10,000 Matching Annotations
  1. Last 7 days
    1. eLife Assessment

      This is a valuable study that tests the functional role of food-washing behavior in removing tooth-damaging sand and grit in long-tailed macaques and whether dominance rank predicts level of investment in the behavior. The evidence that food-washing is deliberate is compelling and the evidence that individual investment in the behavior varies is solid. Overall, the paper should be of interest to researchers interested in foraging behavior, cognition, and primate evolution.

    2. Reviewer #1 (Public review):

      In this paper, the authors had 2 aims:

      (1) Measure macaques' aversion to sand and see if its' removal is intentional, as it likely in an unpleasurable sensation that causes tooth damage.

      (2) Show that or see if monkeys engage in suboptimal behavior by cleaning foods beyond the point of diminishing returns, and see if this was related to individual traits such as sex and rank, and behavioral technique.

      They attempted to achieve these aims through a combination of geochemical analysis of sand, field experiments, and comparing predictions to an analytical model.

      The authors' conclusions were that they verified a long-standing assumption that monkeys have an aversion to sand as it contains many potentially damaging fine grained silicates, and that removing it via brushing or washing is intentional.

      They also concluded that monkeys will clean food for longer than is necessary, i.e. beyond the point of diminishing returns, and that this is rank-dependent.

      High and low-ranking monkeys tended not to wash their food, but instead over-brushed it, potentially to minimize handling time and maximize caloric intake, despite the long-term cumulative costs of sand.

      This was interpreted through the *disposable soma hypothesis*, where dominants maximize immediate needs to maintain rank and increase reproductive success at the potential expense of long-term health and survival.

      Strengths:

      The field experiment seemed well designed, and their quantification of the physical and mineral properties of quartz particles (relative to human detection thresholds) seemed good relative to their feret diameter and particle circularity (to a reviewer that is not an expert in sand). The *Rank Determination* and *Measuring Sand* sections were clear.

      In achieving Aim 1, the authors validated a commonly interpreted, but unmeasured function, of macaque and primate behavior-- a key study/finding in primate food processing and cultural transmission research.

      I commend their approach in trying to develop a quantitative model to generate predictions to compare to empirical data for their second aim.<br /> This is something others should strive for.

      I really appreciated the historical context of this paper in the introduction and found it very enjoyable and easy to read.

      I do think that interpreting these results in the context of the *disposable soma hypothesis* and the potential implications in the *paleolithic matters* section about interpreting dental wear in the fossil record are worthwhile.

    3. Author response:

      The following is the authors’ response to the previous reviews

      We thank the editors and Reviewers 1 and 3 for their though6ul consideration of our manuscript. The present revision is submitted to address comments raised concerning rank determinations and the following sentence in the editorial assessment:

      The evidence that food-washing is deliberate is compelling, but the evidence for variable and adaptive investment depending on rank, including the fitness-relevance and ultimate evolutionary implications of the findings, is incomplete given limitations of the experimental design.

      Close reading of this sentence reveals two parallel threads. The first can be read as “…evidence for variable rank is incomplete given the limitations of the experimental design,” whereas the second can be read as “…evidence for adaptive investment and fitness is incomplete given the limitations of the experimental design.” The first alludes to a critique of our methods, while the second alludes to points of discussion unrelated to our experimental design. Unpacking this sentence is important because it casts the totality of our paper as “incomplete,” a word of consequence for early-career scholars because it prevents indexing in Web of Science.

      For clarity, we will refer to these topics as Thread 1 and Thread 2 in the following response.

      Thread 1 seems rooted in a comment made by Reviewer 1, which is reproduced below:

      I am still struck that there was an analysis of only trials where <3 individuals are present. If rank was important, I would imagine that behavior might be different in social contexts when theA, scrounging, policing, aggression, or other distractions might occur-- where rank would have effects on foraging behavior. Maybe lower rankers prioritize rapid food intake then. If rank should be related to investment in this behavior, we might expect this to be magnified (or different) in social contexts where it would affect foraging. It might just be that the data was too hard to score or process in those settings, or the analysis was limited. Additionally, I think that more robust metrics of rank from more densely sampled focal follow data would be a beJer measure, but I acknowledge the limitations in getting the ideal. Since rank is central to the interpretation of these results, I think that reduced social contexts in which rank was analyzed and the robustness of the data from which rank was calculated and analyzed are the main weaknesses of the evidence presented in this paper.

      We are grateful for this perspective of Reviewer 1, but it puts us in an uncomfortable position. We must respond rather forcefully because of its influence on the above assessment. A problem with R1’s comment is that it uses the word “foraging” (a behavior we did not study) instead of “cleaning” (the behavior we did study). Still, we can substitute the latter word with the former to get the gist of it. 

      R1 criticizes our methods as a prelude for imagining the behaviors of our study animals, a form of conjecture. R1 correctly supposes a positive relationship between the number of animals and the intensity of competition for a limited food resource, a well-known phenomenon; and, yes, the food in each trial was decidedly limited, being fixed at nine cucumber slices. But R1 incorrectly presumes rank effects on cleaning under conditions of intense food competition. When the number of monkeys participating in a trial exceeded the number of feeding stations (n = 3), we saw little or no cleaning effort, either brushing or washing. So, rank effects on cleaning are immaterial under these conditions. As our study goals were narrowly focused on detecting individual propensities, or choices, as a function of rank, we limited our analysis to trials involving three monkeys or fewer. In retrospect, we admit that we should have provided better justification for our choice of trials, so we’ve edited one of our sentences:

      Original sentence 

      Formerly lines 219-220: To minimize the potential confounding effects of dominance interactions, we analyzed trials with ≤ 3 monkeys.

      Revised sentence

      Current lines 219-224: We excluded trials from analysis if the number of participating monkeys exceeded the number of feeding stations, as these conditions produced high levels of feeding competition with scant cleaning behavior. Such conditions effectively erased individual variation in sand removal, the topic motivating our experiment. Accordingly, we analyzed trials with ≤ 3 monkeys, putting 937 food-handling bouts into the GLMM statistical models, which included data on individual rank, sex, and sand treatment.

      R1’s final criticism – “I think that more robust metrics of rank from more densely sampled focal follow data would be a better measure, but I acknowledge the limitations in getting the ideal” – seems to imply that rank data were collected during our experiment. On the contrary, we determined ranks from five years of focal follows preceding the experiment, achieving the very standard that R1 describes as ideal. The relevant text appeared on lines 165-169 in version 2.0:

      To determine the rank-order of adults, we recorded dyadic agonistic interactions and their outcomes (i.e., aggression, supplants, and silent-bared-teeth displays of submission) during 5min focal follows of individuals based on a randomized order of continuous rotation (Tan et al., 2018). In some cases, these data were supplemented with ad libitum observations. This protocol existed during five years (2013-2018) of continual observations before we conducted our experiment in July-August 2018. 

      Naturally, we were puzzled by R1’s dismissal of our methods, as well as R1’s conclusion, reached without evidence, that “[the] reduced social contexts in which rank was analyzed and the robustness of the data from which rank was calculated and analyzed are the main weaknesses of the evidence presented in this paper.” It is unsubstantiated assertation with no definition of robustness, making it difficult for anyone to objectively assess the quality of our data.

      We detect in R1’s words some unfamiliarity with the social organization of our study species, which is fair enough. To better orient readers to the dominance hierarchy of Macaca fascicularis, and to boost reader confidence in the volume and quality of our rank data, we have added several sentences to this section of the manuscript, lines 169-183:

      Macaques form multi-male multi-female (polygynandrous) social groups with individual dominance hierarchies. In M. fascicularis, the hierarchy is strictly linear and extremely steep, meaning aggression is unidirectional (de Waal, 1977; van Noordwijk and van Schaik, 2001) with profound asymmetries in outcomes for individuals of adjacent ranks (Balasubramaniam et al., 2012). Further, the dominance hierarchies of philopatric females are stable and predictable. Daughters follow the pattern of youngest ascendancy, ranking just below their mothers with few known exceptions among older sisters (de Waal, 1977; van Noordwijk and van Schaik, 1999). Taken together, these species traits are conducive to unequivocal rank determinations. 

      To determine the rank-order of adults in our study group, we recorded dyadic agonistic interactions and their outcomes (i.e., aggression, supplants, and silent-bared-teeth displays of submission) during 5-min focal follows of individuals based on a randomized order of continuous rotation (Tan et al., 2018). These data were supplemented with ad libitum observations and all rank determinations were updated monthly, and when males immigrated or emigrated. This protocol predates our experiment in July-August 2018, representing 970 hr of focal data during five years of systematic study (2013-2018). 

      Thread 2 criticizes our evidence for adaptive investment and fitness, describing it is a limitation of our experimental design. Accordingly, the totality of our experiment was classified as “incomplete.” Yet, our experiment was never designed to collect such evidence, and we make no claims of having it. Rather, we discussed potential fitness consequences to highlight the broader significance of our study, connecting it diverse bodies of literature, from evolutionary theory to paleoanthropology. Our intent was to follow the conventions of scientific writing; to put our results into conversation with the wider literature and set an agenda for future research.

      On reflection, Thread 2 seems to pivot around something as arbitrary as structure. Previously, our results and discussion were combined under a single section header (“Results and Discussion”), a stylistic choice to economize words. Our manuscript is a Short Report, which is limited to 1,500 words of main text. But this level of concision proved counterproductive. It blurred our results and discussion in the minds of readers. Indeed, Reviewer 3 described it as “misleading,” a barbed word that accomplishes the same act attributed to us. To counter this perspective, we have simply partitioned our Results (now “Experimental Results”) and Discussion to draw a sharper distinction between the two components of our paper.

    1. eLife Assessment

      This study provides fundamental insights into the regulation of a retained intron in the mRNA coding for OGT, a process known to be regulated by the O-GlcNAc cycling system, and highlights the functional role of the splicing regulator SFSWAP. The evidence supporting the claims of the authors is convincing; the authors performed an elegant state-of-the-art CRISPR knockout strategy and sophisticated bioinformatic analysis to identify SFSWAP as a negative regulator of alternative splicing. The work will be of interest to researchers in the fields of splicing and glycobiology.

    2. Reviewer #1 (Public review):

      Summary:

      Govindan and Conrad use a genome-wide CRISPR screen to identify genes regulating retention of intron 4 in OGT, leveraging an intron retention reporter system previously described (PMID: 35895270). Their OGT intron 4 reporter reliably responds to O-GlcNAc levels, mirroring the endogenous splicing event. Through a genome-wide CRISPR knockout library, they uncover a range of splicing-related genes, including multiple core spliceosome components, acting as negative regulators of OGT intron 4 retention. They choose to follow up on SFSWAP, a largely understudied splicing regulator shown to undergo rapid phosphorylation in response to O-GlcNAc level changes (PMID: 32329777). RNA-sequencing reveals that SFSWAP depletion not only promotes OGT intron 4 splicing but also broadly induces exon inclusion and intron splicing, affecting decoy exon usage. While this study offers interesting insights into intron retention regulation and O-GlcNAc signaling, the RNA-Sequencing experiments lack essential controls needed to provide full confidence to the authors' conclusions.

      Strengths:

      (1) This study presents an elegant genetic screening approach to identify regulators of intron retention, uncovering core spliceosome genes as unexpected positive regulators of intron retention.<br /> (2) The work proposes a novel functional role for SFSWAP in splicing regulation, suggesting that it acts as a negative regulator of splicing and cassette exon inclusion, which contrasts with expected SR-related protein functions.<br /> (3) The authors suggest an intriguing model where SFSWAP, along with other spliceosome proteins, promotes intron retention by associating with decoy exons.

      Weaknesses:

      (1) The conclusions regarding SFSWAP's impact on alternative splicing rely on cells treated with a single pool of two siRNAs for five days. The absence of independent siRNA treatments raises concerns about potential off-target effects, which may reduce confidence in the observed SFSWAP-dependent splicing changes. Rescue experiments or using additional independent siRNA treatments would strengthen the conclusions.<br /> (2) The mechanistic role of SFSWAP in splicing would benefit from further exploration, though this may be more appropriate for future studies.

      Comments on revisions:

      The authors have addressed all my previous recommendations.

    3. Reviewer #2 (Public review):

      Summary:

      The paper describes an effort to identify the factors responsible for intron retention and alternate exon splicing in a complex system known to be regulated by the O-GlcNAc cycling system. The CRISPR/Cas9 system was used to identify potential factors. The bioinformatic analysis is sophisticated and compelling. The conclusions are of general interest and advance the field significantly.

      Strengths:

      - Exhaustive analysis of potential splicing factors in an unbiased screen.<br /> - Extensive genome wide bioinformatic analysis.<br /> - Thoughtful discussion and literature survey

      Weaknesses:

      - No firm evidence linking SFSWA to an O-GlcNAc specific mechanism.<br /> - Resulting model leaves many unanswered questions.

      Comments on revisions:

      I think the authors have adequately dealt with the overall reviewer's comments.

    4. Reviewer #3 (Public review):

      Summary:

      The major novel finding in this study is that SFSWAP, a splicing factor containing an RS domain but no canonical RNA binding domain, functions as a negative regulator of splicing. More specifically, it promotes retention of specific introns in a wide variety of transcripts including transcripts from the OGT gene previously studied by the Conrad lab. The balance between OGT intron retention and OGT complete splicing is an important regulator of O-GlcNAc expression levels in cells.

      Strengths:

      An elegant CRISPR knockout screen employed a GFP reporter, in which GFP is efficiently expressed only when the OGT retained intron is removed (so that the transcript will be exported from the nucleus to allow for translation of GFP). Factors whose CRISPR knockdown cause decreased intron retention therefore increase GFP, and these can be identified by sequencing RNA of GFP-sorted cells. SFSWAP was thus convincingly identified as a negative regulator of OGT retained intron splicing. More focused studies of OGT intron retention indicate that it may function by regulating a decoy exon previously identified in the intron, and that this may extend to other transcripts with decoy exons.

      Weaknesses:<br /> The mechanism by which SFSWAP represses retained introns is unclear, although some data suggests it can operate (in OGT) at the level of a recently reported decoy exon within that intron. Interesting / appropriate speculation about possible mechanism are provided and will likely be the subject of future studies.

      Overall the study is well done and carefully described.

    5. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews: 

      Reviewer #1 (Public review): 

      Summary: 

      Govindan and Conrad use a genome-wide CRISPR screen to identify genes regulating retention of intron 4 in OGT, leveraging an intron retention reporter system previously described (PMID: 35895270). Their OGT intron 4 reporter reliably responds to O-GlcNAc levels, mirroring the endogenous splicing event. Through a genome-wide CRISPR knockout library, they uncover a range of splicing-related genes, including multiple core spliceosome components, acting as negative regulators of OGT intron 4 retention. They choose to follow up on SFSWAP, a largely understudied splicing regulator shown to undergo rapid phosphorylation in response to O-GlcNAc level changes (PMID: 32329777). RNA-sequencing reveals that SFSWAP depletion not only promotes OGT intron 4 splicing but also broadly induces exon inclusion and intron splicing, affecting decoy exon usage. While this study offers interesting insights into intron retention and O-GlcNAc signaling regulation, the RNA sequencing experiments lack the essential controls needed to provide full confidence to the authors' conclusions. 

      Strengths: 

      (1) This study presents an elegant genetic screening approach to identify regulators of intron retention, uncovering core spliceosome genes as unexpected positive regulators of intron retention. 

      (2) The work proposes a novel functional role for SFSWAP in splicing regulation, suggesting that it acts as a negative regulator of splicing and cassette exon inclusion, which contrasts with expected SR-related protein functions. 

      (3) The authors suggest an intriguing model where SFSWAP, along with other spliceosome proteins, promotes intron retention by associating with decoy exons. 

      We thank the reviewer for recognizing and detailing the strengths of our manuscript. 

      Weaknesses: 

      (1) The conclusions on SFSWAP impact on alternative splicing are based on cells treated with two pooled siRNAs for five days. This extended incubation time without independent siRNA treatments raises concerns about off-target effects and indirect effects from secondary gene expression changes, potentially limiting confidence in direct SFSWAP-dependent splicing regulation. Rescue experiments and shorter siRNA-treatment incubation times could address these issues. 

