types.The authors interpret the cell type-specific Gro recruitment as reflecting distinct transcription factor repertoires in BG3, Kc167 and S2R+ cells. This is interesting, but not really shown. To make this point more strongly, the author could provide a map of TF expression across different cell types, especially for the TFs corresponding to the enriched motifs they discuss. Otherwise, this remains speculative.In line, the manuscript discusses enriched motifs in BG3 and compares them to Kc167 and S2R+ cells, but this remains a bit descriptive. A clearer side-by-side comparison would strengthen the paper. This is particularly relevant to the motifs used in interpreting cell type-specific recruitment.

The interpretation that cell type-specific Gro recruitment reflects differences in transcription factor repertoires is based on several previously established observations already described in the manuscript. BG3 cells are derived from the larval CNS, whereas Kc167 and S2R+ cells are embryonic haemocyte-like lines (Cherbas et al., 2011; doi.10.1101/gr.112961.110). Transcriptomic analyses have further shown that these Drosophila cell lines maintain stable and distinct lineage-associated transcriptional identities, including differences in transcription factor expression (Cherbas et al., 2011). Given the diversity of transcription factors known to recruit Gro, the observed cell-type-specific binding patterns and motif enrichments are consistent with the distinct lineage-associated transcriptional programmes previously described for these cell lines.

1. Several overlap analyses could be discussed more in depth. A few statements feel too strong for the actual percentages. For example, the GAF overlap in BG3 is around 51% genome-wide and 56% at TSS, which is meaningful, but not especially high. The text already states that it is not universal, and this point could be discussed more clearly.

We note that the manuscript already explicitly states that overlap between Gro and GAF is not universal. Given the diversity of factors known to recruit Gro and the broad genomic distribution of GAF, we consider overlap frequencies of approximately 50% to represent a substantial association, particularly at transcription start sites. Importantly, the interpretation does not rely on complete co-occupancy between these factors, but rather on the observation that Gro-bound regions show significant enrichment for multiple factors associated with promoter-proximal pausing across different cell types.

Similarly, for the UpSet plot, the wording around the "most frequent" combination could be toned down, because this is not a dominant pattern.

The statement that the overlap between Gro, Nelf-E, GAF, Cdk9 and RNAP II represents the “most frequent” combination refers specifically to the relative frequency of the intersection categories within the UpSet analysis. In this context, the overlap between all five factors represents the largest intersection category identified (306 of 649 Gro peaks), with the next most frequent category containing substantially fewer peaks (90 of 649). We therefore feel that the current wording accurately describes the distribution observed in the analysis.

More generally, I think the manuscript needs a clearer quantitative breakdown of TSS versus non-TSS peaks for the overlap analyses with NELF, GAF, Cdk9 and CycT. Several interpretations depend on this distinction, and right now, this is not always clear enough.

The overlap analyses presented in Figure 3 explicitly distinguish between TSS and non-TSS peaks, and the corresponding quantitative overlap frequencies are described in the Results section. We do not consider that additional breakdowns are required for interpretation of the current data as this distinction is already incorporated into both the analyses and figure presentation.

The "enhancer chromatin" interpretation is interesting, but not fully integrated with the genomic distribution. The observation that Gro is enriched in open enhancer-type chromatin is interesting and supports the idea that Gro does not act mainly through classical repressed chromatin. However, Gro peaks are also enriched at promoters and introns, and this reviewer feels that the manuscript does not fully connect these observations. Where are these enhancer-type peaks located exactly? Are they often intronic? Can this be correlated with the distribution of Gro peaks? This would help the reader and also strengthen the discussion because intronic Gro peaks are present in the data, but are not well integrated into the model.

In the current manuscript, “enhancer chromatin” refers to chromatin states defined by combinations of enhancer-associated histone modifications, including H3K4me1, H3K27ac and H3K56ac as defined by Skalska et al.,2015 (doi.10.15252/embj.201489923), rather than exclusively to distal intergenic regulatory regions. As described in the chromatin-state analysis, these enhancer-associated chromatin signatures do occur at intronic regulatory regions, including regions classified as active intron chromatin. We therefore do not consider the enrichment of Gro peaks at promoters, enhancers and intronic regions to be mutually exclusive observations within this framework.

Intronic enhancer localisation is common in Drosophila, where the compact organisation of the genome results in many developmental regulatory elements residing within introns (Arnold et al., 2013; doi.10.1126/science.1232542). We therefore consider the presence of Gro peaks within intronic regions to be fully consistent with the observed enrichment of Gro binding within enhancer-associated chromatin states.

The in vivo part is a strength, but some important points need clarification.The in vivo section is a clear highlight of the manuscript. It gives functional relevance to the model and moves the paper beyond cell-culture correlations. That said, a few points need to be clearer:-RNAi efficiency is not clear for the tested genes, especially the pausing factors. This is important because the differential effects between NELF subunits could simply reflect differences in knockdown efficiency.

While differences in RNAi efficiency could potentially contribute to variation in phenotype strength between individual knockdowns, multiple biological explanations could also account for the differing effects observed between NELF subunits, including differences in protein stability, residual complex activity, or subunit-specific functions. Importantly, the central conclusion of the manuscript does not depend on quantitative comparison of phenotype strength between individual NELF components, but rather on the observation that perturbation of multiple pausing-associated factors genetically interacts with Gro in vivo.

If RNAi validation is possible with existing reagents, this seems realistic within 3 months.

The manuscript focuses on the genetic interactions observed between Gro and pausing-associated factors in vivo rather than on quantitative comparison between individual RNAi lines. As no specific validation experiments were proposed, we are not currently planning additional RNAi validation analyses for the present study.

The discussion could be expanded, especially because the mechanism is not fully shown.Since the direct mechanism is still missing, the discussion could compensate. Right now, the proposed model is interesting, but it still leaves many open questions. For example:-Is Gro affecting the recruitment or activity of elongation factors?-Could looping or enhancer-promoter communication contribute?-How should the intronic Gro peaks be interpreted in the model?-In the wing, could the phenotype be discussed more mechanistically, in light of what is already known about Gro and derepression of vein-promoting genes?For example, a model figure could help here.

We thank the reviewer for these thoughtful suggestions.

Several of the points raised by the reviewer are discussed in the manuscript already. For example, we discuss the possibility that Gro influences the activity or recruitment of elongation-associated factors. We agree that enhancer-promoter communication and chromatin looping are a plausible component of this mechanism. As the Drosophila genome is compact and intronic enhancers are highly prevalent, topological looping provides a clear physical framework for how Gro molecules distributed at non-TSS sites regulate promoter-proximal machinery. Indeed, we have previously published this model (Kaul, Schuster, and Jennings, 2015; see Figure 1C; doi.10.1080/21541264.2014.1000709). Our current in vivo and genomic findings build directly upon this model, suggesting that within these established looped configurations, Gro acts locally to interface with and stabilize the pausing machinery.

With respect to the wing phenotypes, the Discussion focuses primarily on the interpretation of the observed genetic interactions between Gro and pausing-associated factors rather than on defining the precise downstream target genes contributing to vein phenotypes. We agree that additional mechanistic dissection of these developmental phenotypes would be interesting. However, this would require a substantial expansion of the study into the detailed developmental and signalling mechanisms underlying vein specification, which lies beyond the primary focus of the current manuscript.

OPTIONAL: It would be interesting to know whether the same peak distribution / functional logic is observed in mammalian TLE orthologs. This is not essential for the current conclusions, but it would broaden the impact.

Determining whether similar genomic distributions and functional relationships are conserved for mammalian TLE orthologues will be an important future project. However, relatively little comparable genome-wide TLE occupancy data are currently available, meaning that such analyses would require a substantial independent undertaking beyond the scope of the present study.

Minor comments -Please explain why promoters were defined as {plus minus}250 bp from the TSS. This seems rather narrow.

Promoters were defined as ±250 bp from annotated transcription start sites. This window size is commonly used in Drosophila genomic studies, where the compact organisation of the genome means that broader windows frequently overlap adjacent genes.

-Please clarify why S2R+ cells are included in the comparative part but are not followed in the same way in some downstream analyses.

S2R+ cells were included in the comparative analyses to determine which aspects of Gro recruitment were shared across multiple cell types and which were cell-type specific. Some downstream analyses focused on BG3 and Kc167 cells because these lines had the most extensive corresponding datasets available for the chromatin and pausing-factor analyses performed in the current study.

__Reviewer #3 __ Here Martínez Quiles and Jennings investigate the role of the Groucho repressor in BG3 cells. This extends a previous study that used S2R+ cells, published previously by one of the authors, as well as Kc167 cells. They find that Gro is recruited to gene promoters in a cell-type-specific manner. Gro associates with open chromatin, is mostly associated with enhancer regions, and is primarily excluded from regions of the genome that are repressed by Polycomb. After studying its function in cell culture, the authors investigate the role of Gro in a wing-specific background. The findings here are mostly correlative, showing that loss of Gro results in stronger phenotypic defects when combined with loss of factors including NELF-B or NELF-D, LARP7, and bin3. They propose that Gro acts to attenuate gene expression during early gene expression. This claim would be greatly strengthened if the authors provided RNA-seq data in addition to the ChIP-seq data shown in this manuscript, especially to examine gene expression patterns among the different cell lines studied here. At present, this is a correlative study that does not illuminate the mechanism of Gro in directly regulating promoter-proximal pausing or RNA polymerase behavior.

We thank the reviewer for this suggestion. However, extensive transcriptomic analyses of Drosophila cell lines, including Kc167, S2R+ and BG3-derived lines, have already been published (Cherbas et al., 2011), together with RNA-seq analyses following Gro depletion (Kaul et al., 2014). In addition, the association between Gro occupancy and paused genes has also been reported previously (Kaul et al., 2014; Chambers et al., 2017; doi. 10.1186/s12864-017-3589-6).