      We repeated our SFSWAP knockdown analysis and analyzed both OGT e4-e5 junction splicing and SFSWAP transcript levels by RT-qPCR (now included in Sup. Fig. S4) from day 2 to day 5 post siRNA treatment. We observed that the time point at which OGT intron 4 removal increases (day 2) coincides with the time at which SFSWAP transcript levels start decrease, consistent with a direct effect of SFSWAP knockdown on OGT intron 4 splicing. Moreover, the effect of SFSWAP knockdown on OGT intron 4 splicing peaks between day 4-5, supporting our use of these longer time points to cast a wide net for SFSWAP targets.

      (2) The mechanistic role of SFSWAP in splicing would benefit from further exploration. Key questions remain, such as whether SFSWAP directly binds RNA, specifically the introns and exons (including the decoy exons) it appears to regulate. Furthermore, given that SFSWAP phosphorylation is influenced by changes in O-GlcNAc signaling, it would be interesting to investigate this relationship further. While generating specific phosphomutants may not yield definitive insights due to redundancy and also beyond the scope of the study, the authors could examine whether distinct SFSWAP domains, such as the SR and SURP domains, which likely overlap with phosphorylation sites, are necessary for regulating OGT intron 4 splicing. 

      We absolutely agree with the reviewer that the current work stops short of a detailed mechanistic study, and we have made every attempt to be circumspect in our interpretations to reflect that limitation. In addition, we are very interested in delving more deeply into the mechanistic aspects of this regulation. In fact, we have initiated many of the experiments suggested by the reviewer (and more), but in each case, rigorous interpretable results will require a minimum another year’s time. 

      For example, we have used crosslinking and biotin labeling techniques (using previously available reagents from Eclipsebio) to test whether SFSWAP binds RNA. The results were negative, but the lack of strong SFSWAP antibodies required that we use a transiently expressed myc-tagged SFSWAP. Therefore, this negative result could be an artifact of the exogenous expression and/or tagging. Given the difficulties of “proving the negative”, considerably more work will be required to substantiate this finding. As another example, we intend to develop a complementation assay as suggested. For an essential gene, the ideal complementation system employs a degron system, and we have spent months attempting to generate a homozygous AID-tagged SFSWAP. Unfortunately, we so far have only found heterozygotes. Of course, this could be because the tag interferes with function, the insert was not efficiently incorporated by homologous repair, or that we simply haven’t yet screened a sufficient number of clones. We’re confident that these technical issues that can be addressed, but they will take a significant amount of time to resolve. While we would ideally define a mechanism, we think that the data reported here outlining functions for SFSWAP in splicing represent a body of work sufficient for publication. 

      (3) Data presentation could be improved (specific suggestions are included in the recommendations section). Furthermore, Excel tables with gene expression and splicing analysis results should be provided as supplementary datasheets. Finally, a more detailed explanation of statistical analyses is necessary in certain sections. 

      We have addressed all specific suggestions as detailed in the recommendations below.

      Reviewer #2 (Public review): 

      Summary: 

      The paper describes an effort to identify the factors responsible for intron retention and alternate exon splicing in a complex system known to be regulated by the O-GlcNAc cycling system. The CRISPR/Cas9 system was used to identify potential factors. The bioinformatic analysis is sophisticated and compelling. The conclusions are of general interest and advance the field significantly. 

      Strengths: 

      (1) Exhaustive analysis of potential splicing factors in an unbiased screen. 

      (2) Extensive genome wide bioinformatic analysis. 

      (3) Thoughtful discussion and literature survey. 

      We thank the reviewer for recognizing and detailing the strengths of our manuscript. 

      Weaknesses: 

      (1) No firm evidence linking SFSWAP to an O-GlcNAc specific mechanism. 

      We couldn’t agree more with this critique. Indeed, our intention at the outset for the screen was to find an O-GlcNAc sensor linking OGT splicing with O-GlcNAc levels. As often occurs with high-throughput screens, we didn’t find exactly what we were looking for, but the screen nonetheless pointed us to interesting biology. Prompted by our screen, we describe new insights into the function of SFSWAP a relatively uncharacterized essential gene. Currently, we are testing other candidates from our screen, and we are performing additional studies to identify potential O-GlcNAc sensors.  

      (2) Resulting model leaves many unanswered questions. 

      We agree (see Reviewer 1, point 2 response).  

      Reviewer #3 (Public review): 

      Summary: 

      The major novel finding in this study is that SFSWAP, a splicing factor containing an RS domain but no canonical RNA binding domain, functions as a negative regulator of splicing. More specifically, it promotes retention of specific introns in a wide variety of transcripts including transcripts from the OGT gene previously studied by the Conrad lab. The balance between OGT intron retention and OGT complete splicing is an important regulator of O-GlcNAc expression levels in cells. 

      Strengths: 

      An elegant CRISPR knockout screen employed a GFP reporter, in which GFP is efficiently expressed only when the OGT retained intron is removed (so that the transcript will be exported from the nucleus to allow for translation of GFP). Factors whose CRISPR knockdown causes decreased intron retention therefore increase GFP, and can be identified by sequencing RNA of GFP-sorted cells. SFSWAP was thus convincingly identified as a negative regulator of OGT retained intron splicing. More focused studies of OGT intron retention indicate that it may function by regulating a decoy exon previously identified in the intron, and that this may extend to other transcripts with decoy exons. 

      We thank the reviewer for recognizing the strengths of our manuscript. 

      Weaknesses: 

      The mechanism by which SFSWAP represses retained introns is unclear, although some data suggests it can operate (in OGT) at the level of a recently reported decoy exon within that intron.

      Interesting/appropriate speculation about possible mechanisms are provided and will likely be the subject of future studies. 

      We completely agree that this is a limitation of the current study (see above). Now that we have a better understanding of SFSWAP functions, we will continue to explore SFSWAP mechanisms as suggested. 

      Overall the study is well done and carefully described but some figures and some experiments should be described in more detail. 

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors): 

      (1) Clarify and add missing statistical details across the figures. For example, Figure S2 lacks statistical comparisons, and in Figures 4A and 4C the tests applied should be specified in the legend. 

      We have added appropriate statistical analysis wherever missing and edited figure legends to specify the tests used.

      (2) The authors are strongly encouraged to provide detailed tables of gene expression and alternative splicing analyses from RNA-Seq experiments (e.g., edgeR, rMATS, Whippet, and MAJIQ), as this would enhance transparency and facilitate data interpretation. 

      We have added tables for gene expression and alternate splicing analysis as suggested (Suppl. tables 3-

      6).

      (3) Although the legend sometimes indicates differently (e.g., Figure 3b, 5a, 5c, etc), the volcano plots showing the splicing changes do not contain a cutoff for marginally differential percent spliced in or intron retention values. 

      The legends have been edited to reflect the correct statistical and/or PSI cutoffs.

      (4) For consistency, use a consistent volcano plot format across all relevant figures (Figures 3b, 5a-c, S3, S4, S7, and S8), including cutoffs for differential splicing and the total count of up- and down-regulated events. 

      Due to different statistical frameworks and calculations employed by different alternate splicing pipelines, we could not use the same cutoffs for different pipelines.  However, we have now indicated the number of up- and down-regulated events for consistency among the volcano plots.

      (5) What is the overlap of differentially regulated events between the different analytical methodologies applied? 

      We analyzed the degree of overlap between the three pipelines used in the paper using a Venn diagram (added to Suppl. Fig. S7). However, as widely reported in literature (e.g., Olofsson et al., 2023; Biochem Biophys Res Commun. 2023; doi: 10.1016/j.bbrc.2023.02.053.), the degree of overlap between pipelines is quite low.

      (6) To further substantiate your conclusions, additional validations of RNA-Seq splicing data, ideally visualized on an agarose gel, would be valuable, especially for exons and introns regulated by SFSWAP, and particularly for OGT decoy exons in Figure 4c. 

      We have not included these experiments as we focused on other critiques for this resubmission. Because the RNA-seq, RT-PCR and RT-qPCR data all align, we are confident that the products we are seeing are correctly identified and orthogonally validated (Figs 2d, 4a, 4b, and 4c).  

      (7) It would be more informative if the CRISPR screen data were presented in a format where both the adjusted p-value and LFC values of the hits are presented. Perhaps a volcano plot? 

      We have now included these graphs in revised Supplementary Figure S2. 

      (8) In Figure 2d, a cartoon showing primer binding sites for each panel could aid interpretation, particularly in explaining the unexpected simultaneous increase in OGT mRNA and intron retention upon SFSWAP knockdown. 

      We have added a cartoon showing primer binding sites similar to that shown in Fig. 4a.

      (9) Page 9, line 1, states that SFSWAP autoregulates its expression by controlling intron retention. Including a Sashimi plot would provide visual support for this claim. 

      The data suggesting that SFSWAP autoregulates its own transcript abundance were reported in Zachar et al. (1994), not from our own studies. Validation of those data with our RNA-seq data is confounded by the fact that we are using siRNAs to knockdown the SFSWAP RNA at the transcript level (Fig. S15). 

      (10) In the legend of Figure S2 the authors state that negative results are inconclusive because RNA knockdowns are not verified by western blotting or qRT-PCR. This is correct, but the reviewer would also argue that the positive results are also inconclusive as they are not supported by a rescue experiment to confirm that the effect is not due to off-target effects. 

      This is a fair point with respect to the siRNA experiments on their own. However, the CRISPR screen was performed with sgRNAs, and MAGeCK RRA scores are high only for those genes that have multiple sgRNAs that up-regulate the gene. Examination of the SFSWAP sgRNAs individually shows that three of four SFSWAP sgRNAs had false discovery rates ≤10<sup>-42</sup> for GFP upregulation. Thus, the siRNAs provide an additional orthogonal approach. It seems unlikely that the siRNAs, and three independent sgRNAs will have the same off-target results. Thus, these combined observations support the conclusion that SFSWAP loss leads to decreased OGT intron retention.  

      (11) For clarity in Figure 3a, consider using differential % spliced in or intron retention bar plots with directionality (positive and negative axis) and labeling siSFSWAP as the primary condition. 

      (12) Consider presenting Figure 5D as a box plot with a Wilcoxon test for statistical comparison. 

      For both points 11 and 12, we have tried the graphs as the reviewer suggested. While these were good suggestions, in both cases we felt that the original plots ended up presenting a clearer presentation of the data (see Author response image 1).

      Author response image 1.

      (13) Please expand the Methods section to detail the Whippet and MAJIQ analyses. 

      We have expanded the methods section to include additional details of the alternate splicing analysis.

      (14) Include coordinates for the four possible OGT decoy exon combinations analyzed in the Methods section. 

      We have added the coordinates of all four decoy forms in the methods section.  

      (15) A section on SFSWAP mass spectrometry is listed in Methods but is missing from the manuscript. 

      This section has now been removed.

      Reviewer #2 (Recommendations for the authors): 

      This is an excellent contribution. The paper describes an effort to identify the factors responsible for intron retention and alternate exon splicing in a complex system known to be regulated by the O-GlcNAc cycling system. The CRISPR/Cas9 system was used to identify potential factors. The bioinformatic analysis is sophisticated and compelling. The conclusions are of general interest and advance the field significantly. 

      Some specific recommendations. 

      (1) The plots in Figure 3 describing SI and ES events are confusing to this reader. Perhaps the violin plot is not the best way to visualize these events. The same holds true for the histograms in the lower panel of Figure 3. Not sure what to make of these plots. 

      For Figure 3b, we include both scatter and violin plots to represent the same data in two distinct ways. For Figure 3d, we agree that these are not the simplest plots to understand, and we have spent significant time trying to come up with a better way of displaying these trends in GC content as they relate to SE and RI events. Unfortunately, we were unable to identify a clearer way to present these data. 

      (2) The model (Figure 6) is very useful but confusing. The legend and the Figure itself are somewhat inconsistent. The bottom line of the figure is apparent but I fear that the authors are trying to convey a more complete model than is apparent from this figure. Please revise. 

      We have simplified the figure from the previous submission. As mentioned above, we admit that mechanistic details remain unknown. However, we have tried to generate a model that reflects our data, adds some speculative elements to be tested in the future, but remains as simple as possible. We are not quite sure what the reviewer was referring to as “somewhat inconsistent”, but we have attempted to clarify the model in the revised Discussion and Figure legend.  

      (3) It is unclear how normalization of the RNA seq experiments was performed (eg. Figure S5 and 6).  

      The normalization differences in Fig. S5 and S6 (now Fig S8 and S9) were due to scaling differences during the use of rmats2sashimiplot software. We have now replaced Fig. S5 to reflect correctly scaled images.

      I am enthusiastic about the manuscript and feel that with some clarification it will be an important contribution. 

      Thank you for these positive comments about our study!

      Reviewer #3 (Recommendations for the authors): 

      (1) In Figure 1f, it is clear that siRNA-mediated knockdown of OGT greatly increases spliced RNA as the cells attempt to compensate by more efficient intron removal (three left lanes). However, there is no discussion of the various treatments with TG or OSMI. Might quantitation of these lanes not also show the desired effects of TG and OSMI on spliced transcript levels? 

      The strong effect of OGT knockdown masks the (comparatively modest) effects of subsequent inhibitor treatments on the reporter RNA. We have edited the results section to clarify this.

      (2) In Figure 2c, why is the size difference between spliced RNA and intron-retained RNA so different in the GFP-probed gel (right) compared with the OGT-probed gel (left)? Even recognizing that the GFP probe is directed against reporter transcripts, and the OGT probe (I think) is directed against endogenous OGT transcripts, shouldn't the difference between spliced and unspliced bands be the same, i.e., +/- the intron 4 sequence. Also, why does the GFP probe detect the unspliced transcript so poorly? 

      The fully spliced endogenous OGT mRNA is ~5.5 kb while the fully spliced reporter is only ~1.6kb, so the difference in size (the apparent shift relative to the mRNA) is quite different. Moreover, the two panels in Fig 2c are not precisely scaled to one another, so direct comparisons cannot be made. 

      The intron retained isoform does not accumulate to high levels in this reporter, a phenotype that we also observed with our GFP reporter designed to probe the regulation of the MAT2A retained intron (Scarborough et al., 2021). We are not certain about the reason for these observations, but suspect that the reporter RNA’s retained intron isoforms are less stable in the nucleus than their endogenous counterparts. Alternatively, the lack of splicing may affect 3´ processing of the transcripts so that they do not accumulate to the high levels observed for the wild-type genes. 

      (3) Please provide more information about the RNA-seq experiments. How many replicates were performed under each of the various conditions? The methods section says three replicates were performed for the UPF1/TG experiments; was this also true for the SFSWAP experiments?  

      All RNA-seq experiments were performed in biological triplicates. We have edited the methods section to clarify this.

      (4) Relatedly, the several IGV screenshots shown in Figure 3C presumably represent the triplicate RNA seq experiments. In part D, how many experiments does the data represent? Is it a compilation of three experiments? 

      Fig. 3d is derived from alternate splicing analysis performed on three biological replicates. We have added the number of replicates (n=3) on the figure to clarify this. We have also noted that the three IGV tracks represent biological replicates in the Figure legend for 3c.  

      (5) Please provide more details regarding the qRT-PCR experiments. 

      We have provided the positions of primer sets used for RT-qPCR analysis and cartoon depictions of target sites below the data wherever appropriate.

      (6) In the discussion of decoy exon function (in the Discussion section), several relevant observations are cited to support a model in which decoy exons promote assembly of splicing factors. One might also cite the finding that eCLIP profiling has found enriched binding of U2AF1 and U2AF2 at the 5' splice site region of decoy exons (reference 16). 

      Excellent point. This has now been added to the Discussion. 

      Minor corrections / clarifications: 

      (1) In the Figure 2A legend, CRISPR is misspelled. 

      Corrected.

      (2) In the discussion, the phrase "indirectly inhibits splicing of exons 4 and 5, but promoting stable unproductive assembly of the spliceosome", the word "but" should probably be "by". 

      Corrected.

    1. eLife Assessment

      The role of ACVR2A is potentially of importance to both the biology of trophoblast cells and to the pathogenesis of preeclampsia. In this manuscript, the authors have taken a useful first step towards better understanding this protein using a loss of function model in trophoblast cell lines and then examining invasion, proliferation, and transcription in these cells. The study is solid and further in vivo evidence on how target factors participate in the occurrence of placental structural disorders and diseases through potential downstream pathways will be invaluable in the future.

    2. Reviewer #1 (Public review):

      Summary:

      This study has preliminarily revealed the role of ACVR2A in trophoblast cell function, including its effects on migration, invasion, proliferation, and clonal formation, as well as its downstream signaling pathways.

      Strengths:

      The use of multiple experimental techniques, such as CRISPR/Cas9-mediated gene knockout, RNA-seq, and functional assays (e.g., Transwell, colony formation, and scratch assays), is commendable and demonstrates the authors' effort to elucidate the molecular mechanisms underlying ACVR2A's regulation of trophoblast function. The RNA-seq analysis and subsequent GSEA findings offer valuable insights into the pathways affected by ACVR2A knockout, particularly the Wnt and TCF7/c-JUN signaling pathways.

      Weaknesses:

      The current findings provide valuable insights into the role of ACVR2A in trophoblast cell function and its involvement in the regulation of migration, invasion, and proliferation, further validation in both in vitro and in vivo models would strengthen the conclusions. Additional techniques, such as animal models and more advanced clinical sample analyses, would help strengthen the conclusions and provide a more comprehensive understanding of the molecular pathways involved.

    3. Reviewer #2 (Public review):

      Summary:

      ACVR2A is one of a handful of genes for which significant correlations between associated SNPs and the incidences of preeclampsia have been found in multiple populations. It is one of the TGFB family receptors, and multiple ligands of ACVR2A, as well as its coreceptors and related inhibitors, have been implicated in placental development, trophoblast invasion, and embryo implantation. This useful study builds on this knowledge by showing that ACVR2A knockout in trophoblast-related cell lines reduces trophoblast invasion, which could tie together many of these observations. The implication of cross-talk between the WNT and ACRV2A/SMAD2 pathways is an important contribution to the understanding of the regulation of trophoblast function.

      Strengths:

      (1) ACVR2A is one of very few genes implicated in preeclampsia in multiple human populations, yet its role in pathogenesis is not very well studied and this study begins to address that hole in our knowledge.

      (2) ACVR2A is also indirectly implicated in trophoblast invasion and trophoblast development via its connections to many ligands, inhibitors, and coreceptors, suggesting its potential importance.

      (3) The authors have used multiple cell lines to verify their most important observations.

      Editors' note: Following the first round of peer review, the original reviewers were not available to review the revised manuscript. As several specific weakness detailed by the reviewers were largely addressed in the revised manuscript, they are not included here.

    4. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      This study has preliminarily revealed the role of ACVR2A in trophoblast cell function, including its effects on migration, invasion, proliferation, and clonal formation, as well as its downstream signaling pathways.

      Strengths:

      The use of multiple experimental techniques, such as CRISPR/Cas9-mediated gene knockout, RNA-seq, and functional assays (e.g., Transwell, colony formation, and scratch assays), is commendable and demonstrates the authors' effort to elucidate the molecular mechanisms underlying ACVR2A's regulation of trophoblast function. The RNA-seq analysis and subsequent GSEA findings offer valuable insights into the pathways affected by ACVR2A knockout, particularly the Wnt and TCF7/c-JUN signaling pathways.

      Weaknesses:

      The molecular mechanisms underlying this study require further exploration through additional experiments. While the current findings provide valuable insights into the role of ACVR2A in trophoblast cell function and its involvement in the regulation of migration, invasion, and proliferation, further validation in both in vitro and in vivo models is needed. Additionally, more experiments are required to establish the functional relevance of the TCF7/c-JUN pathway and its clinical significance, particularly in relation to pre-eclampsia. Additional techniques, such as animal models and more advanced clinical sample analyses, would help strengthen the conclusions and provide a more comprehensive understanding of the molecular pathways involved.

      Reviewer #2 (Public review):

      Summary:

      ACVR2A is one of a handful of genes for which significant correlations between associated SNPs and the incidences of preeclampsia have been found in multiple populations. It is one of the TGFB family receptors, and multiple ligands of ACVR2A, as well as its coreceptors and related inhibitors, have been implicated in placental development, trophoblast invasion, and embryo implantation. This useful study builds on this knowledge by showing that ACVR2A knockout in trophoblast-related cell lines reduces trophoblast invasion, which could tie together many of these observations. Support for this finding is incomplete, as reduced proliferation may be influencing the invasion results. The implication of cross-talk between the WNT and ACRV2A/SMAD2 pathways is an important contribution to the understanding of the regulation of trophoblast function.

      Strengths:

      (1) ACVR2A is one of very few genes implicated in preeclampsia in multiple human populations, yet its role in pathogenesis is not very well studied and this study begins to address that hole in our knowledge.

      (2) ACVR2A is also indirectly implicated in trophoblast invasion and trophoblast development via its connections to many ligands, inhibitors, and coreceptors, suggesting its potential importance.

      (3) The authors have used multiple cell lines to verify their most important observations.

      Weaknesses:

      (1) There are a number of claims made in the introduction without attribution. For example, there are no citations for the claims that family history is a significant risk factor for PE, that inadequate trophoblast invasion of spiral arteries is a key factor, and that immune responses, and reninangiotensin activity are involved.

      Thank you for pointing out the lack of citations in some parts of the introduction. We have revised the manuscript to include appropriate references for the claims regarding family history as a risk factor for PE, the role of inadequate trophoblast invasion in spiral arteries, and the involvement of immune responses and the renin-angiotensin system. The revised text now includes citations to well-established studies in the field (Salonen Ros et al., 2000; Chappell LC et al., 2021; Brosens et al., 2002; Knofler et al., 2019; Redman CWG et al., 1999; LaMarca B et al., 2008). We believe these additions improve the scientific rigor of the manuscript.

      (2) The introduction states "As a receptor for activin A, ACVR2A..." It's important to acknowledge that ACVR2A is also the receptor for other TGFB family members, with varying affinities and coreceptors. Several TGFB family members are known to regulate trophoblast differentiation and invasion. For example, BMP2 likely stimulates trophoblast invasion at least in part via ACVR2A (PMID 29846546).

      Thank you for highlighting the broader role of ACVR2A as a receptor for multiple members of the TGF-β superfamily. We have revised the introduction to acknowledge that ACVR2A is not only the receptor for activin A but also interacts with other ligands, such as BMP2, which likely stimulates trophoblast invasion via ACVR2A (PMID: 29846546). This addition provides a more comprehensive view of ACVR2A's function in trophoblast biology. While the focus of our current study is on activin A, we agree that ACVR2A's role in mediating the effects of other TGF-β family members is an important topic for future research.

      (3) An alternative hypothesis for the potential role of ACVR2A in preeclampsia is its functions in the endometrium. In the mouse ACVR2A knockout in the uterus (and other progesterone receptorexpressing cells) leads to embryo implantation failure.

      Thank you for bringing up the potential role of ACVR2A in the endometrium as an alternative hypothesis. We have revised the discussion to acknowledge this possibility and cited relevant studies showing that uterine-specific knockout of ACVR2A in mice leads to embryo implantation failure (Monsivais et al., 2021). This suggests that ACVR2A may play a critical role in uterine receptivity and embryo implantation, which could influence placental development and preeclampsia pathogenesis. While our current study focuses on trophoblast-related functions of ACVR2A, we agree that investigating its role in the uterine environment is an important direction for future research.

      (4) In the description of the patient population for placental sample collections, preeclampsia is defined only by hypertension, and this is described as being in accordance with ACOG guidelines. ACOG requires a finding of hypertension in combination with either proteinuria or one of the following: thrombocytopenia, elevated creatinine, elevated liver enzymes, pulmonary, edema, and new onset unresponsive headache.

      We appreciate the reviewer’s detailed observation regarding the definition of preeclampsia.

      We have reviewed and clarified our description of the diagnostic criteria based on the American College of Obstetricians and Gynecologists (ACOG) guidelines. Specifically, we have revised the definition in the Materials and Methods section under "Collection of Placenta and Decidua Specimens," as follows: In accordance with the guidelines from the American College of Obstetricians and Gynecologists (ACOG, 2023), preeclampsia (PE) is diagnosed as hypertension (systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg on at least two occasions) in combination with one or more of the following: proteinuria (≥300 mg/24-hour urine collection or protein/creatinine ratio ≥0.3), thrombocytopenia, elevated serum creatinine, elevated liver enzymes, pulmonary edema, or new-onset headache unresponsive to treatment.

      (5) I believe that Figures 1a and 1b are data from a previously published RNAseq dataset, though it is not entirely clear in the text. The methods section does not include a description of the analysis of these data undertaken here. It would be helpful to include at least a brief description of the study these data are taken from - how many samples, how were the PE/control groups defined, gestational age range, where is it from, etc. For the heatmap presented in B, what is the significance of the other genes/ why are they being shown? If the purpose of these two panels is to show differential expression specifically of ACVR2A in this dataset, that could be shown more directly.

      Clarification of RNAseq dataset: The Methods section has been revised to specify the dataset source (GEO accession number: GSE114691), which includes 20 PE and 21 control placental samples with gestational ages ranging from 34 to 38 weeks. PE and control groups were defined using clinical criteria such as hypertension and proteinuria, and these details have also been added to the Results section. RNAseq analysis description: We have included details of the differential gene expression analysis in the Methods section. Specifically, the DESeq2 R package was used, with thresholds of FDR < 0.05 and |log2(fold change) | ≥ 1. The selection of WNT pathwayrelated genes in Figure 1B is based on these analyses. Significance of the heatmap genes: The genes displayed in Figure 1B were selected based on their significant differential expression and enrichment in pathways relevant to PE pathogenesis, such as the WNT signaling pathway. We have clarified this in the Results section and updated the figure legend to explain their biological relevance. Purpose of Figures 1A and 1B: Figure 1A emphasizes the downregulation of ACVR2A in PE placentas, while Figure 1B complements this by presenting differentially expressed genes associated with the WNT pathway. These figures collectively highlight the role of ACVR2A in PE and its connection to broader molecular pathways. Text descriptions have been updated to improve clarity and focus.

      (6) More information is needed in the methods section to understand how the immunohistochemistry was quantified. "Quantitation was performed" is all that is provided. Was staining quantified across the whole image or only in anchoring villous areas? How were HRP & hematoxylin signals distinguished in ImageJ? How was the overall level of HRP/DAB development kept constant between the NC and PE groups?

      Thank you for pointing out the need for more details regarding the quantification of immunohistochemistry (IHC). We have now clarified and expanded the description of the IHC quantification process in the Methods section as follows: Quantification Across the Entire Section: IHC staining was assessed across the entire tissue section to account for global expression patterns. For quantitative analysis, representative regions from the anchoring villous areas, where ACVR2A expression is most prominent, were selected for comparison between NC and PE groups. This ensured that the analysis focused on biologically relevant regions. ImageJ Analysis:

      Images of stained sections were captured under identical magnifications and lighting conditions. Hematoxylin (blue, nuclear staining) and DAB/HRP (brown, protein-specific signal) were distinguished using ImageJ's color deconvolution plugin. The DAB/HRP signal was isolated and quantified based on the integrated optical density (IOD) within the selected regions. Consistency in HRP/DAB Development: To maintain consistency between NC and PE groups, all tissue samples were processed under identical experimental conditions, including the same antibody dilution, incubation times, and DAB/HRP development durations. Negative controls (without primary antibody) were included to monitor background staining, and the DAB reaction was stopped simultaneously across all samples to avoid overdevelopment. Statistical Analysis: The quantified DAB signal intensity was normalized to the area of the selected regions, and comparisons between NC and PE groups were performed using statistical tests (e.g., Student’s ttest). Results are reported as mean ± SD. We hope this additional detail addresses your concerns.

      (7) In Figure 1E it is not immediately obvious to many readers where the EVT are. It is probably worth circling or putting an arrow to the little region of ACVR2A+ EVT that is shown in the higher magnification image in Figure 1E. These are actually easier to see in the pictures provided in the supplement Figure 1. Of note, the STB is also staining positive. This is worth pointing out in the results text.

      Thank you for your suggestion regarding Figure 1E. To make the location of the ACVR2A+ extravillous trophoblasts (EVTs) more apparent, we have updated Figure 1E by adding arrows to indicate the regions of EVTs in the higher magnification image. Additionally, we have included annotations in the supplemental Figure S1 to further aid visualization. We appreciate your observation that syncytiotrophoblasts (STBs) also show positive staining for ACVR2A. We have revised the Results section to explicitly mention this finding and its potential significance.

      (8) It is not possible to judge whether the IF images in 1F actually depict anchoring villi. The DAPI is really faint, and it's high magnification, so there isn't a lot of context. Would it be possible to include a lower magnification image that shows where these cells are located within a placental section? It is also somewhat surprising that this receptor is expressed in the cytoplasm rather than at the cell surface. How do the authors explain this?

      Thank you for your suggestion to provide more context for the immunofluorescence (IF) images in Figure 1F. To address this, we have included lower magnification images in Supplementary Figure S2, showing the overall structure of the placental section and the location of the anchoring villi. These images help to contextualize the regions analyzed in Figure 1F, which were selected to clearly illustrate ACVR2A expression in extravillous trophoblasts (EVTs). In Figure 1F, we have focused on higher magnification images for better visualization of ACVR2A staining patterns in EVTs. Regarding the subcellular localization of ACVR2A, the receptor is predominantly expressed on the cell surface, as shown in our images. However, some intracellular staining is also observed, which may reflect receptor trafficking or recycling processes, consistent with the behavior of other activin receptors under physiological or pathological conditions. We have clarified these points in the Results and Discussion sections.

      (9) The results text makes it sound like the data in Figure 2A are from NCBI & Protein atlas, but the legend says it is qPCR from this lab. The methods do not detail how these various cell lines were grown; only HTR-SVNeo cell culture is described. Similarly, JAR cells are used for several experiments and their culture is not described.

      Thank you for pointing out the need for clarification regarding Figure 2A and cell culture methods. The data in Figure 2A were generated using RT-qPCR conducted in our laboratory, not solely based on data from NCBI or the Human Protein Atlas. We have revised the Results section to reflect this more accurately. Regarding the culture conditions, we acknowledge that the methods for other cell lines were not explicitly detailed. For this study, all cell lines, including JAR and other cancer cell lines, were cultured following standard protocols provided by the suppliers. Specifically, JAR cells and other cell lines were purchased from Wuhan Punosei Life Technology and were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin under standard conditions (37°C, 5% CO<sub>2</sub>). This information has been added to the Methods section for clarity.

      (10) Under RT-qPCR methods, the phrase "cDNA reverse transcription cell RNA was isolated..." does not make any sense.

      Thank you for pointing out the unclear phrasing in the RT-qPCR methods section. We agree that the original description was not precise. To address this, we have revised the relevant section to improve clarity and accuracy. Specifically, the methods now explicitly describe the two key steps: RNA isolation and cDNA synthesis. The revised text reads: Total RNA was isolated from cells using a Total RNA Extraction Kit (TIANGEN, China) following the manufacturer’s instructions. The extracted RNA was reverse-transcribed into complementary DNA (cDNA) using a cDNA Synthesis Kit (Takara, Japan) according to the protocol provided by the manufacturer.

      (11) The paragraph beginning "Consequently, a potential association..." is quite confusing. It mentions analyzing ACVR2A expression in placentas, but then doesn't point to any results of this kind and repeats describing the results in Figure 2a, from various cell lines.

      Thank you for your comment regarding the paragraph beginning with "Consequently, a potential association...". We understand that the current wording may create confusion. The primary aim of this section is to compare ACVR2A expression levels across various cell lines, including trophoblast-derived and non-trophoblast cell lines, to highlight the relevance of ACVR2A in trophoblast function, particularly in invasion and migration. To address your concerns, we have revised the paragraph for clarity and logical flow. The updated text explicitly focuses on the comparison of ACVR2A expression across cell lines (Figure 2A) and how this supports the hypothesis that ACVR2A plays a key role in trophoblast invasion and migration. Additionally, the discussion of placental samples has been separated to avoid confusion with cell line results. We hope this revision resolves the issue.

      (12) The authors should acknowledge that the effect of the ACVR2A knockout on proliferation makes it difficult to draw any conclusions from the trophoblast invasion assays. That is, there might be fewer migrating or invading cells in the knockout lines because there are fewer cells, not because the cells that are there are less invasive. Since this is a central conclusion of the study, it is a major drawback.