While additional RNA-seq analyses could further characterise transcriptional differences between cell lines, RNA-seq alone would not directly determine whether altered transcript levels arise specifically through changes in promoter-proximal pausing, as opposed to effects on transcription initiation, transcript stability, or indirect downstream regulatory effects. We therefore do not consider additional RNA-seq analyses necessary to support the central conclusions of the present study.

Figure 2-3: For the ChIP-seq data, scale the y-axis in the same manner to better understand enrichment between the samples.

These ChIP-seq datasets were generated independently using different antibodies and experimental conditions, direct comparison of enrichment magnitudes across datasets would not be biologically meaningful. Accordingly, our analyses focus on significant peak calls and overlap relationships rather than relative signal intensity. Applying identical y-axis scaling across all tracks would obscure significant enrichment in several datasets and could therefore be misleading.

RNA-seq data between different cell lines would greatly enhance the authors findings or Pro-Seq to really show a relationship with Gro binding and promoter proximal pausing.

We note that RNA-seq datasets for Gro depletion in Kc167 and S2R+ cells have already been published previously (Kaul et al., 2014), together with evidence linking Gro occupancy to paused genes (Kaul et al., 2014; Chambers et al., 2017). We therefore do not consider that additional RNA-seq analysis would substantially strengthen the central conclusions of the current manuscript.

Moreover, RNA-seq alone cannot distinguish if altered transcript abundance reflects changes in promoter-proximal pausing from other mechanisms influencing transcript abundance. While PRO-seq approaches could provide further mechanistic information regarding RNAPII dynamics, such experiments are beyond the scope of the present study.

This study helps to further clarify how Gro binds DNA in different cell types and indicates that may intersect with factors involved in promoter proximal pausing. The study is highly correlative and would require additional work to show a mechanistic link between Gro and transcription attenuation due to promoter proximal pausing.

While we agree that PRO-seq approaches could provide additional mechanistic information regarding RNAPII dynamics, establishing an appropriate experimental and analytical framework for these analyses would require a substantial extension beyond the scope of the present study. In addition, several aspects of the relationship between Gro occupancy, transcriptional repression, and promoter-proximal pausing that underpin these suggestions have already been addressed in previously published work, including RNA-seq analyses following Gro depletion (Kaul et al., 2014), evidence linking Gro occupancy with paused genes (Kaul et al., 2014; Chambers et al., 2017), and studies demonstrating that Gro-mediated repression does not occur through inhibition of pre-initiation complex assembly. The current manuscript is therefore intended to build upon these existing findings by integrating comparative genomic analyses with new in vivo genetic interaction data.

• *

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

Here Martínez Quiles and Jennings investigate the role of the Groucho repressor in BG3 cells. This extends a previous study that used S2R+ cells, published previously by one of the authors, as well as Kc167 cells. They find that Gro is recruited to gene promoters in a cell-type-specific manner. Gro associates with open chromatin, is mostly associated with enhancer regions, and is primarily excluded from regions of the genome that are repressed by Polycomb. After studying its function in cell culture, the authors investigate the role of Gro in a wing-specific background. The findings here are mostly correlative, showing that loss of Gro results in stronger phenotypic defects when combined with loss of factors including NELF-B or NELF-D, LARP7, and bin3. They propose that Gro acts to attenuate gene expression during early gene expression. This claim would be greatly strengthened if the authors provided RNA-seq data in addition to the ChIP-seq data shown in this manuscript, especially to examine gene expression patterns among the different cell lines studied here. At present, this is a correlative study that does not illuminate the mechanism of Gro in directly regulating promoter-proximal pausing or RNA polymerase behavior.

Major comments:

Figure 2-3: For the ChIP-seq data, scale the y-axis in the same manner to better understand enrichment between the samples.

The methods section is lacking details on how ChIP-seq was performed in the BG3 cell line. The methods section does a good job of indicating how the data were processed. Information on the antibodies and conditions used is critical, as is whether spike-in controls were used.

RNA-seq data between different cell lines would greatly enhance the authors findings or Pro-Seq to really show a relationship with Gro binding and promoter proximal pausing.

Significance

This study helps to further clarify how Gro binds DNA in different cell types and indicates that may intersect with factors involved in promoter proximal pausing. The study is highly correlative and would require additional work to show a mechanistic link between Gro and transcription attenuation due to promoter proximal pausing.

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 paper describes experiments designed to determine the mechanism of repression by the Groucho co-repressor in flies. The authors first characterize DNA binding by Groucho by ChIP-Seq analysis. This turns out to be consistent with recruitment driven by cell-type specific transcription factors. Nevertheless, its distribution across genomic features is similar across cell types, with enrichment in promoters and introns. It appears to bind in regions otherwise transcriptionally active (ie 'open chromatin'), rather than chromatin that is compacted and repressed. This suggest that Groucho regulates transcription through promoters or promoter-proximal pausing rather than by reducing chromatin accessibility. Groucho binding overlaps with NELF and GAF binding, seemingly consistent with a role in regulating pausing. However, Gro binding was also observed at promoters where P-TEFb components are detected, arguing against Gro repressing transcription P-TEFb exclusion from pausing sites. The authors next switched to investigating the consequences of Groucho kd and tested the idea that co-depletion of pausing factors might inform about the manner of gene repression, the idea being that if Groucho attenuates transcription by promoting or stabilizing promoter proximal pausing, then partial reduction of the pausing factors it affects should enhance the Groucho knock-down phenotype. Interestingly, simultaneous knock-down of Groucho and GAF resulted in enhanced patterning defects relative to Groucho knock-down alone, with the severity of the phenotypes resembling that observed upon increasing Groucho knock-down. Similarly, the knock-down of either Nelf-B or Nelf-D significantly enhanced Groucho phenotype. Finally, Kd of regulators of the pausing regulator CDK9 were tested. The 7SK snRNA complex inhibits CDK9, so any treatment leading to less 7SK will free CDK9 to positively affect pausing release. Larp kd fits that category as it directly leads to less 7SK and thus more CDK9 activity, while Bin3 kd results in less 5'-methyl capping, and thus more 7SK destabilization (less 7SK), again freeing CDK9 from inhibition - so, increasing pause release (like Nelf kd). Gratifyingly, this separate way of de-regulating/decreasing pausing again had an additive effect to Groucho depletion. Together, these genetic data thus overall support the idea that the (non-chromatin regulating) repressor Groucho works by stabilizing pausing complexes at specific genes.

Significance

I think this is nice little paper providing a simple, straightforward take-home message. It does not conceptually shake the world, and the evidence consists of (nice) correlations, with no direct proof put forward for the conclusions. I am not a Drosophila geneticist but probably rather an 'expert' on basic transcription mechanisms. I think the data in the paper are of high quality, if limited in scope, and that the conclusions are supported by the results, but I do not think the results or conclusions will have a big audience. Having said that, I found it interesting to learn about this group of repressors and their likely mode of action.

On the other hand, it is worth emphasizing that proteins such as NELF and CDK9 would arguably be expected to be found at very many genes, as promoter-proximal pausing does exist at a plethora of genes, also genes that are house-keeping genes, ie not regulated by cell type or stimuli. So, lots of genes with pausing are not regulated by modulation of pausing. So, basically, the fact that knockdown of the repressor Groucho and loss of pausing is additive does not in my opinion necessarily mean that Groucho works by stabilizing pausing. Although it is admittedly a reasonably assumption, Grouch could also work by repressing transcription initiation; the genetic outcomes of 'double relief' would be the same, ie higher transcription levels. I think a brief comment to this effect might be appropriate, especially in the absence of (difficult to obtain) direct evidence that the transcription initiation step is not affected by Groucho.

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 entitled "the co-repressor Groucho limits progression through the early transcription elongation checkpoint in vivo", the authors study how the co-repressor Groucho (Gro) may repress transcription in Drosophila. They combine Gro ChIP-seq analysis in BG3 cells with published data from Kc167 and S2R+ cells, chromatin-state and overlap analyses with pausing/elongation factors, and functionally link these interactions in vivo by genetic interaction assays in the wing. The manuscript shows that Gro recruitment is largely cell type-specific, while Gro binding is detected as discrete peaks with similar genomic distribution across cell types. Gro peaks are enriched in open enhancer-type chromatin and overlap with factors linked to promoter-proximal pausing. In vivo, knock-down (KD) of several pausing-related factors enhances the gro RNAi phenotype in the wing. Overall, this is a nice, clean study with interesting data, and the in vivo findings are clearly a strength. However, the mechanistic link between Gro and the early elongation checkpoint remains unclear, and several bioinformatics and presentation points could be strengthened.