      Thank you for highlighting this important point. We agree that the reduced proliferation observed in ACVR2A knockout cells could influence the results of the invasion assays, as fewer cells may inherently lead to reduced invasion. To minimize this effect, we conducted the invasion and migration assays under low-serum conditions (1–2% serum) to limit cell proliferation during the experimental timeframe. This approach was based on optimization trials and existing literature, as serum-free conditions were found to negatively impact cell viability and experimental reproducibility. While these efforts helped to mitigate the impact of proliferation on the results, we acknowledge this as a limitation of our study and have added this discussion to the manuscript. Future studies could incorporate approaches such as normalizing cell numbers or using additional proliferation-independent methods to confirm the findings. We hope this clarification and the steps taken address your concerns.

      (13) The legend and the methods section do not agree on how many fields were selected for counting in the transwell invasion assays in Figure 3C. The methods section and the graph do not match the number of replicate experiments in Figure 3D (the number of replicate experiments isn't described for 3C).

      Thank you for pointing out the inconsistencies regarding the number of fields counted and the number of replicates in the Transwell invasion assays (Figure 3C) and colony formation assays (Figure 3D). We apologize for the lack of clarity in the Methods section and figure legend. To address this, we have revised both the figure legends and the Methods section for consistency and added detailed descriptions. For Figure 3C, cell invasion was quantified by randomly selecting 5 fields of view per sample under 300× magnification. Images shown in the figure were taken at lower magnification to provide a better visual comparison between experimental and control groups. For Figure 3D, each experiment was independently repeated at least 10 times to ensure robust and reproducible results. These clarifications have been incorporated into the revised manuscript. We appreciate your feedback and believe this revision improves the clarity and transparency of our methods.

      (14) Discussion says "Transcriptome sequencing analysis revealed low ACVR2A expression in placental samples from PE patients, consistent with GWAS results across diverse populations." The authors should explain this briefly. Why would SNPs in ACVR2A necessarily affect levels of the transcript?

      Thank you for raising this important point. We acknowledge that our study did not directly investigate how SNPs in the ACVR2A gene affect transcript levels. However, prior studies have suggested that SNPs can influence gene expression through various mechanisms. For example, SNPs in regulatory regions (e.g., promoters, enhancers, or untranslated regions) may affect transcription factor binding, RNA stability, or splicing efficiency, ultimately altering transcript levels. While we did not directly assess the functional consequences of ACVR2A SNPs in this study, the observed downregulation of ACVR2A in PE placentas aligns with the potential regulatory impact of SNPs previously identified in GWAS studies. To address this, we have revised the Discussion section to clarify the relationship between SNPs and transcript levels and acknowledge this limitation.  

      (15) "The expression levels of ACVR2A mRNA were comparable to those of tumor cells such as A549. This discovery suggested a potential pivotal role of ACVR2A in the biological functions of trophoblast cells, especially in the nurturing layer." Alternatively, ACVR2A expression resembles that of tumors because the cell lines used here are tumor cells (JAR) or immortalized cells (HTR8). These lines are widely used to study trophoblast properties, but the discussion should at least acknowledge the possibility that the behavior of these cells does not always resemble normal trophoblasts.

      Thank you for pointing out this important limitation. We agree that the JAR and HTR8/SVneo cell lines, being tumor-derived or immortalized, may not fully replicate the behavior of normal trophoblast cells. While these cell lines are widely used as models for studying trophoblast properties due to their ease of culture and invasive behavior, their gene expression and signaling pathways could partially reflect their tumorigenic or immortalized origins. We have revised the Discussion section to acknowledge this limitation and clarify the interpretation of ACVR2A expression levels in these cells.

      (16) The authors should discuss some of what is known about the relationship between the TCF7/c-JUN pathway and the major signaling pathway activated by ACVR2A, Smad 2/3/4. The Wnt and TGFB family cross-talk is quite complex and it has been studied in other systems.

      Thank you for highlighting the relationship between the TCF7/c-JUN pathway and Smad2/3/4 signaling. In our study, we chose to focus on Smad1/5 due to its strong association with ACVR2A in placental development, as demonstrated in a recent study(DOI: 10.1038/s41467-021-23571-5). This study showed that the BMP signaling pathway, mediated through ACVR2A-Smad1/5, is essential for endometrial receptivity and embryo implantation. While Smad2/3/4 are wellestablished mediators of TGF-β signaling, Smad1/5 activation is more directly linked to ACVR2A in the context of reproductive biology.

      In PE placentas, we observed a significant downregulation of Smad1/5 expression, which supports the hypothesis that ACVR2A-mediated Smad signaling is disrupted in this condition. Although we did not directly assess Smad2/3/4 in this study, prior research has shown that Smad2/3 can interact with TCF/LEF transcription factors to regulate Wnt-related target genes, suggesting potential cross-talk between these pathways. We have now clarified this rationale and included a discussion of these interactions in the revised manuscript.

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      Several points need to be addressed to improve the clarity and robustness of the presented findings:

      (1) From a clinical perspective, several concerns arise regarding the interpretation of these findings. First, the small sample size of 20 patients may not be representative of the broader population, limiting the generalizability of the results. Additionally, although no significant differences in age and pre-pregnancy BMI were observed between the PE and normal control groups, other clinical variables, such as hypertension or gestational diabetes, may also influence ACVR2A expression and contribute to PE development. Furthermore, while the study suggests a correlation between reduced ACVR2A expression and PE, it remains unclear whether this association holds true across different subtypes of PE or whether there are other underlying clinical factors that could account for these changes in gene expression. These factors need to be considered in future studies to better understand the clinical relevance of ACVR2A in PE.

      Thank you for raising these insightful concerns about the clinical interpretation of our findings. We agree that the small sample size of 20 patients may limit the generalizability of our results. To address this, we are actively expanding our cohort by collecting additional clinical samples from PE patients and normotensive controls. This effort aims to strengthen the robustness of our findings and provide stronger evidence for the role of ACVR2A in PE. We would also like to clarify that, during the initial sample collection, we specifically included only PE patients without comorbidities such as gestational diabetes, chronic hypertension, or other pregnancy-related complications. This strict selection criterion was implemented to minimize the potential influence of confounding clinical variables and ensure that our findings specifically reflect the association between ACVR2A expression and PE. While our study provides important initial insights, we recognize the need for larger-scale studies to validate these findings. The ongoing collection of clinical samples will allow us to address this limitation and enhance the translational relevance of our research. We have revised the manuscript to reflect these points and highlight our plans to strengthen the study by increasing the sample size.

      (2) The section "Precision Genome Surgery: ACVR2A Knockout via CRISPR/Cas9" in the results contains some issues with expression details. The results section should be more structured, with data presented in a more detailed and clear manner, ensuring that there is a clear connection between each experimental step and its corresponding result. For example, the sentence "Following multiple rounds of monoclonal culture, genotype identification, RT-qPCR and Western blotting (WB) analysis for screening, specific double-knockout monoclonal cell lines were distinctly chosen" contains redundant phrasing and unnecessary details, which affect the flow of the text.

      Thank you for your constructive feedback on the “Precision Genome Surgery: ACVR2A Knockout via CRISPR/Cas9” section. We agree that this section can be better structured to present the data in a more detailed and coherent manner. To address this, we have reorganized the results into distinct steps, ensuring a clear connection between each experimental step and its corresponding result. Redundant phrasing has been removed to improve the flow and readability of the text. The revised section emphasizes the purpose of each step, the screening process, and the specific results obtained.

      (3) The figure legends and panel labels in Figure 3 should be revised to ensure clarity and consistency. The figure legend should specify the exact panels (e.g., Figure 3A, 3B, 3C, etc.) and clearly describe the experimental conditions and results shown in each part.

      Thank you for pointing out the need for improved clarity and consistency in the figure legends and panel labels for Figure 3. We have revised the figure legend to specify each panel (e.g., Figure 3A, 3B, 3C, etc.) and included detailed descriptions of the experimental conditions and results displayed in each part. These updates aim to ensure better understanding and alignment between the figure legend and the panels.

      (4) Lack of In Vivo Validation of ACVR2A Knockout: The study does not include in vivo experiments to validate the effects of ACVR2A knockout. It would be important to investigate whether similar regulatory effects of ACVR2A on trophoblast cell migration and invasion can be observed in animal models or in larger clinical studies. The lack of in vivo data raises questions about the translational relevance of the findings.

      Thank you for highlighting the importance of in vivo validation to assess the translational relevance of our findings. While we acknowledge that in vivo experiments could provide additional insights into the role of ACVR2A in trophoblast migration and invasion, this study was primarily designed as an in vitro investigation to explore the molecular mechanisms underlying ACVR2A function in trophoblast cells. The choice of an in vitro model allowed us to perform precise and controlled mechanistic analyses, which are critical for establishing a foundation for future research. We agree that in vivo studies using animal models or larger clinical cohorts are important next steps to validate the regulatory effects of ACVR2A on trophoblast function and its contribution to PE pathogenesis. These directions will be pursued in future research to further establish the translational potential of our findings. We have included this perspective in the revised Discussion section.

      (5) TCF7/c-JUN Pathway in Clinical Samples: In the study of the TCF7/c-JUN pathway, the authors mention assessing protein expression in clinical samples through immunohistochemistry (IHC). However, the manuscript does not provide a clear explanation of how the findings from laboratory cell models (such as HTR8/SVneo and JAR) relate to the clinical samples. Specifically, while ACVR2A knockout is shown to affect these proteins at the cellular level, it is unclear whether this effect is observed in clinical samples. Therefore, further validation of the TCF7/c-JUN pathway in the cell models and exploration of its relationship with protein expression in clinical samples is necessary. Additional experiments, such as immunofluorescence staining or mass spectrometry, could further confirm the role of the TCF7/c-JUN pathway in cells and provide a more direct comparison with clinical data.

      Thank you for highlighting the need to connect findings from cell models to clinical samples, particularly with respect to the TCF7/c-JUN pathway. In response to your comment, we conducted additional experiments using Western blot analysis to evaluate the expression of ACVR2A, SMAD1/5, SMAD4, pSMAD1/5/9, and TCF7L1/TCF7L2 in PE placental tissues compared to normotensive controls (Figure 7A). The results demonstrated significantly reduced expression of these proteins in PE placentas, providing evidence that disruptions in the ACVR2A-SMAD and TCF7/c-JUN signaling pathways observed in vitro are also present in clinical samples.

      These findings strengthen the translational relevance of our study by directly linking the molecular mechanisms identified in cell models to clinical observations. We have updated the Results and Discussion sections to incorporate these new data, and we believe this addition addresses your concern about the relationship between in vitro and clinical findings.

    1. eLife Assessment

      This important study highlights the key role of the gut-liver axis mediated by LPS in causing hepatic steatosis. The authors provide solid evidence, in vivo, in vitro, and in silico, for the role of acyloxyacyl hydrolase in mediating this effect using KO mice subjected to MASD-inducing diets. The findings are significant for the liver research community and others interested in the gut-liver axis.

    2. Reviewer #1 (Public review):

      Lu et. al. proposed here a direct role of LPS in inducing hepatic fat accumulation and that metabolism of LPS therefore can mitigate fatty liver injury. With an Acyloxyacyl hydrolase whole-body KO mice, they demonstrated that Acyloxyacyl hydrolase deletion resulted in higher hepatic fat accumulation over 7 months of high glucose/high fructose diet. Previous literature has found that hepatocyte TLR4 (which is a main receptor for binding LPS) KO reduced fatty liver in MAFLD model, and this paper complement this by showing that degradation/metabolism of LPS can also reduce fatty liver. Using clodronate-liposomes to deplete KC, the authors went on to show that AOAH level decreased significantly with increased SREBP1 level, suggesting that KCs were the major source of AOAH in the liver. To explain the mechanism of LPS induced lipogenesis, the authors demostrated in vitro that LPS alone without KC can induce SREBP1 level in primary hepatocytes via mTOR activation. This result proposed a very interesting mechanism, and the translational implications of utilizing Acyloxyacyl hydrolase to decrease LPS exposure is intriguing.

      The strengths of the present study include that they raised a very simplistic mechanism with LPS that is of interest in many diseases. The phenotype shown in the study is strong. The mechanism proposed by the findings are generally well supported. Manuscript significantly improved with revision. Overall, this work adds to the current understanding of the gut-liver axis and development of MAFLD, and will be of interest to many readers.

    3. Reviewer #2 (Public review):

      The authors of this article investigated the impact of the host enzyme AOAH on the progression of MASLD in mice. To achieve this, they utilized whole-body Aoah-/- mice. The authors demonstrated that AOAH reduced LPS-induced lipid accumulation in the liver, probably by decreasing the expression and activation of SREBP1. In addition, AOAH reduced hepatic inflammation and minimized tissue damage.

      The authors have effectively addressed some key questions I raised. However, I still have some lingering concerns regarding the mechanisms underlying AOAH's effects.

      (1) AOAH is expressed in the intestine, where it may inactivate LPS before it enters systemic circulation. In Fig. 3F, fecal LPS is significantly higher in Aoah⁻/⁻ mice compared to Aoah⁺/⁺ mice, indicating that AOAH in the intestine reduces bioactive LPS levels at the source. This implies that differences in hepatic LPS levels are already influenced by the gut environment, raising doubts about how much Kupffer cells contribute to inactivating LPS in the liver.

      (2) The reliance on Kupffer cell depletion with clodronate-liposomes may overestimate the role of Kupffer cells because clodronate does not exclusively target hepatic Kupffer cells. Clodronate liposomes are taken up by macrophages systemically, potentially depleting macrophages in other organs, including the intestine and circulation. This means observed effects could also be due to loss of AOAH activity in non-hepatic macrophages.

    4. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      Lu et. al. proposed here a direct role of LPS in inducing hepatic fat accumulation and that the metabolism of LPS therefore can mitigate fatty liver injury. With an Acyloxyacyl hydrolase whole-body KO mice, they demonstrated that Acyloxyacyl hydrolase deletion resulted in higher hepatic fat accumulation over 8 months of high glucose/high fructose diet. Previous literature has found that hepatocyte TLR4 (which is a main receptor for binding LPS) KO reduced fatty liver in the MAFLD model, and this paper complements this by showing that degradation/metabolism of LPS can also reduce fatty liver. This result proposed a very interesting mechanism and the translational implications of utilizing Acyloxyacyl hydrolase to decrease LPS exposure are intriguing.

      The strengths of the present study include that they raised a very simplistic mechanism with LPS that is of interest in many diseases. The phenotype shown in the study is strong. The mechanism proposed by the findings is generally well supported.

      There are also several shortcomings in the findings of this study. As AOAH is a whole-body KO, the source production of AOAH in MAFLD is unclear. Although the authors used published single-cell RNA-seq data and flow-isolated liver cells, physiologically LPS degradation could occur in the blood or the liver. The authors linked LPS to hepatocyte fatty acid oxidation via SREBP1. The mechanism is not explored in great depth. Is this signaling TLR4? In this model, LPS could activate macrophages and mediate the worsening of hepatocyte fatty liver injury via the paracrine effect instead of directly signaling to hepatocytes, thus it is not clear that this is a strictly hepatocyte LPS effect. It would also be very interesting to see if administration of the AOAH enzyme orally could mitigate MAFLD injury. Overall, this work will add to the current understanding of the gut-liver axis and development of MAFLD and will be of interest to many readers.

      We thank the reviewers for their important questions and comments.

      In previous studies we found that AOAH is expressed in Kupffer cells and dendritic cells cells (Shao et al., 2007). Single-cell RNAseq analysis of mouse livers by others has found AOAH in Kupffer cells, monocytes, NK cells and ILC1 cells (Remmerie et al.,2020). We also analyzed human liver single-cell RNAseq data and found that AOAH is expressed in monocytes, macrophages, resident and circulating NK cells, and some T cells (Ramachandran et al., 2019) (Please see new Figure 3E). Using clodronate-liposomes to deplete Kupffer cells we found that hepatic AOAH mRNA diminished and nSREBP1 increased (Please see new Figure 5D). These results suggest that Kupffer cells are the major source of AOAH in the liver and that LPS needs to be inactivated in the liver to prevent hepatocyte lipid accumulation.

      Using primary hepatocyte culture, we found that LPS can stimulate hepatocytes directly to induce mTOR activation and SREBP1 activation (new Figure 6E). Adding purified Kupffer cells to the hepatocyte culture did not further increase SREBP1 activation. These results suggest that LPS may directly stimulate hepatocyte to accumulate fat, at least in vitro.