Major comments

1. The main weakness is the lack of a mechanistic link between Gro and the early elongation checkpoint. This is really the main point for this reviewer. The manuscript builds an interesting model, and the data support a functional connection between Gro and pausing-related factors, but the mechanistic link is absent. At present, the paper relies on co-localisation of ChIP peaks and genetic interaction in vivo. This is interesting and supportive, but with several possible interpretations. The title and some parts of the text are thus a bit stronger than what is directly demonstrated. Two possibilities could be proposed: either tone down the mechanistic claim or strengthen it experimentally. A more direct assay of pause release or productive elongation after Gro depletion at endogenous targets would be highly valuable. For example, Gro-KD followed by Pol II Ser2-P ChIP, or promoter vs. gene body analysis on Gro-bound genes, ideally comparing genes with Gro at TSS vs. not-TSS, would greatly support the proposed model. If the assay is established, this seems feasible in about 4 months.
2. The bioinformatic part could be strengthened on "distinct TF repertoires" between cell types. The authors interpret the cell type-specific Gro recruitment as reflecting distinct transcription factor repertoires in BG3, Kc167 and S2R+ cells. This is interesting, but not really shown. To make this point more strongly, the author could provide a map of TF expression across different cell types, especially for the TFs corresponding to the enriched motifs they discuss. Otherwise, this remains speculative. In line, the manuscript discusses enriched motifs in BG3 and compares them to Kc167 and S2R+ cells, but this remains a bit descriptive. A clearer side-by-side comparison would strengthen the paper. This is particularly relevant to the motifs used in interpreting cell type-specific recruitment.
3. Several overlap analyses could be discussed more in depth. A few statements feel too strong for the actual percentages. For example, the GAF overlap in BG3 is around 51% genome-wide and 56% at TSS, which is meaningful, but not especially high. The text already states that it is not universal, and this point could be discussed more clearly. Similarly, for the UpSet plot, the wording around the "most frequent" combination could be toned down, because this is not a dominant pattern. More generally, I think the manuscript needs a clearer quantitative breakdown of TSS versus non-TSS peaks for the overlap analyses with NELF, GAF, Cdk9 and CycT. Several interpretations depend on this distinction, and right now, this is not always clear enough.
4. The "enhancer chromatin" interpretation is interesting, but not fully integrated with the genomic distribution. The observation that Gro is enriched in open enhancer-type chromatin is interesting and supports the idea that Gro does not act mainly through classical repressed chromatin. However, Gro peaks are also enriched at promoters and introns, and this reviewer feels that the manuscript does not fully connect these observations. Where are these enhancer-type peaks located exactly? Are they often intronic? Can this be correlated with the distribution of Gro peaks? This would help the reader and also strengthen the discussion because intronic Gro peaks are present in the data, but are not well integrated into the model.
5. The in vivo part is a strength, but some important points need clarification. The in vivo section is a clear highlight of the manuscript. It gives functional relevance to the model and moves the paper beyond cell-culture correlations. That said, a few points need to be clearer:
   • RNAi efficiency is not clear for the tested genes, especially the pausing factors. This is important because the differential effects between NELF subunits could simply reflect differences in knockdown efficiency.
   • The figures and text could lay out the logic of the genetic interactions for non-Drosophila readers. For example, the comparison of single and double copies of Gro-RNAi to combinatorial knockdowns, when it is additive, and when it is interpreted as synergistic.
   • The statistical analysis of the phenotype distributions should be shown more clearly (Fig. 5B). If RNAi validation is possible with existing reagents, this seems realistic within 3 months.
6. The discussion could be expanded, especially because the mechanism is not fully shown. Since the direct mechanism is still missing, the discussion could compensate. Right now, the proposed model is interesting, but it still leaves many open questions. For example:
   • Is Gro affecting the recruitment or activity of elongation factors?
   • Could looping or enhancer-promoter communication contribute?
   • How should the intronic Gro peaks be interpreted in the model?
   • In the wing, could the phenotype be discussed more mechanistically, in light of what is already known about Gro and derepression of vein-promoting genes? For example, a model figure could help here.

OPTIONAL:

It would be interesting to know whether the same peak distribution / functional logic is observed in mammalian TLE orthologs. This is not essential for the current conclusions, but it would broaden the impact.

Minor comments

• Please explain why promoters were defined as {plus minus}250 bp from the TSS. This seems rather narrow.
• Please clarify why S2R+ cells are included in the comparative part but are not followed in the same way in some downstream analyses.
• Figure 5 would gain clarity if the phenotype classes/panel letters were shown more clearly on the images.
• The legends of the wing figures should be expanded, especially for readers outside the Drosophila field.
• "in vivo" should be italicised consistently.

Referee cross-commenting

My main concerns are broadly echoed by Reviewer 2, notably regarding the need to clarify the level of mechanistic support for the proposed model. Reviewer 3 also raises related points about the correlative nature of the evidence. Overall, I think the reports converge on the need to better align the conclusions with the current data, while recognising the value of the functional in vivo results, though with different levels of requested additional analysis.

Significance

General assessment

This is a nice paper, with clean data and an interesting model. The strongest point is the attempt to connect the Gro genomic localisation with functional interaction in a developmental context. The observation that Gro is found in open enhancer-type chromatin, together with the in vivo genetic interactions, makes the study significant. The main limitation is that the mechanistic link is still missing. Overall, this reviewer finds the study convincing as a functional and descriptive paper but less convincing as a mechanistic one.

Advance

The study extends previous work on Gro by comparing several cell types and by adding in vivo genetic data in the wing. The main advance is thus conceptual and functional: it supports the idea that Gro acts in concert with the pausing/elongation machinery rather than simply through repressed chromatin. However, the mechanistic advance remains limited because a direct link to the early elongation checkpoint has not yet been demonstrated. This is the main thing preventing the paper from being stronger.

Audience

This reviewer feels that the manuscript will mainly interest a specialised basic research audience: scientists working on transcriptional regulation, co-repressors, RNA polymerase II pausing, chromatin regulation, and Drosophila developmental genetics. It can also be relevant to those broadly interested in Gro/TLE biology.

Expertise

This reviewer's expertise includes gene regulation and its nuclear organisation, transcriptional/co-transcriptional and post-transcriptional regulations, transcription factors biology, and Drosophila genetics. This reviewer is comfortable evaluating the developmental genetics, the conceptual aspect, and the interpretation of genomic analyses, but has less competence in evaluating bioinformatic ChIP-seq processing pipelines.

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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-2025-03227R

Corresponding author(s): Dr. David Skerrett-Byrne & Prof. Brett Nixon

1. General Statements

We are grateful to the reviewers and editorial team for their thoughtful and constructive evaluation of our manuscript. The comments provided were insightful and have substantially strengthened the rigor, clarity, and presentation of the study. In response, we have carefully revised the manuscript throughout, including clarification of conceptual interpretations, expansion of methodological detail, refinement and condensation of the Discussion, as well as addition of new supplementary analyses and figures. Collectively, we believe these revisions have improved both the transparency and accessibility of the work while reinforcing the central conclusions of the study.

At its core, this study sought to address a major unresolved question in reproductive biology: how spermatozoa, which are transcriptionally and translationally inert, achieve functional competence during post-testicular maturation. Using deep, stage-resolved phosphoproteomics integrated with functional validation approaches, we demonstrate that the majority of sperm phosphoproteomic remodelling occurs during epididymal maturation rather than during capacitation, challenging long-standing paradigms in the field. Beyond generating one of the deepest sperm phosphoproteomic resources currently available (>14,000 phosphosites), the study also provides functional and physiological context through kinase inhibition studies, in vivo knockout phenotypes, and the development of the ShinySpermPhospho online resource to facilitate community access and future discovery.

Importantly, through the review process we have worked carefully to ensure that the manuscript more clearly distinguishes data-driven conclusions from hypothesis-generating interpretations, particularly in areas relating to kinase prediction, metabolic regulation, and phosphoproteomic remodelling. We believe the revised manuscript now presents a more balanced and rigorous framework while preserving the significance of the central findings.

Overall, we hope the revised manuscript now provides a valuable resource and conceptual advance for the reproductive biology community, with implications extending from fundamental sperm cell biology to translational opportunities in male infertility and contraceptive development.

2. Point-by-point description of the revisions

REVIEWER #1

The manuscript by Skerrett-Byrne and collaborators represents a comprehensive and technically sophisticated phosphoproteomics study. Using high-resolution mass spectrometry on mouse sperm obtained from the caput and cauda regions of the epididymis, both before and after capacitation, the authors generated a more complete database of phosphorylation changes in these cells. One of the most interesting outcomes is that most of these changes occur during sperm maturation, rather than sperm capacitation. The work is important and relevant, and the information obtained could be valuable for reproductive biologists working in basic science, as well as for the identification of novel contraceptive targets.

__Answer: __We thank the reviewer for their positive assessment of our work and for recognising the value of the datasets we have generated for supporting future innovations in both fundamental reproductive biology and the identification of novel contraceptive targets. We are also delighted that the reviewer has recognised the significance that, contrary to previously thought, the majority of the phosphorylation changes we report occur during epididymal maturation, rather than subsequently during capacitation.

• The title should include a reference to sperm capacitation, as most of the study focuses on comparisons between epididymal maturation and capacitation, and the functional experiments are based on the latter. __Answer: __We thank the reviewer for this suggestion and have revised the title to reflect the importance of our focus on both phases of post-testicular sperm maturation, namely epididymal sperm maturation and sperm capacitation (please see line 1).

• Considering the newly reported changes in phosphosites, it would be desirable to include validation at the individual protein level for at least a few examples, using an independent technique such as western blotting. __Answer: __We thank the reviewer for this thoughtful suggestion and fully appreciate the motivation to seek orthogonal validation of phosphoproteomic findings. However, we respectfully wish to express our reservations regarding the use of antibody-based validation of site-specific phosphorylation events, a technique that is increasingly being recognised as problematic and, in many cases, less reliable than modern MS-based approaches (Nature, PMID: 39506148). Indeed, high-resolution mass spectrometry provides direct, site-resolved identification and quantification of phosphorylation events with substantially greater specificity, accuracy, and proteoform resolution than antibody-based methods. For this reason, MS-based phosphoproteomics is now widely regarded as the gold standard for mapping phosphorylation dynamics.

With regard to the use of antibodies, many commercially available phospho-specific antibodies lack sufficient site specificity, often have poorly defined or undocumented epitope recognition, and frequently fail to discriminate between closely related proteoforms or neighbouring phosphorylation sites. Indeed, recent large-scale evaluations have demonstrated that many widely used antibodies do not reliably bind their intended targets, raising concerns about reproducibility and interpretability across the biomedical sciences (PMID: 37995198). In one study, testing the utility of >600 antibodies, two thirds failed to work as described (PMID: 37995198), while the literature also features other studies (e.g. PMID: 31612854) reporting that certain antibodies (SC-138763) do not bind their stated target despite having been "used in 15 published manuscripts to ascribe specific properties to the protein in normal and disease states", collectively cited >3,000 times.