      Both TLR4 and caspase 11 are reported to play important roles in MASLD development (Sharifnia et al., 2015; Zhu et al., 2021). We have crossed Aoah<sup>-/-</sup> mice with TLR4<sup>-/-</sup> mice and found that Aoah<sup>-/-</sup>TLR4<sup>-/-</sup> and Aoah<sup>-/-</sup> mice had similarly severe MASLD. This is probably because TLR4 is required for gut homeostasis (Rakoff-Nahoum et al., 2004); in TLR4 whole-body KO mice compromised gut homeostasis may result in more severe MASLD. By specifically deleting TLR4 on hepatocytes, Yu et al found that NASH-induced fibrosis was mitigated (Yu et al., 2021). In future studies we therefore would need to specifically delete TLR4 in hepatocytes to test whether excessive gut-derived LPS in Aoah<sup>-/-</sup> mice stimulates hepatic TLR4 to induce more severe MASLD. We would also test whether Caspase 11 is required for hepatic fat accumulation in Aoah<sup>-/-</sup> mice.

      It is intriguing to test whether providing exogenous AOAH may mitigate MASLD. We will use an AAV expressing AOAH to test this idea.

      Reviewer #2 (Public review):

      The authors of this article investigated the impact of the host enzyme AOAH on the progression of MASLD in mice. To achieve this, they utilized whole-body Aoah<sup>-/-</sup> mice. The authors demonstrated that AOAH reduced LPS-induced lipid accumulation in the liver, probably by decreasing the expression and activation of SREBP1. In addition, AOAH reduced hepatic inflammation and minimized tissue damage.

      However, this paper is descriptive without a clear mechanistic study. Another major limitation is the use of whole-body KO mice so the cellular source of the enzyme remains undefined. Moreover, since LPS-mediated SREBP1 regulation or LPS-mediated MASLD progression is already documented, the role of AOAH in SREBP1-dependent lipid accumulation and MASLD progression is largely expected.

      Specific comments:

      (1) The overall human relevance of the current study remains unclear.

      It is a good point. We have studied human relevance and show the results in Figure 3E. AOAH expression increased in the hepatic macrophages and monocytes of MASLD patients.

      (2) Is AOAH secreted from macrophages or other immune cells? Are there any other functions of AOAH within the cells?

      AOAH can be secreted from kidney proximal tubule cells and the released AOAH can be taken up by cells that do not express AOAH (Feulner et al., 2004). AOAH can also deacylate oxidized phospholipids, DAMP molecules (Zou et al., 2021).

      (3) Due to using whole-body KO mice, the role of AOAH in specific cell types was unclear in this study, which is one of the major limitations of this study. The authors should at least conduct in vitro experiments using a co-culture system of hepatocytes and Kupffer cells (or other immune cells) isolated from WT or Aoah<sup>-/-</sup> mice.

      Thanks for the suggestion.

      Using clodronate-liposomes, we depleted Kupffer cells and found that hepatic AOAH mRNA diminished and nSREBP1 increased in the liver (Please see new Figure 5D). These results confirm that Kupffer cells are the major source of AOAH in the liver and LPS needs to be inactivated in the liver to prevent hepatocyte lipid accumulation.  Using primary hepatocyte culture, we found that LPS can stimulate hepatocytes directly to induce mTOR activation and SREBP1 activation (new Figure 6E).  These results suggest that LPS may directly stimulate hepatocytes to accumulate fat, at least in vitro.

      (4) It has been well-known that intestinal tight junction permeability is increased by LPS or inflammatory cytokines. However, in Figure 3E, intestinal permeability is comparable between the groups in both diet groups. The authors should discuss more about this result. In addition, intestinal junctional protein should be determined by Western blot and IHC (or IF) to further confirm this finding.

      We have stained ZO-1 (Please see Author response image 1, ZO-1- green fluorescence) in Aoah<sup>+/+</sup> and Aoah<sup>-/-</sup> mouse colonic sections. We did not see a big difference between the two strains of mice.

      Author response image 1.

      Feeding a high fat diet in our mouse facility for 28 weeks has led to increased gut permeability, but there was no difference between Aoah<sup>+/+</sup> and Aoah<sup>-/-</sup>mice. Thus, the more severe MASLD in Aoah<sup>-/-</sup> mice is mainly caused by elevated bioactive LPS instead of increased LPS translocation from the intestine to the liver.

      (5) In Figure 6, the LPS i.g. Aoah<sup>-/-</sup> group is missing. This group should be included to better interpret the results.

      Please see new Figure 6. When we orally gavaged Aoah<sup>-/-</sup> mice with LPS, fecal LPS levels did not increase further. Their liver SREBP1 did not increase further while the SREBP1 target gene expression increased when compared with Aoah<sup>-/-</sup> mice i.g. PBS.

      (6) The term NAFLD has been suggested to be changed to MASLD as the novel nomenclature according to the guidelines of AASLD and EASL.

      Thanks for the suggestion. We have changed NAFLD to MASLD.

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      Consider using MAFLD rather than NAFLD.

      Thanks for the suggestion. We have changed NAFLD to MASLD.

      References

      Feulner, J.A., M. Lu, J.M. Shelton, M. Zhang, J.A. Richardson, and R.S. Munford. 2004. Identification of acyloxyacyl hydrolase, a lipopolysaccharide-detoxifying enzyme, in the murine urinary tract. Infection and immunity 72:3171-3178.

      Zou, B., M. Goodwin, D. Saleem, W. Jiang, J. Tang, Y. Chu, R.S. Munford, and M. Lu. 2021. A highly conserved host lipase deacylates oxidized phospholipids and ameliorates acute lung injury in mice. eLife 10:

    1. eLife Assessment

      This valuable work investigates the social interactions of mice living together in a system of multiple connected cages. It provides solid evidence for a statistical approach capturing changes in social interactions after manipulating prefrontal cortical plasticity. This research will be of broad interest to researchers studying animal social behavior.

    2. Reviewer #3 (Public review):

      Summary:

      Chen et al. present a thorough statistical analysis of social interactions, more precisely, co-occupying the same chamber in the Eco-HAB measurement system. They also test the effect of manipulating the prelimbic cortex by using TIMP-1 that inhibits the MMP-9 matrix metalloproteinase. They conclude that altering neural plasticity in the prelimbic cortex does not eliminate social interactions, but it strongly impacts social information transmission.

      Strengths:

      The quantitative approach to analyzing social interactions is laudable and the study is interesting. It demonstrates that the Eco-HAB can be used for high throughput, standardized and automated tests of the effects of brain manipulations on social structure in large groups of mice.

      Weaknesses:

      A demonstration of TIMP-1 impairing neural plasticity specifically in the prelimbic cortex of the treated animals would greatly strengthen the biological conclusions. The Eco-HAB provides coarser spatial information compared to some other approaches, which may influence the conclusions.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Public Reviews:

      Reviewer #2 (Public review):

      The authors have constructively responded to previous referee comments and I believe that the manuscript is a useful addition to the literature. I particularly appreciate the quantitative approach to social behavior, but have two cautionary comments.

      (1) Conceptually it is important to further justify why this particular maximum entropy model is appropriate. Maximum entropy models have been applied across a dizzying array of biological systems, including genes, neurons, the immune system, as well as animal behavior, so would seem quite beneficial to explain the particular benefits here, for mouse social behavior as coarse-grained through the eco-hab chamber occupancy. This would be an excellent chance to amplify what the models can offer for biological understanding, particularly in the realm of social behavior

      We thank the reviewer for this comment. Maximum entropy models, along with other statistical inference methods that learn interaction patterns from simultaneously-measured degrees of freedom, help distinguish various types of interactions, e.g. direct vs. indirect interactions among animals, individual preference to food vs. social interaction with pairs. As research on social behavior expands from focusing on pairs of animals to studying groups in (semi-)naturalistic environments, maximum entropy models serve as a crucial link between high-throughput data and the need to identify and distinguish interaction rules. Specifically, among all possible maximum entropy models, the pairwise maximum entropy model is one of the simplest that can describe interactions among individuals, which serves as an excellent starting point to understand collective and social behavior in animals.

      Although the Eco-HAB setup currently records spatially coarse-grained data, it still provides more spatial information compared to the traditional three-chamber tests used to assess sociability for rodents. By showing that the maximum entropy model can effectively analyze Eco-HAB data, we hope to highlight its potential in research of social behavior in animals.

      To amplify what the models can offer for biological understanding particularly in the realm of social behavior, We have updated the Introduction to add a more logical structure to the need of using maximum entropy models to identify interactions among mice. Additionally, we updated the first paragraph of the Discussion to make it specific that it is the use of maximum entropy models that identifies interaction patterns from the high-throughput data. Finally, we have also added in the Discussion (line 422-425) arguments supporting the specific use of pairwise maximum entropy models to study social behaviors.

      (2) Maximum entropy models of even intermediate size systems involve a large number of parameters. The authors are transparent about that limitation here, but I still worry that the conclusion of the sufficiency of pairwise interactions is simply not general, and this may also relate to the differences from previous work. If, as the authors suggest in the discussion, this difference is one of a choice of variables, then that point could be emphasized. The suggestion of a follow up study with a smaller number of mice is excellent.

      We thank the reviewer for raising the issue and agree that the caveat of how general pairwise interactions can describe social behavior of animals needs to be discussed. We have added a sentence in the Discussion to point out this important caveat. “More generally, this discrepancy when looking at different choices of variables raises the issue that when studying social behavior of animals in a group, it is important to test and compare interaction models with different complexity (e.g. pairwise or with higher-order interactions).” We have also toned down our conclusion to limit our results of pairwise interactions describing mice co-localization patterns to the data collected in Eco-HAB (also see Reviewer 3 Major Point 2).

      Reviewer #3 (Public review):

      Summary:

      Chen et al. present a thorough statistical analysis of social interactions, more precisely, co-occupying the same chamber in the Eco-HAB measurement system. They also test the effect of manipulating the prelimbic cortex by using TIMP-1 that inhibits the MMP-9 matrix metalloproteinase. They conclude that altering neural plasticity in the prelimbic cortex does not eliminate social interactions, but it strongly impacts social information transmission.

      Strengths:

      The quantitative approach to analyzing social interactions is laudable and the study is interesting. It demonstrates that the Eco-HAB can be used for high throughput, standardized and automated tests of the effects of brain manipulations on social structure in large groups of mice.

      Weaknesses:

      A demonstration of TIMP-1 impairing neural plasticity specifically in the prelimbic cortex of the treated animals would greatly strengthen the biological conclusions. The Eco-HAB provides coarser spatial information compared to some other approaches, which may influence the conclusions.

      Recommendations for the authors:  

      Reviewer #3 (Recommendations for the authors):

      Major points

      (1) Do the Authors have evidence that TIMP-1 was effective, as well as specific to the prelimbic cortex?

      We refer to the literature for the effectiveness and specificity of TIMP-1 to the prelimbic cortex.

      Specifically, the study by Okulski et al. (Biol. Psychiatry 2007) provides clear evidence that TIMP1 plays a role in synaptic plasticity in the prefrontal cortex. They showed that TIMP-1 is induced in the medial prefrontal cortex (mPFC) following stimulation that triggers late long-term potentiation (LTP), a key model of synaptic plasticity. Overexpression of TIMP-1 in the mPFC blocked the activity of matrix metalloproteinases (MMPs) and prevented the induction of late LTP in vivo. Similar effects were observed with pharmacological inhibition of MMP-9 in vitro, reinforcing the idea that TIMP-1 regulates extracellular proteolysis as part of the plasticity mechanism in the prefrontal cortex. These findings confirm that TIMP-1 is both effective and active in this specific brain region.

      Further evidence comes from Puścian et al. (Mol. Psychiatry 2022), who used TIMP-1-loaded nanoparticles to influence neuronal plasticity in the amygdala. They found that TIMP-1 affected MMP expression, LTP, and dendritic morphology, showing its impact on synaptic modifications. More directly relevant, Winiarski et al. (Sci. Adv. 2025) demonstrated that injecting TIMP-1-loaded nanoparticles into the prelimbic cortex altered responses to social stimuli, further supporting the idea that TIMP-1 has region-specific effects on behavioral processes.

      We have also updated the main text (page 8, 1st paragraph of “Effect of impairing neuronal plasticity in the PL on subterritory preferences and sociability”) of the manuscript to include the above references.

      (2) The Authors seem to suggest that one main reason for the different results compared to Shemesh et al. 2013 was the coarseness of the Eco-HAB data. In this case, I think this conclusion should be toned down because of this significant caveat.

      We thank the reviewer for pointing this out, and agree that this caveat and difference should be emphasized. To tone down the conclusion, we have

      (1) added details about the Eco-HAB (it being coarse-grained, etc.) in the abstract to tone down the conclusion.

      (2) added to the results summary in the Discussion (top of page 12) that the results are “within in the setup of the semi-naturalistic Eco-HAB experiments”

      (3) added to the Discussion (page 13) that the different results compared to Shemesh et al 2013 means that general studies of social behavior need to compare models with different levels of complexity (e.g. pairwise vs. higher-order interactions). (Also see Reviewer 2 Comment 2.)

      Minor points

      (1) Please explain what is measured in Fig. 1C (what is on the y axis?).

      Figure 1C shows the activity of the mice as measured by the rate of transitions, i.e. the number of times the mice switch boxes during each hour of the day, averaged over all N = 15 mice and T = 10 days (cohort M1). The error bars represent variability of activities across individuals or across days. For mouse-to-mouse variability (blue), we first compute for each mouse its number of transitions averaged over the same hour for all 10 days, then we compute its standard deviation across all 15 mice and plot it as error bars. For day-to-day variability (orange), we first compute for each day the number of transitions for each hour averaged over all mice, then compute its standard deviation across all 10 days as the errorbar. We have added the detailed explanation in the caption of Figure 1C.

      (2) In Fig. 3, it would be better to present the control group also in the main figure instead of the supplementary.

      We have merged Figure 3 and Figure 3 Supplementary 1 to present the control group also in the main figure.

      (3) In Fig. 3 and corresponding supplements, there seems to be a large difference between males and females. I think this would deserve some more discussion.

      While not being the main focus of this paper, we agree with the reviewer that the difference between male and female is important and deserves attention in the discussion and also future study. Thus we have added a paragraph in the Discussion (line 394-399, bottom of page 12).

    1. eLife Assessment

      This manuscript provides potentially important findings examining in 2D and 3D models in MYC liver cancer cells changes in DNA repair genes and programs in response to hypoxia. The authors use convincing methodology in most cases, but there is some concern that the analysis is incomplete.

    2. Reviewer #1 (Public review):

      Summary:

      In this report, the authors made use of a murine cell line derived from a MYC-driven liver cancer to investigate the gene expression changes that accompany the switch from normoxic to hypoxia conditions during 2D growth and the switch from 2D monolayer to 3D organoid growth under normoxic conditions. They find a significant (ca. 40-50%) overlap among the genes that are dysregulated in response to hypoxia in 2D cultures and in response to spheroid formation. Unsurprisingly, hypoxia-related genes were among the most prominently deregulated under both sets of conditions. Many other pathways pertaining to metabolism, splicing, mitochondrial electron transport chain structure and function, DNA damage recognition/repair and lipid biosynthesis were also identified.

      Comments on the revised manuscript:

      In my original review of this manuscript, I raised 11 points that I thought needed to be addressed and/or clarified by the authors. In response, they have provided an adequate answer to only one of these (point 6), which is little more than a more thorough description of how spheroids were generated. The remaining points that I raised, which would have provided more mechanistic insight into their study were addressed by the authors with the following such comments:

      - It is not the focus of this study (Points 1 and 4)

      - It is worthy of further validation (Point 2)

      - We apologize for not being able to validate everything (Point 3)

      - This reviewer has raised an interesting question. We are investigating this hypothesis and hopefully we can give a clear answer in the future (Point 5)

      - This is an excellent idea that we certainly will do it in our future experiments (Point 7)

      As to responses that the authors made to the other two reviewers' comments: Most pertained to cosmetic alterations involving clarification of methods, inclusion of a new figure or rearrangement of old figures. These were generally answered. However, in response to the last point raised by Rev. 3 to compare "sgRNA abundances at the earliest harvesting time with the distribution in the library...to see whether and to what extent selection has already taken place before the three culture conditions were established", the authors responded with the comment: "This is great point. Unfortunately, we did not perform such an analysis."

      I understand that it is often impossible to address all points raised by the reviewers. This can be for a variety of reasons and many times the omissions can be overlooked and accepted if the reviewer can be convinced that a good faith attempt has otherwise been made to address the other deficiencies. However, no such effort has been made here and the study remains deficient and largely descriptive as I pointed out in my original review.