Accordingly, while we recognise the importance of independent validation, we contend that antibody-based validation may not be the most appropriate strategy to improve the robustness of the conclusions in this study. It is for this reason that we elected to strengthen confidence in our findings through multiple complementary approaches, including rigorous statistical filtering, extensive in-silico pathway and kinase analyses, selective pharmacological inhibition of target proteins, and in vivo functional interrogation using knockout mouse models. Together, these orthogonal strategies provide additional biological validation linking the reported phosphorylation changes to aspects of sperm function.

We have clarified this rationale in the revised manuscript and briefly expanded the discussion to touch on these methodological strengths and limitations (please see lines 754 - 759).

• In the knockout models, it is not possible to distinguish between defects in spermatogenesis and those arising during maturation or capacitation. A parameter directly related to spermatogenesis should therefore be included, for example, testicular weight or histology, sperm number, and sperm morphology. __Answer: __We thank the reviewer for raising this important point. We agree that systemic knockout models do not allow definitive discrimination between defects arising during spermatogenesis versus those occurring downstream during post-testicular sperm maturation or capacitation. Unfortunately, the additional parameters suggested by the reviewer do not form part of the standardised phenotyping pipeline implemented by the International Mouse Phenotyping Consortium (IMPC) and the European Mouse Mutant Archive (EMMA). As such, these data are not available for the knockout lines examined in this study and cannot be retrospectively generated. We have therefore clarified this limitation more explicitly in the revised manuscript and have framed the knockout data as physiological validation concerning the functional relevance of the parent protein rather than as definitive evidence of stage-specific or phosphorylation-dependent mechanisms of action. Importantly, the consistency of impaired sperm motility and fertilisation outcomes across multiple independent knockout lines supports the biological importance of the parent proteins identified, while acknowledging that the precise developmental window of their action remains to be resolved. While we regrettably concede that it is beyond the scope of this study, we do acknowledge that future studies will be required to dissect these mechanisms with greater resolution, ideally using germ cell-specific or temporally controlled knockout models, or targeted manipulation of key phosphoproteins and/or their phosphorylation motifs. Such approaches will be essential if we are to be able to distinguish roles of target proteins in spermatogenesis from those that occur downstream during epididymal maturation and capacitation (please see lines 725 - 733).

• Error values, sample size, and statistical analyses are missing from Figure 7 and should be provided for clarity. __Answer: __We apologise for this omission and have now updated Figure 7 and its legend to include sample sizes, error values, and details of the statistical analyses used, thereby improving clarity and reproducibility of these data.

In addition, we have clarified that sperm functional data derived from EMMA knockout lines are generated from cryopreserved samples comprising pooled cauda epididymal spermatozoa collected from 10 heterozygous males per line (PMID: 17709347, 38839949). As such, each data point represents a pooled biological sample, consistent with standardised EMMA/INFRAFRONTIER protocols (PMID: 25414328, 27262858, 38839949). Where appropriate, we have also included additional reproductive metrics at the level of IVF cycle (where available) and individual litters, including average litter size and fetal sex distribution (with exceptions for specific lines where such data are not available). These details are now captured in both the Methods and relevant figure legends (please see lines 453 - 457, 1228 - 1234, 1431 - 1434, 1533 - 1536, 1572 - 1575, Figure 7 & S6).

REVIEWER #2

In this manuscript, the authors examine dynamic modifications of the sperm phosphoproteome during epididymal transit and capacitation. They compare three distinct populations differing in anatomical localization and activation status: caput sperm, non capacitated cauda sperm, and capacitated cauda sperm. Using high resolution tandem mass spectrometry, they reveal that phosphorylation changes during epididymal passage are far more extensive than previously appreciated. These findings are further validated in genetically modified animal models, where disruption of selected genes encoding for phosphoproteins results in marked defects in sperm motility and fertilization capacity.

__Answer: __We thank the reviewer for their positive and thoughtful evaluation of our study and for recognising both the depth of the phosphoproteomic dataset and the importance of the functional validation experiments; sentiments that we whole heartly agree with.

• Throughout the text, and particularly in the paragraph entitled 'Epididymal maturation accounts for the majority of maturation associated sperm cell signaling,' it seems that phosphorylation is interpreted as inherently activatory and dephosphorylation as inhibitory (lines 248-252). Since this relationship is not universally applicable, it would be valuable to address this issue at the outset of the paragraph and to discuss how phosphorylation events are context dependent in their effects on protein function. __Answer: __We thank the reviewer for highlighting this important conceptual point. We fully agree that phosphorylation is not inherently activatory, nor is dephosphorylation necessarily inhibitory, and that the functional consequences of phosphorylation are highly context dependent. We have revised the indicated paragraph to explicitly acknowledge this at the outset to ensure that phosphorylation changes are interpreted as regulatory rather than intrinsically directional (please see lines 219 - 222).

• Lines 392-393: the claim that "the introduction of each inhibitor to populations of capacitating spermatozoa led to a significant reduction..." is not fully supported by data and should be toned down. In fact, two out of three inhibitors, do not significantly affect the acrosome reaction. __Answer: __We thank the reviewer for this careful assessment and agree that the original wording overstated the nuances of the effects of individual inhibitors. We have revised the text to explicitly report the corresponding p-values and to distinguish between statistically significant and non-significant trends. Specifically, inhibition of PAK1 produced a statistically significant reduction in the acrosome reaction, whereas inhibition of STK33 (p = 0.0574) and HIPK4 (p = 0.0911) resulted in consistent, but non-significant, reductions. Importantly, combined inhibition of all three kinases yielded a robust and statistically significant suppression of acrosomal exocytosis. The revised wording now accurately reflects the quantitative data (please see lines 419 - 424, 700 - 703).

• The discussion section, spanning 11 pages, is overly long and contains considerable repetition. I recommend transferring the detailed description of experiments to the 'Results' section and using the discussion primarily to synthesize and highlight the novel findings while limiting speculative content. For example, the content in lines 509-530 could be condensed and relocated to the Results. Likewise, other detailed examples would be more appropriately presented within their respective result paragraphs. __Answer: __We thank the reviewer for this constructive feedback. We agree that the Discussion was overly long and on reflection does contain some unnecessary repetition. In response, we have substantially condensed the Discussion (shorten by 641 words), relocated and shorten detailed descriptions of experimental observations to the Results section where appropriate (including the suggestion made), and focused the revised Discussion on synthesis of the key findings and their broader implications. We should note, to address certain review comments, this require further additions to the discussion but we have endeavour to keep this brief (please see lines 309 - 326 and throughout the discussion).

• Minor points:

• To improve reproducibility, the suppliers of all reagents should be specified together with their catalogue numbers
• Figure 7: it is unclear which data are statistically significant
• Figure 7B: fertilization capacity should be assessed at an earlier stage, as the cleavage rate to 2-cell embryo may be affected by factors unrelated to the sperm ability to fertilize

__Answer: __We thank the reviewer for these suggestions. We have now added supplier information and catalogue numbers for all reagents to the Methods section to improve reproducibility (please see lines 1256 - 1257, 1295, 1299, 1310, 1333 - 1334, 1343 - 1344, 1388 - 1389, 1401, 1406). We have revised Figure 7 and its legend to clearly indicate statistically significant differences, including sample sizes and statistical tests used. Lastly, we agree that assessment at earlier fertilisation stages would complement our featured assessment of sperm fertilisation competence. Regrettably, all IVF data were generated via standardised and unbiased IMPC/EMMA pipelines. As such, cleavage rate to the 2-cell stage represents the earliest uniformly available endpoint across all knockout lines. We have clarified this limitation in the revised manuscript (please see lines 723 - 724).

REVIEWER #3

This is technically sophisticated phosphoproteomic study of mouse sperm maturation across the epididymis and during capacitation. The dataset is deep (>14,000 phosphosites) and the analyses integrate high-resolution MS, immunofluorescence, IPA, kinase mapping, pharmacological inhibition, and knockout mouse models. The manuscript represents a nice resource for the field. However, several issues limit clarity, mechanistic interpretation, and robustness of the conclusions. In particular, the manuscript's scope is extremely large, making some conclusions insufficiently supported, and some analyses require better control, methodological transparency, or deeper mechanistic connection. It gives the impression that some mechanistic data was added to descriptive data in order to increase the manuscript's impact, although the current mechanistic data is not convincing.

Major concerns

1. Conceptual Overreach - "Epididymal maturation accounts for 86% of phosphorylation changes" The manuscript repeatedly emphasizes that epididymal maturation causes the majority of phosphoproteomic remodeling. While the data indeed show large quantitative differences, several conceptual issues remain:

2. • The caput vs. cauda comparison includes differences in protein abundance, not only phosphorylation*
3. • Many phosphosites lost in the cauda may reflect protein loss, not dephosphorylation (the authors acknowledge this, but quantitative controls are insufficient)*
4. • The normalization method for phosphopeptide abundance vs total protein abundance is needed*
5. • It is unclear whether the same amount of starting material and equal protein loading were used across stages I would suggest to perform (or explicitly describe) normalization using matched proteome intensities. Provide supplementary plots showing phosphosite/parent-protein normalization to avoid overinterpreting phosphosite loss as dephosphorylation*

__Answer: __We thank the reviewer for this important and constructive critique and agree that interpretation of phosphoproteomic changes during epididymal maturation must carefully consider concurrent remodelling of the underlying sperm proteome.