    3. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      In this report, the authors made use of a murine cell life derived from a MYC-driven liver cancer to investigate the gene expression changes that accompany the switch from normoxic to hypoxia conditions during 2D growth and the switch from 2D monolayer to 3D organoid growth under normoxic conditions. They find a significant (ca. 40-50%) overlap among the genes that are dysregulated in response to hypoxia in 2D cultures and in response to spheroid formation. Unsurprisingly, hypoxia-related genes were among the most prominently deregulated under both sets of conditions. Many other pathways pertaining to metabolism, splicing, mitochondrial electron transport chain structure and function, DNA damage recognition/repair, and lipid biosynthesis were also identified.

      We thank this reviewer for his/her time and efforts, and the insightful comments.

      Major comments:

      (1) Lines 239-240: The authors state that genes involved in DNA repair were identified as being necessary to maintain survival of both 2D and 3D cultures (Figure S6A). Hypoxia is a strong inducer of ROS. Thus, the ROS-specific DNA damage/recognition/repair pathways might be particularly important. The authors should look more carefully at the various subgroups of the many genes that are involved in DNA repair. They should also obtain at least a qualitative assessment of ROS and ROS-mediated DNA damage by staining for total and mitochondrial-specific ROS using dyes such as CM-H2-DCFDA and MitoSox. Actual direct oxidative damage could be assessed by immunostaining for 8-oxo-dG and related to the sub-types of DNA damage-repair genes that are induced. The centrality of DNA damage genes also raises the question as to whether the previously noted prominence of the TP53 pathway (see point 5 below) might represent a response to ROS-induced DNA damage.

      We thank this reviewer for the insightful comments, and agreed that ROS induced by hypoxia could play a role in modulating DNA repair and consequently cellular essentiality. Although pathway enrichment in Figure S6A (now as Figure 2-figure supplement 4A) showed that DNA repair pathway was essential to cell survival in hypoxia and 3D cultures, the genes associated with this pathway (Ddb1;Brf2;Gtf3c5;Guk1;Taf6) are not typical DNA repair genes. They are more likely involved in gene transcription. However, it will be interesting to see if they are specifically involved in DNA damage in response to ROS, which is out of focus of this study.

      (2) Because most of the pathway differences that distinguish the various cell states from one another are described only in terms of their transcriptome variations, it is not always possible to understand what the functional consequences of these changes actually are. For example, the authors report that hypoxia alters the expression of genes involved in PDH regulation but this is quite vague and not backed up with any functional or empirical analyses. PDH activity is complex and regulated primarily via phosphorylation/dephosphorylation (usually mediated by PDK1 and PDP2, respectively), which in turn are regulated by prevailing levels of ATP and ADP. Functionally, one might expect that hypoxia would lead to the down-regulation of PDH activity (i.e. increased PDH-pSer392) as respiration changes from oxidative to non-oxidative. This would not be appreciated simply by looking at PDH transcript levels. This notion could be tested by looking at total and phospho-PDH by western blotting and/or by measuring actual PDH activity as it converts pyruvate to AcCoA.

      We agreed with this reviewer that PDH activity regulation could be affected by multi-factors, and it is worthy of further validation by other approaches.

      (3) Line 439: Related to the above point: the authors state: "It is likely that blockade of acetyl-CoA production by PDH knockout may force cells to use alternative energy sources under hypoxic and 3D conditions, averting the Warburg effect and promoting cell survival under limited oxygen and nutrient availability in 3D spheroids." This could easily be tested by determining whether exogenous fatty acids are more readily oxidized by hypoxic 2D cultures or spheroids than occurs in normoxic 2D cultures.

      We thank for this suggestion. We apologized for not being able to validate everything.

      (4) Line 472: "Hypoxia induces high expression of Acaca and Fasn in NEJF10 cells indicating that hypoxia promotes saturated fatty acid synthesis...The beneficial effect of Fasn and Acaca KO to NEJF10 under hypoxia is probably due to reduction of saturated fatty acid synthesis, and this hypothesis needs to be tested in the future.". As with the preceding comment, this supposition could readily be supported directly by, for example, performing westerns blots for these enzymes and by showing that incubation of hypoxic 2D cells or spheroids converted more AcCoA into lipid.

      We thank for this suggestion. However, functional validation for the Fasn and Acaca KO is out of focus in this study.

      (5) In Supplementary Figure 2B&C, the central hub of the 2D normoxic cultures is Myc (as it should well be) whereas, in the normoxic 3D, the central hub is TP53 and Myc is not even present. The authors should comment on this. One would assume that Myc levels should still be quite high given that Myc is driven by an exogenous promoter. Does the centrality of TP53 indicate that the cells within the spheroids are growtharrested, being subjected to DNA damage and/or undergoing apoptosis?

      The predicted transcription factor activity analysis was based on the differential ATAC-seq peaks among different culture through pairwise comparisons. If TP53 and MYC were not present under that condition, it did not mean their activity was absent.

      “…the centrality of TP53 indicate that the cells within the spheroids are growth-arrested, being subjected to DNA damage and/or undergoing apoptosis?” This reviewer has raised an interesting question. We are investigating this hypothesis and hopefully we can give a clear answer in the future.

      (6) In the Materials and Methods section (lines 711-720), the description of how spheroid formation was achieved is unclear. Why were the cells first plated into non-adherent 96 well plates and then into nonadherent T75 flasks? Did the authors actually utilize and expand the cells from 144 T75 flasks and did the cells continue to proliferate after forming spheroids? Many cancer cell types will initially form monolayers when plated onto non-adherent surfaces such as plastic Petri dishes and will form spheroid-like structures only after several days. Other cells will only aggregate on the "non-adherent" surface and form spheroid-like structures but will not actually detach from the plate's surface. Have the authors actually documented the formation of true, non-adherent spheroids at 2 days and did they retain uniform size and shape throughout the collection period? The single photo in Supplementary Figure 1 does not explain when this was taken. The authors include a schematic in Figure 2A of the various conditions that were studied. A similar cartoon should be included to better explain precisely how the spheroids were generated and clarify the rationale for 96 well plating. Overall, a clearer and more concise description of how spheroids were actually generated and their appearance at different stages of formation needs to be provided.

      The cells were initially plated in non-adherent 96-well plates to facilitate the formation of spheroids in a controlled and uniform manner. As correctly mentioned by the reviewer, during the initial stages, cells cultured on non-adherent surfaces often form aggregates or clumps, and it takes a few days for them to develop into solid spheroids.

      In our study, we aimed to achieve 3D spheroid formation immediately following the transduction process to allow for screening under both 2D and 3D conditions. Plating the cells into 96-well plates enabled us to monitor and control the formation of spheroids in smaller volumes before scaling up the culture in non-adherent T75 flasks for subsequent experimental steps. This setup allows us to maintain gene editing processes under both 2D and 3D conditions.

      Regarding the proliferation and uniformity of spheroids:

      • Yes, the spheroids continued to proliferate after their formation.

      • True, non-adherent spheroids were documented as early as the next day. This was visually confirmed under microscopy, and size uniformity was maintained throughout the collection period by following optimized culture protocols.

      We also agreed with the reviewer’s suggestion to include a cartoon schematic similar to Figure 2A, illustrating the spheroid generation process and clarifying the rationale for using 96-well plates. We have included such a cartoon and speroid growth curve monitored by Incucyte as Figure 2-figure supplement 2.

      (7) The authors maintained 2D cultures in either normoxic or hypoxic (1% O2) states during the course of their experiments. On the other hand, 3D cultures were maintained under normoxic conditions, with the assumption that the interiors of the spheroids resemble the hypoxic interiors of tumors. However, the actual documentation of intra-spheroid hypoxia is never presented. It would be a good idea for the authors to compare the degree of hypoxia achieved by 2D (1% O2) and 3D cultures by staining with a hypoxia-detecting dye such as Image-iT Green. Comparing the fluorescence intensities in 2D cultures at various O2 concentrations might even allow for the construction of a "standard curve" that could serve to approximate the actual internal O2 concentration of spheroids. This would allow the authors to correlate the relative levels of hypoxia between 2D and 3D cultures.

      This is an excellent idea that we certainly will do it in our future experiments.

      (8) Related to the previous 2 points, the authors performed RNAseq on spheroids only 48 hours after initiating 3D growth. I am concerned that this might not have been a sufficiently long enough time for the cells to respond fully to their hypoxic state, especially given my concerns in Point 6. Might the results have been even more robust had the authors waited longer to perform RNA seq? Why was this short time used?

      We agreed with this reviewer. We were unsure if 48hours was an ideal timepoint. It might be necessary to perform a longitudinal experiment to harvest samples under different timepoints in the future experiments.

      (9) What happens to the gene expression pattern if spheroids are re-plated into standard tissue culture plates after having been maintained as spheroids? Do they resume 2D growth and does the gene expression pattern change back?

      This is a great question and we have never thought about what the gene expression pattern would be if speroids are re-plated in 2D. This could be a challenging experiment because the gene expression and epigenetic changes are timing related. However, the cells do grow well after re-plated in 2D.

      (10) Overall, the paper is quite descriptive in that it lists many gene sets that are altered in response to hypoxia and the formation of spheroids without really delving into the actual functional implications and/or prioritizing the sets. Some of these genes are shown by CRISPR screening to be essential for maintaining viability although in very few cases are these findings ever translated into functional studies (for example, see points 14 above). The list of genes and gene pathways could benefit from a better explanation and prioritization of which gene sets the authors believe to be most important for survival in response to hypoxia and for spheroid formation.

      This was a genome-wide study that integrated RNA-seq, ATAC-seq and CRISPR KO, providing resource to understand the oncogenic pathways in different culture conditions. We believe we have clearly articulated the important genes/pathways in our abstract.

      (11) The authors used a single MYC-driven tumor cell line for their studies. However, in their original paper (Fang, et al. Nat Commun 2023, 14: 4003.) numerous independent cell lines were described. It would help to know whether RNAseq studies performed on several other similar cell lines gave similar results in terms of up & down-regulated transcripts (i.e. representative of the other cell lines are NEJF10 cells).

      We have not generated RNA-seq data for these cell lines cultured in different conditions.

      Reviewer #2 (Public review):

      Summary:

      The manuscript by Fang et al., provides a tour-de-force study uncovering cancer cell's varied dependencies on several gene programs for their survival under different biological contexts. The authors addressed genomic differences in 2D vs 3D cultures and how hypoxia affects gene expression. They used a Myc-driven murine liver cancer model grown in 2D monolayer culture in normoxia and hypoxia as well as cells grown as 3D spheroids and performed CRISPR-based genome-wide KO screen to identify genes that play important roles in cell fitness. Some context-specific gene effects were further validated by in-vitro and in-vivo gene KO experiments.

      Strengths:

      The key findings in this manuscript are:

      (1) Close to 50% of differentially expressed genes were common between 2D Hypoxia and 3D spheroids conditions but they had differences in chromatin accessibility.

      (2) VHL-HIF1a pathway had differential cell fitness outcomes under 2D normoxia vs 2D hypoxia and 3D spheroids.

      (3) Individual components of the mitochondrial respiratory chain complex had contrasting effects on cell fitness under hypoxia.

      (4) Knockout of organogenesis or developmental pathway genes led to better cell growth specifically in the context of 3D spheroids and knockout of epigenetic modifiers had varied effects between 2D and 3D conditions.

      (5) Another key program that leads to cells fitness outcomes in normoxia vs hypoxia is the lipid and fatty acid metabolism.

      (6) Prmt5 is a key essential gene under all growth conditions, but in the context of 3D spheroids even partial loss of Prmt5 has a synthetic lethal effect with Mtap deletion and Mtap is epigenetically silenced specifically in the 3D spheroids.

      We appreciate this reviewer for acknowledging the strengths of our study.

      Issues to address:

      (1) The authors should clarify the link between the findings of the enrichment of TGFb-SMAD signaling REACTOME pathway to the findings that knocking out TGFb-SMAD pathway leads to better cell fitness outcomes for cells in the 3D growth conditions.

      We have clarified this link in abstract by saying “Notably, multicellular organogenesis signaling pathways including TGFb-SMAD, which is upregulated in 3D culture, specifically constrict the uncontrolled cell proliferation in 3D while inactivation of epigenetic modifiers (Bcor, Kmt2d, Mettl3 and Mettl14) has opposite outcomes in 2D vs. 3D:

      (2) Supplementary Figure 4C has been cited in the text but doesn't exist in the supplementary figures section.

      Sorry for this typo. It should be 5C which is Figure 2-figure supplement 3C in the new version of MS. We have corrected it now.

      (3) A small figure explaining this ABC-Myc driven liver cancer model in Supplementary Figure 1 would be helpful to provide context.

      We appreciate this suggestion. We have added a cartoon as Figure 1-figure supplement 1A to indicate the procedure for generation of this model.

      (4) The method for spheroids formation is not found in the method section.

      We described the method in our previous publication (Nature Communications 2023 Jul 6;14(1):4003.). However, we have added the information in method now, and the procedure is very simple (line 623-624). We found the murine liver cancer cell lines can readily form spheroids when they are cultured in low-attachment dish with standard DMEM complete media.

      (5) In Supplementary Figure 1b, the comparisons should be stated the opposite way - 3D vs 2D normoxia and 2D-Hypoxia vs 2D-Normoxia.

      We have made correction in the Figure legend of Figure S1B which is Figure 1B now in the new version of MS.

      (6) There are typos in the legend for Supplementary Figure 10.

      We have checked the typos.

      (7) Consider putting Supplementary Figure 1b into the main Figure 1.

      We have moved both Supplementary Figure 1a and 1b into main Figure 1 as Figure 1A and 1B. Hopefully, this will help the readers to catch the information easily.

      (8) Please explain only one timepoint (endpoint) for 3D spheroids was performed for the CRISPR KO screen experiment, while several timepoints were done for 2D conditions? Was this for technical convenience?

      As this reviewer speculated, indeed this was for technical convenience. We found that it was technically challenging to split the spheroids for CRISPR screening.

      (9) In line 372, it is indicated that Bcor KO (Fig 5e) had growth advantage - this was observed in only one of the gRNA -- same with Kmt2d KO in the same figure where there was an opposite effect. Please justify the use of only one gRNA.

      We actually used 4 gRNAs for each gene. In the heatmap, although one of the gRNA for each gene showed some levels of enrichment under hypoxic 2D condition, they were all highly enriched in 3D.

      (10) Why was CRISPR based KO strategy not used for the PRMT5 gene but rather than the use of shRNA.? Note that one of the shRNA for PRMT5 had almost no KO (PRMT5-shRNA2 Figure 7B) but still showed phenotype (Figure 7D) - please explain.

      We used shRNA as second approach for cross-validation. We agreed that the knockdown efficiency of shRNA2 was not as good as the others, with only about 40% knockdown efficiency.

      (11) In Figure 7D, which samples (which shRNA group) were being compared to do the t-test?

      The comparisons were for shCtrl and each of the shPRMT5. We have clarified this in figure legend.

      (12) In line 240, it is stated that oxphos gene set is essential for NEJF10 cell survival in both normoxia and hypoxia conditions. But shouldn't oxphos be non-essential in hypoxia as cells move away from oxphos and become glycolytic?

      This is a great question. While indeed hypoxia may promote the switch from oxphos to glycolysis, several studies showed that the low oxygen concentrations in hypoxic regions of tumors may not be limiting for oxphos, and ATP is generated by oxphos in tumors even at very low oxygen tensions (please see review Clin Cancer Res (2018) 24 (11): 2482–2490.). We therefore speculated that NEJF10 cells were still dependent on oxphos for ATP production under hypoxia. However, this needs further investigation. We have added this discussion in our manuscript (line 250-254).

      (13) In line 485 it is mentioned that Pmvk and Mvd genes which are involved in cholesterol synthesis when knocked out had a positive effect on cell growth in 3D conditions and since cholesterol synthesis is essential for cell growth how does this not matter much in the context of 3D - please explain.

      We thank this reviewer for this note. It seemed that only two gRNA for each were upregulated in 3D and it could be due to technical issue or clonal selection. We have deleted this sentence in our new version of MS.

      Reviewer #3 (Public review):

      Summary:

      In this study, Fang et al. systematically investigate the effects of culture conditions on gene expression, genome architecture, and gene dependency. To do this, they cultivate the murine HCC line NEJF10 under standard culture conditions (2D), then under similar conditions but under hypoxia (1% oxygen, 2D hypoxia) and under normoxia as spheroids (3D). NEJF10 was isolated from a marine HCC model that relies exclusively on MYC as a driver oncogene. In principle, (1) RNA-seq, (2) ATAC-seq and (3) genetic screens were then performed in this isogenic system and the results were systematically compared in the three cultivation methods. In particular, genome-wide screens with the CRISPR library Brie were performed very carefully. For example, in the 2D conditions, many different time points were harvested to control the selection process kinetically. The authors note differential dependencies for metabolic processes (not surprisingly, hypoxia signaling is affected) such as the regulation and activity of mitochondria, but also organogenesis signaling and epigenetic regulation.

      Strengths:

      The topic is interesting and relevant and the experimental set-up is carefully chosen and meaningful. The paper is well written. While the study does not reveal any major surprises, the results represent an important resource for the scientific community.

      We thank this reviewer for his/her positive comments.

      Weaknesses:

      However, this presupposes that the statistical analysis and processing are carried out very carefully, and this is where my main suggestions for revision begin. Firstly, I cannot find any information on the number of replicates in RNA- and ATAC-seq. This should be clearly stated in the results section and figure legends and cut-offs, statistical procedures, p-values, etc. should be mentioned as well. In principle, all NGS experiments (here ATAC- and RNA-seq) should be performed in replicates (at least duplicates, better triplicates) or the results should be validated by RT-PCR in independent biological triplicates. Secondly, the quantification of the analyses shown in the figures and especially in the legends is not sufficiently careful. Units are often not mentioned. Example Figure 4a: The legend says: 'gRNA reads' but how can the read count be -1? I would guess these are FC, log2FC, or Z-values. All figure legends need careful revision.

      Based upon the reviewer’s suggestions, we have added details about the replicates in figure legend. For gRNA read heatmap, the scale bar indicates the Z score. We have added the information in figure legends.

      Furthermore, I would find a comparison of the sgRNA abundances at the earliest harvesting time with the distribution in the library interesting, to see whether and to what extent selection has already taken place before the three culture conditions were established (minor point).

      This is great point. Unfortunately, we did not perform such an analysis.

      Recommendations for the authors:

      Reviewing Editor:

      There are three general issues:

      First, there is a lack of detail regarding much of the analysis. In some cases, this makes it difficult to assess the value of the data, albeit, there is generally a consensus the information is really interesting.

      Second, the findings - although provocative - lack mechanistic details and are focused more on descriptive findings. Hence, the manuscript would be improved by some effort at evaluating identified programs and providing some suggestions of mechanisms.

      Third, the authors need to put much more effort into the clarity and tightness of the presentation.

      We have made clarification in response to the reviewer’s comments.

      Reviewer #1 (Recommendations for the authors):

      Figure S1C. the labeling of the lower x-axis is inverted.

      Due to space limitation, we changed the figure orientation in our old version of MS. We have tilted the figure back in the new version, which is Figure 1-figure supplement 1B now.

    1. eLife Assessment

      Centromeres are specific sites on chromosomes that are essential for mitosis and genome fidelity. This valuable research advance builds upon previous studies to convincingly show that the centromere-histone core contributes to force transduction through the kinetochore. The centromere mainly strengthens one of the two paths of force transduction, influenced by the centromeric DNA sequence. The mechanism underlying this phenomenon will be an exciting future avenue of research, given that centromeric DNAs are not conserved. This work will be of interest to those studying cell division and chromosome segregation.

    2. Reviewer #1 (Public review):

      Summary:

      The authors address the role of the centromere histone core in force transduction by the kinetochore

      Strengths:

      They use a hybrid DNA sequence that combines CDEII and CDEIII as well as Widom 601 so they can make stable histones for biophysical studies (provided by the Widom sequence) and maintain features of the centromere (CDE II and III).

      Weaknesses:

      The main results are shown in one figure (Fig 2). Indeed the Centromere core of Widom and CDE II and III contribute to strengthening the binding force for the OA-beads. The data are very nicely done and convincingly demonstrate the point. The weakness is that this is the entire paper. It is certainly of interest to investigators in kinetochore biology, but beyond that the impact is fairly limited in scope.

      Comments on revisions:

      The additional information provided by the authors will help the reader understand and interpret the manuscript.

    3. Reviewer #2 (Public review):

      Summary:

      This paper provides a valuable addendum to the findings described in Hamilton et al. 2020 (https://doi.org/10.7554/eLife.56582). In the earlier paper, the authors reconstituted the budding yeast centromeric nucleosome together with parts of the budding yeast kinetochore and tested which elements are required and sufficient for force transmission from microtubules to the nucleosome. Although budding yeast centromeres are defined by specific DNA sequences, this earlier paper did not use centromeric DNA but instead the generic Widom 601 DNA. The reason is that it has so far been impossible to stably reconstitute a budding yeast centromeric nucleosome using centromeric DNA.

      In this new study, the authors now report that they were able to replace part of the Widom 601 DNA with centromeric DNA from chromosome 3. This makes the assay more closely resemble the in vivo situation. Interestingly, the presence of the centromeric DNA fragment makes one type of minimal kinetochore assembly, but not the other, withstand stronger forces.

      Which kinetochore assembly turned out to be affected was somewhat unexpected, and can currently not be reconciled with structural knowledge of the budding yeast centromere/kinetochore. This highlights that, despite recent advances (e.g. Guan et al., 2021; Dendooven et al., 2023), aspects of budding yeast kinetochore architecture and function remain to be understood and that it will be important to dissect the contributions of the centromeric DNA sequence.

      In the future, it will be interesting to pinpoint which interactions contribute to the enhanced force resistance in the presence of centromeric DNA.

      Strength:

      - The paper demonstrates that centromeric DNA can increase the attachment strength between budding yeast microtubules and centromeric nucleosomes.

      Weakness:

      - How centromeric DNA exerts this effect remains unclear.

      Comments on revisions:

      I appreciate the authors' detailed response and their decision to list all the tested in chimeras in Table 3.

      All my prior comments have been addressed.

    4. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1:

      Summary:

      The authors address the role of the centromere histone core in force transduction by the kinetochore.

      Strengths:

      They use a hybrid DNA sequence that combines CDEII and CDEIII as well as Widom 601 so they can make stable histones for biophysical studies (provided by the Widom sequence) and maintain features of the centromere (CDE II and III).

      Weaknesses:

      The main results are shown in one figure (Figure 2). Indeed the Centromere core of Widom and CDE II and III contribute to strengthening the binding force for the OA-beads. The data are very nicely done and convincingly demonstrate the point. The weakness is that this is the entire paper. It is certainly of interest to investigators in kinetochore biology, but beyond that, the impact is fairly limited in scope.

      This reviewer might have missed that this is a Research Advance, not an article. Research Advances are limited in scope by definition and provide a new development that builds on research reported in a prior paper. They can be of any length. Our Research Advance builds on our prior work, Hamilton et al., 2020 and provides the new result that native centromere sequences strengthen the attachment of the kinetochore to the nucleosome.

      Reviewer #2:

      Summary:

      This paper provides a valuable addendum to the findings described in Hamilton et al. 2020 (https://doi.org/10.7554/eLife.56582). In the earlier paper, the authors reconstituted the budding yeast centromeric nucleosome together with parts of the budding yeast kinetochore and tested which elements are required and sufficient for force transmission from microtubules to the nucleosome. Although budding yeast centromeres are defined by specific DNA sequences, this earlier paper did not use centromeric DNA but instead the generic Widom 601 DNA. The reason is that it has so far been impossible to stably reconstitute a budding yeast centromeric nucleosome using centromeric DNA.

      In this new study, the authors now report that they were able to replace part of the Widom 601 DNA with centromeric DNA from chromosome 3. This makes the assay more closely resemble the in vivo situation. Interestingly, the presence of the centromeric DNA fragment makes one type of minimal kinetochore assembly, but not the other, withstand stronger forces.

      We thank the reviewer for their careful and positive assessment of our work.

      Which kinetochore assembly turned out to be affected was somewhat unexpected, and can currently not be reconciled with structural knowledge of the budding yeast centromere/kinetochore. This highlights that, despite recent advances (e.g. Guan et al., 2021; Dendooven et al., 2023), aspects of budding yeast kinetochore architecture and function remain to be understood and that it will be important to dissect the contributions of the centromeric DNA sequence.

      We couldn’t agree more.

      Given the unexpected result, the study would become yet more informative if the authors were able to pinpoint which interactions contribute to the enhanced force resistance in the presence of centromeric DNA.

      Strength:

      The paper demonstrates that centromeric DNA can increase the attachment strength between budding yeast microtubules and centromeric nucleosomes.

      Weakness:

      How centromeric DNA exerts this effect remains unclear.

      Recommendations for the authors:

      Reviewer #2 (Recommendations for the authors):

      (1) Additional specific mutants would be helpful in interpreting the effect observed. The authors speculate that a small segment of OA near the DNA (based on Dendooven et al., 2023) could be important. Would it be possible to introduce specific mutations and test this?

      This would be an interesting study but is far beyond the scope of a Research Advance. In fact, it would make a nice thesis project for a new student. Although perhaps not obvious, these studies require a large set of reagents including wrapped nucleosomes, which must be made fresh (they cannot be frozen) and five purified recombinant complexes, purified by specialized protocols that maintain their activity. Moreover, each datapoint is gathered one at a time. For example, the data in Figure 2 in this manuscript includes 343 datapoints acquired one at a time over the course of 1.5 years.  

      (2) Please provide the sequences of the other CEN3-W601 chimeras that were tested and did NOT stably wrap centromeric histone octamers. This may help others to design yet different constructs in the future. (Maybe the information is there and I didn't see it?)

      We fully agree and thank the reviewer for this excellent suggestion. The sequences and summaries of their wrapping stability are now provided in Table 3, page 17.

      (3) I wonder whether the authors tested the C0N3 sequence used in Dendooven et al., 2023. If not, could it be tested? This would more tightly couple the functional assay shown here with the structural work.

      We did not test the CON3 sequence, which was published several years after the start of this work. We agree that a tight coupling between the functional assay and the structural work would be useful. However, we also see the advantage of being able to go beyond the structural work and include even more CEN3 sequence than has so far been possible in the structural work.  

      In addition to measuring the role of DNA sequence in Okp1/Ame1 attachment to the nucleosome, we were interested in the role of DNA sequence in the attachment of Mif2. Therefore, we included all 35 bp of the Mif2 footprint in our chimeric CCEN DNA sequence. CON3 only includes 8 bp from CDEII. We did produce stable nucleosomes using CEN3-601 from Guan et al. (see Table 3). Again, CEN3-601 only includes 8 bp of the Mif2 footprint so we opted to study nucleosomes wrapped in our CCEN DNA with the entire Mif2 footprint. Curiously we found that even the entire Mif2 footprint was not enough to find the DNA sequence specificity seen in the EMSA experiments reported by Xiao et al., 2017.

      To help readers understand the differences between all these constructs, we have included them in Table 3.

      (4) Would an AlphaFold 3 prediction of the assemblies used in this paper be feasible and useful?

      The structures of the Dam1 complex (Jenni et al., 2018), Ndc80 complex (Zahm, et al., 2023 and references therein), MIND complex (Dimitrova et al., 2016), OA complex (Dendooven et al., 2023), and the nucleosome (Xaio et al., 2017; Yan et al., 2019; Guan et al., 2021; Dendooven et al., 2023) are published. The interactions between many of these complexes are understood beyond the level that AlphaFold3 could provide (Dimitrova et al., 2016; Dendooven et al., 2023). One of the main questions is how Mif2 interacts with the nucleosome and the other components of the kinetochore. Even structural analyses that included Mif2 in the assembly detect little or no Mif2 in the final structure. Unfortunately, AlphaFold3 is also not helpful as it predicts only the structure of the dimerization domain, which was already known (Cohen et al., 2008).

      AlphaFold3 predicts the rest of Mif2 is largely unstructured with several alpha helices predicted with low confidence.

      (5) Given that the centromeric DNA piece included should be able to bind the CBF3 complex, would it be possible to add this complex and test the effect on force transmission?

      This would be an interesting experiment, and we do expect CBF3 to bind. As stated above, this is far beyond the scope of this Research Advance. In our experience, with each new kinetochore subcomplex that we add into our reconstitutions, there are new challenges purifying the subcomplex in active form and in sufficient quantity. We are eager to add CBF3 but this is not something we can pull off in the context of this Research Advance. Thank you again for the time and energy spent reviewing our manuscript

    1. eLife Assessment

      The bacterial cell wall is crucial to maintain viability. It has previously been suggested that Gram positive bacteria have a periplasmic region between the cell membrane and peptidoglycan cell wall that this is maintained by the presence of teichoic acids. In this valuable study, Nguyen et al. make clever use of electron microscopy and metabolic labelling to interrogate the role of teichoic acids in supporting the maintenance of the periplasmic region in Streptococcus pneumoniae. The findings are solid and close some crucial knowledge gaps whilst providing novel tools to further interrogate discrepancies in the field. This work will be of broad interest to microbiologists.

    2. Reviewer #1 (Public review):

      The authors set out to analyse the roles of the teichoic acids of Streptococcus pneumoniae in supporting the maintenance of the periplasmic region. Previous work has proposed the periplasm to be present in Gram positive bacteria and here advanced electron microscopy approach was used. This also showed a likely role for both wall and lipo-teichoic acids in maintaining the periplasm. Next, the authors use a metabolic labelling approach to analyse the teichoic acids. This is a clear strength as this method cannot be used for most other well studied organisms. The labelling was coupled with super-resolution microscopy to be able to map the teichoic acids at the subcellular level and a series of gel separation experiments to unravel the nature of the teichoic acids and the contribution of genes previously proposed to be required for their display. The manuscript could be an important addition to the field but there are a number of technical issues which somewhat undermine the conclusions drawn at the moment. These are shown below and should be addressed. More minor points are covered in the private

      Recommendations for Authors.

      Weaknesses to be addressed:

      (1) l. 144 Was there really only one sample that gave this resolution? Biological repeats of all experiments are required.

      (2) Fig. 4A. Is the pellet recovered at "low" speeds not just some of the membrane that would sediment at this speed with or without LTA? Can a control be done using an integral membrane protein and Western Blot? Using the tacL mutant would show the behaviour of membranes alone.

      (3) Fig. 4A. Using enzymatic digestion of the cell wall and then sedimentation will allow cell wall associated proteins (and other material) to become bound to the membranes and potentially effect sedimentation properties. This is what is in fact suggested by the authors (l. 1000, Fig. S6). In order to determine if the sedimentation properties observed are due to an artefact of the lysis conditions a physical breakage of the cells, using a French Press, should be carried out and then membranes purified by differential centrifugation. This is a standard, and well-established method (low-speed to remove debris and high-speed to sediment membranes) that has been used for S. pneumoniae over many years but would seem counter to the results in the current manuscript (for instance Hakenbeck, R. and Kohiyama, M. (1982), Purification of Penicillin-Binding Protein 3 from Streptococcus pneumoniae. European Journal of Biochemistry, 127: 231-236).

      (4) l. 303-305. The authors suggest that the observed LTA-like bands disappear in a pulse chase experiment (Fig. 6B). What is the difference between this and Fig. 5B, where the bands do not disappear? Fig. 5C is the WT and was only pulse labelled for 5 min and so would one not expect the LTA-like bands to disappear as in 6B?

      (5) Fig. 6B, l. 243-269 and l. 398-410. If, as stated, most of the LTA-like bands are actually precursor then how can the quantification of LTA stand as stated in the text? The "Titration of Cellular TA" section should be re-evaluated or removed? If you compare Fig. 6C WT extract incubated at RT and 110oC it seems like a large decrease in amount of material at the higher temperature. Thus, the WT has a lot of precursors in the membrane? This needs to be quantified.

      (6) L. 339-351, Fig. 6A. A single lane on a gel is not very convincing as to the role of LytR. Here, and throughout the manuscript, wherever statements concerning levels of material are made, quantification needs to be done over appropriate numbers of repeats and with densitometry data shown in SI.

      (7) 14. l. 385-391. Contrary to the statement in the text, the zwitterionic TA will have associated counterions that results in net neutrality. It will just have both -ve and +ve counterions in equal amounts (dependent on their valency), which doesn't matter if it is doing the job of balancing osmolarity (rather than charge).