To directly address the concern that phosphosites lost in the cauda may reflect protein loss, not dephosphorylation, we have now explicitly compared these phosphoproteins lost during caput-to-cauda transit with proteins shown to be lost or reduced over the same maturation window in a previously published matched proteomic analysis of the same sperm populations. This comparison revealed that 527 phosphoproteins, out of a total of 966 phosphoproteins lost, overlapped with proteins lost during epididymal maturation, while a further 88 phosphoproteins aligning with proteins exhibiting reduced abundance during transit. While these data indicate that a subset of phosphosite loss can be attributed to complete loss of the parent protein, the remaining phosphoproteomic changes (45.4%) cannot be fully explained by protein disappearance alone and are therefore consistent with extensive phosphoproteomic remodelling. We have documented this information in a new panel of Supplementary Figure 1 (Figure S1B) and the corresponding text has been revised accordingly (please see lines 182 - 188).

With respect to normalisation strategies, we respectfully note that normalisation of phosphopeptide intensities to total protein abundance is not universally accepted in large-scale phosphoproteomic analyses (PMID: 30190555, 34857927, 38576152), particularly in systems undergoing extensive proteome remodelling such as maturing spermatozoa. In many contexts, including our own previous work, phosphoproteomic analyses are performed on equal protein input and interpreted at the level of phosphopeptide abundance, with functional relevance established through orthogonal biological validation rather than ratio-based correction to total protein levels.

Lastly, all samples in this study were diluted to equal total protein amounts prior to phosphopeptide enrichment, ensuring consistent input material across all sperm populations (originally noted in the manuscript, please see line 1339). We have now clarified this explicitly in the Results section to ensure this is not missed (please see lines 146 - 147). Moreover, our conclusions are supported by independent in-silico analyses, pharmacological inhibition studies, and in vivo knockout models, collectively providing functional validation that extends beyond phosphosite quantification alone.

Finally, to address concerns regarding potential conceptual overreach, we have revised the language surrounding the statement that epididymal maturation accounts for ~86% of phosphorylation changes to ensure precise interpretation. Specifically, we have clarified that this value refers to the proportion of statistically significant differences in phosphopeptide abundance detected across maturation stages, to avoid implying direct measurement of net enzymatic dephosphorylation (please see lines 519 - 520).

Importantly, having addressed the reviewer's concerns detailed above, we believe the data do support the conclusion that the majority of sperm phosphoproteomic remodelling occurs during epididymal maturation rather than during capacitation. While we have tempered our language to improve clarity, the central quantitative observation that epididymal transit represents the dominant phase of phosphoproteomic remodelling remains supported by the revised analyses.

• Capacitation analysis is underpowered and oversimplified*

The authors state that capacitation leads to "modest" changes. However:

• • The capacitation protocol uses dibutyryl-cAMP + pentoxifylline, which may bypass early physiological signaling. This is a important red flag __Answer: __We thank the reviewer for this important point and agree that the choice of capacitation conditions influences the nature and magnitude of signalling events detected. The use of dibutyryl cAMP and pentoxifylline represents a well-established and widely adopted experimental model to induce robust and synchronised capacitation-associated signalling in mouse spermatozoa, acting specifically via the activation of the canonical cAMP/PKA signalling axis (PMID: 36384108, 22458710, 16221991). While we acknowledge that this approach bypasses some upstream physiological signalling events that initiate capacitation during sperm transit of the female reproductive tract, it is intentionally employed to provide a reproducible capacitation stimulus, specifically enabling us to discriminate phosphorylation changes associated with the attainment of sperm fertilization competence. This strategy also directly addresses a limitation of working with mouse spermatozoa in which these cells rapidly succumb to cell senescence/death within a matter of ~1-3 hours in an in vitro* setting. In our previous studies, we have noted that this time period is insufficient to achieve high levels of capacitation among populations of mouse spermatozoa, unless pharmacological agents (i.e. dibutyryl-cAMP + pentoxifylline) are supplemented to accelerate capacitation (PMID: 15252132). This is now a widely accepted paradigm in the field and one that enables us to deliver on our stated objective of assessing the phosphorylation status of fertilization competent spermatozoa, as opposed to those that are captured during early phases of the capacitation cascade.

Importantly, our conclusion that capacitation is associated with comparatively fewer phosphoproteomic changes is based on direct quantitative comparison with epididymal maturation under identical analytical conditions, and is not intended to downplay the biological importance of this critical maturation event. Even under the capacitation-inducing conditions employed herein, the scale of phosphoproteomic remodelling observed was substantially smaller than that occurring during epididymal transit, underscoring the influence of epididymal maturation over the status of the sperm phosphoproteome.

To address this concern, we have revised the manuscript to clarify that the capacitation-associated phosphoproteomic changes reported here are specific to the experimental model used and likely represent a conservative estimate of signalling complexity under physiological conditions. We have also tempered language implying generalisation beyond this context (please see lines 329 - 332, 488 - 491, 673 - 677).

• *

• Kinase prediction and functional validation require more rigor*

The identification of 343 kinases that may regulate phospho-changes is extremely broad. Issues:

• • The kinase-substrate assignments rely heavily on in silico predictions (IPA, PhosphoSitePlus), which often contain non-sperm data. __Answer: __We thank the reviewer for this important observation and fully agree that kinase-substrate assignments inferred from in-silico* resources such as IPA and PhosphoSitePlus are largely derived from non-sperm systems and therefore must be interpreted cautiously.

Importantly, this limitation reflects a broader and well-recognised gap in the field; regrettably comprehensive, experimentally validated kinase-substrate networks do not currently exist for mammalian spermatozoa on this scale, particularly in the context of epididymal maturation and capacitation. The primary objective of the present study was therefore not to define definitive kinase-substrate relationships, but to generate a high-depth, sperm-specific phosphoproteomic resource that can serve as a foundation for hypothesis generation and future mechanistic interrogation.

Accordingly, in-silico kinase prediction tools were employed to contextualise the phosphoproteomic data and to prioritise candidate kinases for functional testing, rather than to assert sperm-specific kinase-substrate specificity. We have revised the manuscript to clarify that these predictions represent informed starting points in a system where such information is currently lacking, and that functional relevance was subsequently assessed using complementary pharmacological and genetic approaches (please see lines 383 - 388, 682 - 686).

By providing a deep, stage-resolved phosphoproteomic dataset encompassing more than 14,000 phosphosites, this study establishes a much-needed reference framework for the reproductive biology community, enabling future targeted validation of kinase-substrate relationships in sperm. We believe this resource-based contribution represents a major strength of the work and addresses a critical knowledge gap in the field.

• *

• • Please explain the rationale by which, from 343 candidate kinases, 3 (STK33, HIPK4, PAK1) are selected.*
• • The pharmacological inhibitors used have off-target effects (ML281 inhibits multiple CMGC kinases; Foretinib inhibits MET/VEGFR; NVS-PAK1-1 inhibits PAK1/2/3).*
• • No control experiments are included to confirm kinase inhibition in sperm (e.g., phosphosite-specific Western blots)* __Answer: __We split this comment from the above, to best address this important critique. We agree that kinase-substrate relationships inferred from phosphoproteomic data must be interpreted with caution. The identification of 343 kinases in this study was not intended to represent a definitive catalogue of all sperm-specific kinase-substrate interactions, but rather to provide insights into kinases that potentially contribute to phosphoproteomic remodelling of mouse spermatozoa during the different phases of their post-testicular maturation. These kinases were identified through integration of multiple complementary approaches, including direct detection within the phosphoproteome, upstream regulator prediction using IPA, curated kinase-substrate databases, and comparison with previously published epididymal sperm proteomes.

From this broader resource, we deliberately restricted functional interrogation to a small subset of kinases putatively associated with capacitation-induced phosphoproteomic changes. STK33, HIPK4, and PAK1 were selected based on their predicted association with capacitation-specific phosphorylation events, representation across distinct kinase families, lack of prior functional characterisation in terms of either sperm maturation or function, and availability of well-characterised pharmacological inhibitors suitable for functional perturbation. We fully acknowledge that the inhibitors employed are not absolutely kinase-specific and may exhibit off-target effects. Accordingly, we have revised the manuscript to clarify that these experiments are intended to test functional dependence on kinase activity rather than to establish direct kinase-substrate relationships. The observation that combined inhibition of three mechanistically distinct kinases produced a robust and additive suppression of the acrosome reaction supports the conclusion that kinase activity is required for this process, while avoiding overinterpretation of individual kinase specificity.

We have revised the language throughout the manuscript to more clearly reflect these limitations and to frame the kinase inhibition experiments as functional validation of phosphoproteomic predictions rather than definitive mechanistic proof (please see lines 383 - 388, 405 - 406, 682 - 686, 705 - 710).

• *

• *

• The knockout-mouse validation section is underdeveloped*

The linkage of KO phenotypes to phosphorylation changes is potentially powerful but currently weak.

Issues:

• • Most KOs are systemic deletions, not sperm-specific; phenotypes could stem from developmental defects.*
• • Some proteins validated (e.g., ACO2, CMPK1) regulate core metabolism; their phenotypes may not reflect phosphoregulation but loss of essential protein function.*
• • No evidence is provided that the KO affects the specific phosphosites detected in the MS dataset.* __Answer: __We thank the reviewer for this important clarification. We agree that since the knockout models employed represent systemic deletions, they cannot directly resolve sperm-specific or phosphosite-specific mechanisms, and it was not our intention to suggest otherwise. We have revised the manuscript to explicitly frame the knockout phenotypes as evidence of physiological relevance of the identified phosphoproteins, rather than as direct validation of individual phosphorylation events (please see lines 723 - 734).

We further clarify that for proteins with central metabolic roles, the observed phenotypes likely reflect loss of essential protein function rather than isolated disruption of phosphoregulation. Accordingly, we have tempered our language and emphasise that these data support functional importance while highlighting the need for future studies employing germ cell-specific or phosphosite-targeted models (please see discussion).

• *

• Immunofluorescence and Western blots need improved quantification*

Figures showing PKA substrate, pY, pT, pS changes are visually compelling but lack:

• • quantification across biological replicates*
• • explanation of antibody specificity (e.g., pan-PKA sites include RRXS/T motifs; cross-reactivity possible). __Answer: __We thank the reviewer for this comment and appreciate the emphasis on rigor in antibody-based analyses. We would like to clarify that the immunofluorescence and immunoblot data presented in this study do include densitometric based quantification taking into account data generated from three independent biological replicates. This is indicated by the inclusion of error bars and as stated in the relevant figure captions. (please see lines 1164 - 1168 "All immunoblotting experiments were repeated with at least three biological replicates. Densitometric data normalization was performed against the loading control protein GAPDH, and each value subsequently expressed as a fold change relative to the caput sperm. Data were analyzed by one way ANOVA with GraphPad Prism.).*

With respect to antibody specificity, we fully agree that phospho-specific and motif-based antibodies have inherent limitations, including epitope ambiguity and the inability to resolve site-specific phosphorylation with amino acid precision (please see our answer to Comment #2 from Reviewer #1 above). For this reason, the antibodies employed here represent well-established, widely used markers in sperm biology and were included to illustrate global phosphorylation trends rather than to validate individual phosphosites. Importantly, quantitative and site-resolved interpretation of phosphorylation dynamics throughout the manuscript is derived from mass spectrometry based phosphoproteomics, which provides substantially greater specificity and resolution than antibody-based approaches.

• *

• Many interpretations about metabolism, storage, oxidative stress, and quiescence are speculative*

The discussion provides attractive models linking phosphorylation to:

• • suppression of glycolysis*
• • quiescent metabolic state in cauda epididymis*
• • activation of antioxidant pathways*
• • UPR and proteostasis modifications However, no direct functional evidence is provided for any of these pathways.*

__Answer: __We thank the reviewer for this thoughtful observation and agree that several interpretations linking phosphorylation changes to metabolic regulation, oxidative stress, proteostasis, and cellular quiescence are necessarily inferential in the absence of direct functional assays. With this in mind, we have revised the Discussion to more clearly distinguish data-driven observations from hypothesis-generating interpretations and have tempered language accordingly. These models are now explicitly framed as conceptual frameworks arising from large-scale phosphoproteomic analysis, intended to guide future targeted investigation rather than to assert definitive mechanistic conclusions (please see lines 588, 590, 596 - 599, 609, 613, 621 - 625 ).

Given the breadth and depth of the dataset, only a limited number of functional pathways could be explored experimentally within the scope of the current study. We anticipate that the phosphoproteomic resource generated here, supported by the accompanying ShinySpermPhospho application, will enable the wider community to interrogate additional pathways and to design focused mechanistic studies building on these findings.

• Acrosome reaction evaluation*

I have encountered significant deficiencies in this approach. On one side, testing the effect of a single dose of inhibitors on a specific readout is too preliminary, as stated above. In addition, and due to the presence of possible off target effects, more than one inhibitor is expected to be tested, or a direct biochemical assay to confirm at least targeted action. Even KO models, as proposed for other proteins in Figure 7.

Acrosome reaction values are expected to be presented, as regularly done, by indicating acrosome reacted percentages, without normalizations that complicate understanding. In addition, consider Pg as a more physiological stimulus instead of A23187 for triggering AR.

__Answer: __We thank the reviewer for these constructive comments and agree that careful assessment of inhibitor effects on sperm viability and motility is essential. We would like to clarify, in case this was overlooked, that these controls were performed and are presented in Supplementary Figure S4. Specifically, sperm were exposed to four concentrations of each inhibitor and assessed over time (0 and 60 minutes) for viability, total motility, and progressive motility. Across all concentrations, time points, and treatment conditions, including combined inhibitor treatments, no significant reductions in sperm viability or motility parameters were observed. These data support the conclusion that the effects on acrosome reaction are not secondary to general sperm toxicity.

With respect to data presentation, acrosome reaction values were expressed relative to matched capacitated vehicle-treated controls to account for biological variability in absolute acrosome reaction rates observed between independent sperm preparations and experimental days. This normalisation strategy was used to facilitate direct comparison between treatment groups, and respectfully, we have elected to retain this presentation format in Figure 6. Nonetheless, in the interest of transparency, we have now included the raw acrosome reaction values/ranges in the supplementary material (Table S6) and have provide these as a figure to the reviewer for reference.

We agree that the use of progesterone represents a more physiological stimulus for inducing the acrosome reaction. However, there is no single universally accepted approach for acrosome reaction induction, and calcium ionophore-based assays remain widely used to assess the capacity for acrosomal exocytosis under defined experimental conditions. In the present study, this approach was selected to provide a robust and reproducible functional readout suitable for comparative analysis.

We have revised the manuscript to more clearly describe the dose-response and viability control experiments, to acknowledge potential off-target effects of kinase inhibitors, and framed the acrosome reaction assays as functional screening experiments rather than definitive mechanistic dissection (please see lines 415, 705 - 710, Table S7).

• *

• Figure legend to Figure 7. How many oocytes, how many replicates were performed. How many transfers. Please add important data to the legend.*

__Answer: __We thank the reviewer for highlighting this omission. In line with comment 4 from Reviewer #1 (please see above), we have revised the Figure 7 legend and Methods section to provide detailed information regarding sample size and experimental design, including the number of oocytes used per IVF experiment performed and the number of biological replicates.

Specifically, IVF and oocyte isolation procedures were conducted according to standardised INFRAFRONTIER protocols (PMID: 25414328, 27262858, 38839949). Across knockout lines, IVF experiments were performed over 1 - 6 independent cycles per line, with an average of 24.2 oocytes used per cycle. To provide transparency regarding this variability, we have included a new supplementary figure (Figure S6) summarising the average number of oocytes used per IVF cycle alongside the corresponding cleavage rates (%CR).

Sperm samples used in these assays were derived from cryopreserved cauda epididymal spermatozoa pooled from 10 heterozygous males per knockout line, as per EMMA guidelines (PMID: 17709347). Additionally , where available, we have incorporated reproductive outcome measures at the level of individual litters (e.g. average pup number and sex distribution) to provide further biological context. These additions improve transparency and ensure that the experimental design and data interpretation are clearly defined (please see lines 453 - 457, 1228 - 1234, 1431 - 1444, 1533 - 1536, 1572 - 1575, Figure 7 & S6).

• *

This is technically sophisticated phosphoproteomic study of mouse sperm maturation across the epididymis and during capacitation. The dataset is deep (>14,000 phosphosites) and the analyses integrate high-resolution MS, immunofluorescence, IPA, kinase mapping, pharmacological inhibition, and knockout mouse models.

__Answer: __We thank the reviewer for this positive assessment and for recognising the technical sophistication and integrative nature of the study. We are also grateful for the constructive feedback provided, which has helped us to substantially strengthen the clarity, rigor, and presentation of the 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 #3

Evidence, reproducibility and clarity

This is technically sophisticated phosphoproteomic study of mouse sperm maturation across the epididymis and during capacitation. The dataset is deep (>14,000 phosphosites) and the analyses integrate high-resolution MS, immunofluorescence, IPA, kinase mapping, pharmacological inhibition, and knockout mouse models. The manuscript represents a nice resource for the field. However, several issues limit clarity, mechanistic interpretation, and robustness of the conclusions. In particular, the manuscript's scope is extremely large, making some conclusions insufficiently supported, and some analyses require better control, methodological transparency, or deeper mechanistic connection. It gives the impression that some mechanistic data was added to descriptive data in order to increase the manuscript's impact, although the current mechanistic data is not convincing.

Major concerns

1. Conceptual Overreach - "Epididymal maturation accounts for 86% of phosphorylation changes" The manuscript repeatedly emphasizes that epididymal maturation causes the majority of phosphoproteomic remodeling. While the data indeed show large quantitative differences, several conceptual issues remain:
   • The caput vs. cauda comparison includes differences in protein abundance, not only phosphorylation.
   • Many phosphosites lost in the cauda may reflect protein loss, not dephosphorylation (the authors acknowledge this, but quantitative controls are insufficient).
   • The normalization method for phosphopeptide abundance vs total protein abundance is needed.
   • It is unclear whether the same amount of starting material and equal protein loading were used across stages.

I wwould suggest to perform (or explicitly describe) normalization using matched proteome intensities. Provide supplementary plots showing phosphosite/parent-protein normalization to avoid overinterpreting phosphosite loss as dephosphorylation. 2. Capacitation analysis is underpowered and oversimplified The authors state that capacitation leads to "modest" changes. However: - The capacitation protocol uses dibutyryl-cAMP + pentoxifylline, which may bypass early physiological signaling. This is a important red flag 3. Kinase prediction and functional validation require more rigor The identification of 343 kinases that may regulate phospho-changes is extremely broad. Issues: - The kinase-substrate assignments rely heavily on in silico predictions (IPA, PhosphoSitePlus), which often contain non-sperm data. - Please explain the rationale by which, from 343 candidate kinases, 3 (STK33, HIPK4, PAK1) are selected. - The pharmacological inhibitors used have off-target effects (ML281 inhibits multiple CMGC kinases; Foretinib inhibits MET/VEGFR; NVS-PAK1-1 inhibits PAK1/2/3). - No control experiments are included to confirm kinase inhibition in sperm (e.g., phosphosite-specific Western blots). 4. The knockout-mouse validation section is underdeveloped The linkage of KO phenotypes to phosphorylation changes is potentially powerful but currently weak. Issues: - Most KOs are systemic deletions, not sperm-specific; phenotypes could stem from developmental defects. - Some proteins validated (e.g., ACO2, CMPK1) regulate core metabolism; their phenotypes may not reflect phosphoregulation but loss of essential protein function. - No evidence is provided that the KO affects the specific phosphosites detected in the MS dataset. 5. Immunofluorescence and Western blots need improved quantification Figures showing PKA substrate, pY, pT, pS changes are visually compelling but lack: - quantification across biological replicates - explanation of antibody specificity (e.g., pan-PKA sites include RRXS/T motifs; cross-reactivity possible). 6. Many interpretations about metabolism, storage, oxidative stress, and quiescence are speculative The discussion provides attractive models linking phosphorylation to: - suppression of glycolysis - quiescent metabolic state in cauda epididymis - activation of antioxidant pathways - UPR and proteostasis modifications However, no direct functional evidence is provided for any of these pathways. 7. Acrosome reaction evaluation I have encountered significant deficiencies in this approach. On one side, testing the effect of a single dose of inhibitors on a specific readout is too preliminary, as stated above. In addition, and due to the presence of possible off target effects, more than one inhibitor is expected to be tested, or a direct biochemical assay to confirm at least targeted action. Even KO models, as proposed for other proteins in Figure 7. Acrosome reaction values are expected to be presented, as regularly done, by indicating acrosome reacted percentages, without normalizations that complicate understanding. In addition, consider Pg as a more physiological stimulus instead of A23187 for triggering AR. 8. Figure legend to Figure 7. How many oocytes, how many replicates were performed. How many transfers. Please add important data to the legend.

Significance

This is technically sophisticated phosphoproteomic study of mouse sperm maturation across the epididymis and during capacitation. The dataset is deep (>14,000 phosphosites) and the analyses integrate high-resolution MS, immunofluorescence, IPA, kinase mapping, pharmacological inhibition, and knockout mouse models.

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 examine dynamic modifications of the sperm phosphoproteome during epididymal transit and capacitation. They compare three distinct populations differing in anatomical localization and activation status: caput sperm, non‑capacitated cauda sperm, and capacitated cauda sperm. Using high‑resolution tandem mass spectrometry, they reveal that phosphorylation changes during epididymal passage are far more extensive than previously appreciated. These findings are further validated in genetically modified animal models, where disruption of selected genes encoding for phosphoproteins results in marked defects in sperm motility and fertilization capacity.

Major points:

Throughout the text, and particularly in the paragraph entitled 'Epididymal maturation accounts for the majority of maturation‑associated sperm cell signaling,' it seems that phosphorylation is interpreted as inherently activatory and dephosphorylation as inhibitory (lines 248-252). Since this relationship is not universally applicable, it would be valuable to address this issue at the outset of the paragraph and to discuss how phosphorylation events are context‑dependent in their effects on protein function.

Lines 392-393: the claim that "the introduction of each inhibitor to populations of capacitating spermatozoa led to a significant reduction..." is not fully supported by data and should be toned down. In fact, two out of three inhibitors, do not significantly affect the acrosome reaction.

The discussion section, spanning 11 pages, is overly long and contains considerable repetition. I recommend transferring the detailed description of experiments to the 'Results' section and using the discussion primarily to synthesize and highlight the novel findings while limiting speculative content. For example, the content in lines 509-530 could be condensed and relocated to the Results. Likewise, other detailed examples would be more appropriately presented within their respective result paragraphs.

Minor points:

• To improve reproducibility, the suppliers of all reagents should be specified together with their catalogue numbers.
• Figure 7: it is unclear which data are statistically significant
• Figure 7B: fertilization capacity should be assessed at an earlier stage, as the cleavage rate to 2-cell embryo may be affected by factors unrelated to the sperm ability to fertilize

Significance

A key novel finding of this work is that extensive changes in the sperm phosphoproteome occur during epididymal maturation, whereas capacitation is associated with comparatively modest modifications. This research provides a finely resolved description of phosphorylation events associated with the signaling pathways underlying functional sperm maturation. The methodological innovation -high‑resolution MS‑based phosphoproteomics- unlocks a level of detail and comprehensiveness in phosphorylation analysis that was previously unattainable. Moreover, the identification of previously unrecognized phosphoproteins in sperm cells, together with the development of a dedicated application hosting the complete dataset, represents a valuable resource for researchers in reproductive biology and particularly for experts in sperm development and maturation.

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 Skerrett-Byrne and collaborators represents a comprehensive and technically sophisticated phosphoproteomics study. Using high-resolution mass spectrometry on mouse sperm obtained from the caput and cauda regions of the epididymis, both before and after capacitation, the authors generated a more complete database of phosphorylation changes in these cells. One of the most interesting outcomes is that most of these changes occur during sperm maturation, rather than sperm capacitation. The work is important and relevant, and the information obtained could be valuable for reproductive biologists working in basic science, as well as for the identification of novel contraceptive targets.

Minor comments:

1) The title should include a reference to sperm capacitation, as most of the study focuses on comparisons between epididymal maturation and capacitation, and the functional experiments are based on the latter.

2) Considering the newly reported changes in phosphosites, it would be desirable to include validation at the individual protein level for at least a few examples, using an independent technique such as western blotting.

3) In the knockout models, it is not possible to distinguish between defects in spermatogenesis and those arising during maturation or capacitation. A parameter directly related to spermatogenesis should therefore be included, for example, testicular weight or histology, sperm number, and sperm morphology.

4) Error values, sample size, and statistical analyses are missing from Figure 7 and should be provided for clarity.

Significance

This study shows by high-resolution phosphoproteomics that most phosphorylation changes occur during epididymal transit rather than capacitation, challenging long-standing assumptions. The integration of the new datasets with functional validation of key kinases and knockout models strengthens the study; however, the work lacks single-protein validation of phosphorylation events, and the use of systemic knockouts does not allow confirmation of sperm-specific effects. The open ShinySpermPhospho dataset will be worthwhile to a broad audience of reproductive biologists and cell signaling specialists, and may be of value for future studies on male fertility and the development of novel male contraceptives.

Field of expertise: reproductive physiology, sperm biology and capacitation, gamete interaction.

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 and we are glad that they acknowledge this work to be a timely contribution to a quickly moving field and a valuable tool to generate testable hypothesis. We are pleased that reviewer #2 highlights that “a major strength is the combination of orthogonal evidence types” and that the tool serves to generate novel hypothesis. The revised manuscript will sharpen the positioning of the study within this context. Additional experimental evidence will be provided to address the points raised by reviewers #1 and #3.

Reviewer #1* 1.The authors do not co-IP ARF1. This does not surprise me as small GTPases often hydrolyse their GTP during lysis. *

We agree that this is likely due to transient association and GTP hydrolysis during lysis and will add a section to the manuscript.

There have been a number of ARF1 bioID screens done- have the authors checked if their complex has turned up here?

We will include this in the revised manuscript.

1. I am a bit confused by some of the interpretation about KO and loss of JTB staining. They interpret: "The SYS1 acts as a Golgi recruitment factor for both ARFRP1 and JTB". The ARFRP1 has been published and is a cytosolic protein, so that makes sense. However, the JTB is not cytosolic by a membrane protein, so cannot be "recruited". Now maybe it is retained in the Golgi by this interaction, but if that is the case you would still expect signal on another organelle or the plasma membrane (and we see it isnt degraded in the lysosome due to the western blot). I am confused by the authors model here.

We will clarify the phrasing and will provide a clearer interpretation, also considering the other improved imaging experiments that will be included in the revised manuscript.

4.The authors validate their JTB antibody and confirm the fact that there are not reduced SYS1 levels in the JTBKO- this is very clear (albeit unquantified). What I do not see validated is the SYS1KO. I think this is quite important.

We will validate SYS1 KO using TIDE and/or western blotting.

5.The colocalisation in panel 3D is weak and unclear to me. It is not quantified. It is not clear if there have been 3 repeats.

The revised manuscript will include improved imaging data. We will repeat relevant experiments, include appropriate controls and quantify where necessary.

6.The imaging in figure 3 is not clear in places, and it stands out in a very clear manuscript. I cannot see the JTB in panel F. There are no scale bars. The dynamic range of the image is not utalised. I do not see the stain in the JTB in either of the sys1 KO, i do not see the SYS1-FLAG staining in the complement, and it is not quantified at all. It may all seem trivial, but (to me) this is an absolutely critical bit of biology data to support the informatics.

The revised manuscript will include improved imaging data. We will repeat relevant experiments, include appropriate controls and quantify where necessary.

7.I am a bit unconvinced by the interpretation of it being a retrograde trafficking complex. This is for 2 key reasons- 1) the VSV-G is antrograde (despite unusually they interpret a "severe defect in retrograde transport"). 2) Even if it was only having an effect in the retrograde direction I would still remain a little open minded about it as you can easily mistake trafficking of a protein in one direction for another if an unknown protein (SNARE for example) has defective trafficking.

We used VSVG-KDEL in this assay. This setup specifically measures retrograde trafficking. We will clarify this in the revised manuscript. We will clarify in the Discussion that we confirmed a role in retrograde trafficking but cannot exclude a role in anterograde trafficking

Reviewer #2

Major comment: scope and interpretation of DepMap-derived functional evidence The manuscript could benefit from more clearly defining the scope of the functional evidence used to nominate complexes. The central co-dependency signal is derived from DepMap 24Q2 CRISPR gene-effect profiles, which are primarily cancer cell-line fitness/proliferation data. This is an important limitation because the resulting correlations may preferentially capture complexes or pathways that influence viability in proliferating cancer cells, while missing complexes active in differentiated, tissue-specific, stimulus-dependent, or non-proliferative contexts. Conversely, some correlations may reflect shared cancer-lineage or fitness dependencies rather than direct participation in a stable complex. The authors are appropriately cautious in stating that DepCom is not a complete inventory of human protein complexes, but the title, framing, and resource description could still be read as implying a more general catalogue of functional protein complexes. The authors might consider adding a clearer introduction to DepMap and explicitly discuss how the cancer-cell-line origin of the data affects interpretation of the 518 predicted complexes. This could be addressed without new experiments, for example by adding text early in the Results section explaining what the CRISPR gene-effect scores measure, and by expanding the Discussion to clarify that DepCom represents structurally plausible complexes prioritized by co-dependency across cancer cell lines, rather than an unbiased or context-independent map of human protein complexes. The selection of highlighted examples would also benefit from clearer justification. The peroxisome, actin, WNK/TSC22D2, and Golgi/JASS examples are biologically interesting, but the rationale for choosing them is not always explicit. Were they selected because they were novel, high-confidence, disease-associated, experimentally tractable, or representative of different resource categories? Briefly stating the selection criteria would help readers understand whether these examples are illustrative case studies or representative outcomes of the pipeline.

We agree with the reviewers' assessment that this resource should be viewed as hypothesis-generating and that the overall framing should be improved. We will revise the manuscript at the appropriate sections, according to the more detailed comments of all reviewers.

Minor comments

1. Clarify post-clustering removal of large/problematic protein families and complexes. In the Methods, the authors state that "clusters of histones and keratin clusters, as well as the mito-ribosome, complexes of the electron transport chain and the mediator complex" were removed because of their large sizes. This filtering step would benefit from additional detail. Please specify the criteria used to define these removed clusters, how many clusters/proteins were removed at this stage, and whether removal was based only on size or also on biological/manual curation. It would also be helpful to explain why these proteins or clusters were removed after clustering rather than excluded before graph construction and clustering, since highly connected or compositionally biased protein families could potentially influence neighboring cluster assignments. If available, a brief robustness check showing that pre-removal of these proteins gives similar candidate complexes would strengthen confidence in the clustering procedure.

We will add the requested information to the relevant section. Alongside the manuscript we will also provide lists of the complexes before and after every filtering step

1. Clarify the rationale for excluding complexes larger than 5000 residues. The 5000-residue cutoff is understandable for AF3 computational cost, but the manuscript should briefly state how many candidate complexes were excluded by this cutoff and whether this preferentially removes known large assemblies. This would help readers understand the scope of complexes that DepCom is expected to miss.

Alongside the manuscript we will now also provide lists of the complexes before and after every filtering step.

1. Improve wording in the CAP1/CFL1/WDR1/ACTB example. The sentence "Additionally, CAP1 works in concert with CFL1 to accelerate depolymerisation, though if a four-protein complex consisting of actin, WDR1, CAP1 and CFL1 is relevant is not clear" is difficult to parse. Possible revision might be something like: "Additionally, CAP1 works in concert with CFL1 to accelerate depolymerisation, although it remains unclear whether actin, WDR1, CAP1 and CFL1 form a stable four-protein complex in cells." This more clearly separates known biology from the speculative interpretation of the DepCom prediction.

Wording will be improved.

1. Improve reproducibility details for AF3 predictions. The Methods state that predictions were run using a local AF3 installation, but reproducibility would be improved by reporting relevant AF3 settings, number of seeds/models per complex, whether templates were used, how disordered regions were handled, and whether predictions were repeated for all complexes or only selected examples. This is especially important because the manuscript notes that multiple predictions can yield different subunit arrangements.

We will provide detailed settings in the methods section. Regarding disordered parts: All predictions used full length sequences (canonical UNIPROT ID) for each protein, so disordered residues are included. If disordered regions have low PLDDT and poor PAE, these regions will simply not score as interfaces in AlphaBridge. The one exception where we did crop structures is Figure 2D, but purely for visualization purposes, the full length complex did score in the pipeline (uncropped).

Reviewer #3

Co-essentiality is not the same as physical complex membership. This is the biggest conceptual concern. Genes in the same pathway are co-essential whether or not their products bind. The authors lean on the structural prediction step to filter this out, but that means the entire pipeline rests on AF3+AlphaBridge being correct about who interacts with whom. There is no independent benchmarking shown of how often AlphaBridge calls a true positive vs a false positive at the chosen 0.5 cutoff. Why 0.5? Where does that number come from? A short benchmarking section using known complexes (CORUM 5.0, hu.MAP 2.0, the PDB) would make the choice defensible. Right now it reads as arbitrary.

We thank the reviewer for bringing up the need for such an important clarification. We fully agree that co-essentiality does not equal physical interaction and structure predictions are imperfect. This is precisely the logic underlying our pipeline design, not a limitation we overlooked. The two data sources are used sequentially and serve distinct roles: first, we construct protein sets that are connected through networks of predicted binary physical interactions; then we cluster these based on DepMap correlations, selecting likely physical complexes that display co-essentiality between their components.

In other words, clustering on DepMap data alone would certainly return many spurious correlations: as the referee points out “Co-essentiality is not the same as physical complex membership”. Anchoring the search space with structural predictions substantially reduces this noise. Neither data source alone is sufficient, nor do we claim either is definitively "correct": the value lies in their combination. We hope improved phrasing in the revised manuscript will highlight this better.

Regarding benchmarking AlphaBridge score: we have benchmarked AlphaBridge, in response to reviewer feedback on the original AlphaBridge paper (Structure, Cell Press). In the figure here it is clear that in our benchmark of PDB structures (with

Comparison to existing resources is incomplete. I can't help but wonder what was found here that would not have been possible by analysing existing resources. CORUM 5.0 (7,193 mammalian complexes, ~71% human-derived; Tsitsiridis et al. 2024 NAR), hu.MAP 2.0 (Drew et al. 2021, ~6,965 complexes from >15,000 MS experiments), BioPlex 3.0 (Huttlin et al. 2021, 118,162 interactions in HEK293T), ad the Complex Portal already cover a large fraction of the human complexome. The authors compare to PDB, the original interactome paper, and Complex Portal, but they explicitly skip CORUM and hu.MAP, both of which are central reference resources in this space. Without including these, the "60 complexes unique to DepCom" number is not really meaningful. This needs to be redone properly.

We will add the comparison with Corum and hu-MAP in the revision.

Validation rate is one out of 518. The JASS work is solid, but a single experimentally validated complex out of 518 gives the reader essentially no estimate of how often the rest of the predictions are correct. Even a smaller systematic effort, say IP-MS on five to ten predicted novel complexes in the same cell line, would do an enormous amount to establish how trustworthy the resource is. The authors already have the V5/IP-MS pipeline running. Right now the manuscript implicitly asks the reader to trust 517 predictions on the strength of one validation.

In this paper we validated one out of the 60 complexes we claim are new. Notably we provide new biological data and demonstrate how consulting our resource, or following the same logic of combining functional and structural information, can lead to new exciting discoveries. We note that out of the 518 complexes we list, 69 complexes are exactly mirrored in the PDB and/or Complex Portal, while for another 389 there is partial evidence. Thus, our dataset is amply validated, and at the same time contains data to enable new discoveries. We also note, that following the release of our resource eight months ago, a new high-impact publication “validated” a complex we have independently picked in DepMap (Oosterheert et al, Choreography of rapid actin filament by coronin, cofilin and AIP1, Cell, 2025). We will rephrase relevant sections (also in response to reviewer 2) to increase clarity about validation.

The functional and disease clustering is potentially circular. GO terms and STRING associations are themselves derived in large part from the published literature on protein function, including text mining channels in STRING, much of which is downstream of complex membership. Of course complexes cluster into "DNA repair" and "vesicle trafficking" if you cluster on GO and STRING. The same applies to Open Targets, which integrates GWAS Catalog, ClinVar, literature mining, and other sources. The clustering is fine as a navigation aid for the website, but it is not, as currently presented, an independent validation of anything. I would tone the discussion down accordingly.

We did not mean to present the clustering as an independent validation. We will tone down the discussion accordingly.

AF3 limitations on this class of problem. AF3 itself acknowledges limitations (Abramson et al. 2024, including the December 2024 addendum), and subsequent benchmarking has flagged disordered regions, dynamic/large assemblies, and certain transmembrane systems as known weak points. The JASS complex is largely transmembrane, the WNK1-TSC22D2 example involves disorder-to-order transitions, and several flagship examples involve large multi-domain proteins. The authors acknowledge some of this in passing but should state explicitly which complexes were trimmed, how the trimming choices were made, and whether predictions were repeated with different seeds to check stability. Figure S4 is a good start, but for a resource paper a more systematic seed-stability analysis is warranted.

No complexes were trimmed for the initial AF3 predictions. The WNK1-TSC22D2 example was trimmed and re-predicted only for visualization purposes. We apologize for the misunderstanding and will state this more clearly.

AF3 certainly has limitations. Regarding disordered regions, these will almost always be assigned a poor pLDDT (also if AF3 wrongly folds them into helices). AlphaBridge will not pickup these low pLDDT regions as interfaces. Regarding dynamic assemblies, these might again lead to poor confidence scores and consequently these will not be picked up as interfaces by AlphaBridge. If AF3 confidence metrics are analyzed properly, the main concern for both disordered regions and dynamic assemblies is to miss true positive interactions, rather than finding false positive. As we did not aim to identify all possible human complexes, we consider focusing on the most confidently predicted interactions to be a fair trade off.

While the JASS complex is indeed a membrane protein complex, the predictions are exceptionally confident across multiple seeds (we can provide predictions from multiple seeds for revision), and validates experimentally. Of course, structure predictions are no substitute for experimental structures, as cautioned multiple times throughout the manuscript.

Figure S4 shows that despite the complex overall geometry being flexible, the interaction sites are predicted with high confidence across different poses. Since the aim of this study was to identify proteins interacting with each other, not accurate structures (which need to be solved experimentally), we argue that recomputing all structures with multiple seeds is disproportionately expensive computationally and would delay publication of a timely study while adding little.

Statistics are thin in several places. On the Fisher exact test for Golgi/ER enrichment in V5-JTB IP-MS (Supplemental Table 1), an odds ratio of 2.77 is modest, and there is no comparison to a matched control IP. Is this more than expected by chance against an appropriate background? The IP-MS volcano plots show many significant proteins, but how was the background controlled? On the LLM section, no quantitative evaluation is presented at all and the assessment is admitted to be subjective.