      Comments on revisions:

      The resubmitted manuscript now contains new data and changes to the text.

      The authors have largely covered my previous points in both sets of reviews (Public/Recommendations).

      Public Review Points:

      1 & 6: I still do not see a reproducibility statement as such, with details of the number of biological repeats etc.

      2 & 3. Fig S7 seems to be quite telling. As predicted after physical breakage the membrane proteins sediment at high speed (rather than low speed). This presumably also means that the LTA comes down at high and not low speed. LTA was not measured due to cost of reagents. The Microfluidizer breaks the cells using a shear force and thus is unlikely to create very small membrane fragments. Thus, the sedimentation properties of membranes containing LTA are likely dependent on the way in which the cells are lysed. It is therefore worthwhile qualifying the statements on l. 35-36, 46-47 and 212 (as Ref 8 used mechanical breakage). This will give better direction to those in the field following up the findings.

      It is also a little alarming that the mutanolysin is contaminated by protease and one hopes this does not affect any of the properties of the materials being analysed.

    3. Reviewer #2 (Public review):

      The Gram-positive cell wall contains for a large part of TAs, and is essential for most bacteria. However, TA biosynthesis and regulation is highly understudied because of the difficulties in working with these molecules. This study closes some of our important knowledge gaps related to this and provides new and improved methods to study TAs. It also shows an interesting role for TAs in maintaining a 'periplasmic space' in Gram positives. Overall, this is an important piece of work. Future work will need to address the possible causal link between TAs and periplasmic space, for instance using complemented mutants and CEMOVIS. It will be interesting to see what happens with the periplasmic space in other mutants besides TA or also in strains with capsules/without capsules and in PG mutants, or in lafB (essential for production of another glycolipid) mutants. Overall, I support the publication of this revised work as it pioneers some new methods that will definitively move the field forward.

    4. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      The authors set out to analyse the roles of the teichoic acids of Streptococcus pneumoniae in supporting the maintenance of the periplasmic region. Previous work has proposed the periplasm to be present in Gram positive bacteria and here advanced electron microscopy approach was used. This also showed a likely role for both wall and lipo-teichoic acids in maintaining the periplasm. Next, the authors use a metabolic labelling approach to analyse the teichoic acids. This is a clear strength as this method cannot be used for most other well studied organisms. The labelling was coupled with super-resolution microscopy to be able to map the teichoic acids at the subcellular level and a series of gel separation experiments to unravel the nature of the teichoic acids and the contribution of genes previously proposed to be required for their display. The manuscript could be an important addition to the field but there are a number of technical issues which somewhat undermine the conclusions drawn at the moment. These are shown below and should be addressed. More minor points are covered in the private Recommendations for Authors.

      Weaknesses to be addressed:

      (1) l. 144 Was there really only one sample that gave this resolution? Biological repeats of all experiments are required.

      CEMOVIS is a very challenging method that is not amenable to numerous repeats. However, multiple images were recorded from at least two independent samples for each strain. Additional sample images are shown in a new Fig. S3.

      CETOVIS is even more challenging (only two publications in Pubmed since 2015) and was performed on a single ultrathin section that, exceptionally, laid perfectly flat on the EM grid, allowing tomography data acquisition on ∆tacL cells. The reconstructed tomogram confirmed the absence of a granular layer in the depth of the section. Additionally, the numbering of Fig. S4A-B (previously misidentified as Fig. S2A-B) has been corrected in the text of V2.

      (2) Fig. 4A. Is the pellet recovered at "low" speeds not just some of the membrane that would sediment at this speed with or without LTA? Can a control be done using an integral membrane protein and Western Blot? Using the tacL mutant would show the behaviour of membranes alone.

      We think that the pellet is not just some of the membrane but most of it. In support of this view, the “low” speed pellets after enzymatic cell lysis contain not just some membrane lipids, but most of them (Fig. S10A). We therefore expect membrane proteins to be also present in this fraction. We performed a Western blot using antibodies against the membrane protein PBP2x (new Fig. S7C). Unfortunately, no signal was detected most likely due to protein degradation from contaminant proteases that we could trace to the purchased mutanolysin. The same sedimentation properties were observed with the ∆tacL strain as shown in Fig. 6A. However, in the ∆tacL strain the membrane pellet still contains membrane-bound TA precursors. It is therefore impossible to test definitely if pneumococcal membranes totally devoid of TA would sediment in the same way.

      (3) Fig. 4A. Using enzymatic digestion of the cell wall and then sedimentation will allow cell wall associated proteins (and other material) to become bound to the membranes and potentially effect sedimentation properties. This is what is in fact suggested by the authors (l. 1000, Fig. S6). In order to determine if the sedimentation properties observed are due to an artefact of the lysis conditions a physical breakage of the cells, using a French Press, should be carried out and then membranes purified by differential centrifugation. This is a standard, and well-established method (low-speed to remove debris and high-speed to sediment membranes) that has been used for S. pneumoniae over many years but would seem counter to the results in the current manuscript (for instance Hakenbeck, R. and Kohiyama, M. (1982), Purification of Penicillin-Binding Protein 3 from Streptococcus pneumoniae. European Journal of Biochemistry, 127: 231-236).

      Thank you for this suggestion. We have tested this hypothesis by breaking cells with a Microfluidizer followed by differential centrifugation. This experiment, which requires an important minimal volume, was performed with unlabeled cells (due to the cost of reagents) and assessed by Western blot using antibodies against the membrane protein PBP2x (new Fig. S7C). In this case, the majority of the membrane material was found in the high-speed pellet, as expected.

      We also applied the spheroplast lysis procedure of Flores-Kim et al. to the labeled cells, and found that most of the labeled material sedimented at low speed (new Fig. S7B), as observed with our own procedure.

      With these new results, the section on membrane density has been removed from the Supplementary Information. Instead, the fractionation is further discussed in terms of size of membrane fragments and presence of intact spheroplasts in the notes in Supplementary Information preceding Fig. S7.

      (4) l. 303-305. The authors suggest that the observed LTA-like bands disappear in a pulse chase experiment (Fig. 6B). What is the difference between this and Fig. 5B, where the bands do not disappear? Fig. 5C is the WT and was only pulse labelled for 5 min and so would one not expect the LTA-like bands to disappear as in 6B?

      Fig. 6B shows a pulse-chase experiment with strain ∆tacL, whereas Fig. 5C shows a similar experiment with the parental WT strain. The disappearance of the LTA-like band pattern with the ∆tacL strain (Fig. 6B), and their persistence in the WT strain (Fig. 5C), indicate that these bands are the undecaprenyl-linked TA in ∆tacL and proper LTA in the WT. A sentence has been added to better explain this point in V2.

      Note that we have exchanged the previous Fig. 5C and Fig. S13B, so that the experiments of Fig. 5A and 5C are in the same medium, as suggested by Reviewer #2.

      (5) Fig. 6B, l. 243-269 and l. 398-410. If, as stated, most of the LTA-like bands are actually precursor then how can the quantification of LTA stand as stated in the text? The "Titration of Cellular TA" section should be re-evaluated or removed? If you compare Fig. 6C WT extract incubated at RT and 110oC it seems like a large decrease in amount of material at the higher temperature. Thus, the WT has a lot of precursors in the membrane? This needs to be quantified.

      Indeed, the quantification of the ratio of LTA and WTA in the WT strain rests on the assumption that the amount of membrane-linked polymerized TA precursors is negligible in this strain. This assumption is now stated in the Titration section. We think it is the case. The true LTA and TA precursors do not have exactly the same electrophoretic mobility, being shifted relative to each other by about half a ladder “step”. This difference is visible when samples are run in adjacent lanes on the same gel, as in the new Fig. 6C. The difference of migration was well documented in the original paper about the deletion of tacL, although tacL was known as rafX at that time, and the ladders were misidentified as WTA (Wu et al. 2014. A novel protein, RafX, is important for common cell wall polysaccharide biosynthesis in Streptococcus pneumoniae: implications for bacterial virulence. J Bacteriol. 196, 3324-34. doi: 10.1128/JB.01696-14). This reference was added in V2. The experiment in the new Fig. 6C was repeated to have all samples on the same gel and treated at a lower temperature. The minor effect on the amount of LTA when WT cells are heated at pH 4.2 may be due to the removal of some labeled phosphocholine. We have NMR evidence that the phosphocholine in position D is labile to acidic treatment of LTA, which may lack in some cases, as reported by Hess et al. (Nat Commun. 2017 Dec 12;8(1):2093. doi: 10.1038/s41467-017-01720-z).

      (6) L. 339-351, Fig. 6A. A single lane on a gel is not very convincing as to the role of LytR. Here, and throughout the manuscript, wherever statements concerning levels of material are made, quantification needs to be done over appropriate numbers of repeats and with densitometry data shown in SI.

      Yes indeed. Apart from the titration of TA in the WT strain, we haven’t yet carried out a thorough quantification of TA or LTA/WTA ratio in different strains and conditions, although we intend to do so in a follow-up study, using the novel opportunities offered by the method presented here.

      However, to better substantiate our statement regarding the ∆lytR strain, we have quantified two experiments performed in C-medium with azido-choline, and two experiments of pulse labeling in BHI medium. The results are presented in the additional supplementary Fig. S14. The value of 51% was a calculation error, and was corrected to 41%. Likewise, the decrease in the WTA/LTA ratio was corrected to 5 to 7-fold.

      (7) 14. l. 385-391. Contrary to the statement in the text, the zwitterionic TA will have associated counterions that result in net neutrality. It will just have both -ve and +ve counterions in equal amounts (dependent on their valency), which doesn't matter if it is doing the job of balancing osmolarity (rather than charge).

      Thank you for pointing out this point. The paragraph has been corrected in V2.

      Reviewer #2 (Public review):

      The Gram-positive cell wall contains for a large part of TAs, and is essential for most bacteria. However, TA biosynthesis and regulation is highly understudied because of the difficulties in working with these molecules. This study closes some of our important knowledge gaps related to this and provides new and improved methods to study TAs. It also shows an interesting role for TAs in maintaining a 'periplasmic space' in Gram positives. Overall, this is an important piece of work. It would have been more satisfying if the possible causal link between TAs and periplasmic space would have been more deeply investigated with complemented mutants and CEMOVIS. For the moment, there is clearly something happening but it is not clear if this only happens in TA mutants or also in strains with capsules/without capsules and in PG mutants, or in lafB (essential for production of another glycolipid) mutants. Finally, some very strong statements are made suggesting several papers in the literature are incorrect, without actually providing any substantiation/evidence supporting these claims. Nevertheless, I support the publication of this work as it pioneers some new methods that will definitively move the field forward.

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      (1) l. 55 It is stated that TA are generally not essential. This needs to be introduced in a little more detail as in several species they are collectively. Need some more references here to give context.

      We have expended the paragraph and added a selection of references in V2.

      (2) l. 63 and Fig. 1A. Is the model based on the images from this paper? Is the periplasm as thick as the peptidoglycan layer? Would you not expect the density of WTA to be the same throughout the wall, rather than less inside? Do the authors think that the TA are present as rods in the cell envelope and because of this the periplasm looks a little like a bilayer, is this so? Is the relative thickness of the layers based on the data in the paper (Table 1)?

      The model proposed in Fig. 1A is not based on our data. It is a representation of the model proposed by Harold Erickson, and the appropriate reference has been added to the figure legend in V2. We do not speculate on the relative density of WTA inside the peptidoglycan layer, at the surface or in the periplasm. The only constraint from the model is that the density of WTA in the periplasm should be sufficient for self-exclusion and allow the brush polymer theory to apply. The legend has been amended in V2.

      We indeed think that the bilayer appearance of the periplasmic space in the wild type strain, and the single layer periplasmic space in the ∆tacL and ∆lytR support the Erickson’s model. Although the model was drawn arbitrarily, it turns out that the relative thickness of the peptidoglycan and periplasmic scale is in rough agreement with the measurements reported in Table 1.

      (3) Fig. 2. It is hard to orient oneself to see the layers. The use of the term periplasmic space (l. 132) and throughout is probably not wise as it is not a space.

      We prefer to retain this nomenclature since the term periplasmic space has been used in all the cell envelope CEMOVIS publications and is at the core of Erickson’s hypothesis about these observations and teichoic acids.

      (4) L. 147. This is not referring to Fig. S2A-B as suggested but Fig. S3A-B.

      This has been corrected.

      (5) l. 148. How do you know the densities observed are due to PG or certainly PG alone? Perhaps it is better to call this the cell wall.

      Yes. Cell wall is a better nomenclature and the text and Table 1 have been corrected in V2, in accordance with Fig. 2.

      (6) l. 165. It is also worth noting that peripheral cell wall synthesis also happens at the same site so this may well not be just division.

      Yes. We have replaced “division site” by “mid-cell” in V2.

      (7) l. 214 What is the debris? If PG digestion has been successful then there will be marginal debris. Is this pellet translucent (like membranes)? If you use fluorescently labelled PG in the preparation has it all disappeared, as would be expected by fully digested and solubilised material?

      In traditional protocols of bacterial membrane preparation, a low-speed centrifugation is first performed to discard “debris” that to our knowledge have not been well characterized but are thought to consist of unbroken cells and large fragments of cell wall. After enzymatic degradation of the pneumococcal cell wall, the low-speed pellet is not translucent as in typical membrane pellets after ultracentrifugation, but is rather loose, unlike a dense pellet of unbroken cells. A description of the pellet appearance was added in V2.

      It is a good idea to check if some labeled PG is also pelleted at low-speed after digestion. In a double labeling experiment using azido-choline and a novel unpublished metabolic probe of the PG, we found that the PG was fully digested and labeled fragments migrated as a couple of fuzzy bands likely corresponding to different labeled peptides. These species were not pelleted at low speed.

      (8) l. 219. Can you give a reference to certify that the low mobility material is WTA? Why does it migrate differently than LTA? Or is the PG digestion not efficient?

      WTA released from sacculi by alkaline lysis were found to migrate as a smear at the top of native gels revealed by alcian-blue silver staining, which is incompatible with SDS (Flores-Kim, 2019, 2022). The references have be added in V2. It could be argued in this case that the smearing was due to partial degradation of the WTA by the alkaline treatment.

      Bui et al. (2012) reported the preparation of WTA by enzymatic digestion of sacculi, but the resulting WTA were without muropeptide, presumably due to a step of boiling at pH 5 used to deactivate the enzymes.

      To our knowledge, this is the first report of pneumococcal WTA prepared by digestion of sacculi and analyzed by SDS-PAGE. Since the migration of WTA in native and SDS-PAGE is similar, we hypothesize that they do not interact significantly with the dodecyl sulphate, in contrast to the LTA, which bear a lipidic moiety. The fuzziness of the WTA migration pattern may also result from the greater heterogeneity due to the attached muropeptide, such as different lengths (di-, tetra-saccharide…), different peptides despite the action of LytA (tri-, tetra-peptide…), different O-acetylation status, etc.

      (9) L. 226-227, Fig S8. Presumably several of the major bands on the Coomassie stained gel are the lysozyme, mutanolysin, recombinant LytA, DNase and RNase used to digest the cell wall etc.? Can the sizes of these proteins be marked on the gel. Do any of them come down with the material at low-speed centrifugation?

      We have provided a gel showing the different enzymes individually and mixed (new Fig. S9G). While performing several experiments of this type, we found that the mutanolysin might be contaminated with proteases. The enzymes do not appear to sediment at low speed.

      (10) Fig. S9B. It is difficult to interpret what is in the image as there appear to be 2 populations of material (grey and sometimes more raised). Does the 20,000 g material look the same?

      Fig. S10B is a 20,000 × g pellet. We agree that there appears to be two types of membrane vesicles, but we do not know their nature.

      (11) l. 277 and Fig. 5A. Why is it "remarkable" that there are apparently more longer LTA molecules as the cell reach stationary phase?

      This is the first time that a change of TA length is documented. Such a change could conceivably have consequences in the binding and activity of CBPs and the physiology of the cell envelope in general. These questions should be adressed in future studies.

      (12) l. 280. How do you know which is the 6-repeat unit?

      It is an assumption based on previous analyses by Gisch et al.( J Biol Chem 2013, 288(22):15654-67. doi: 10.1074/jbc.M112.446963). The reference was added.

      (13) Fig. 5A and C. Panel C, the cells were grown in a different medium and so are not comparable to Panel A. Why is Fig. S12B not substituted for 5B? Presumably these are exponential phase cells.

      We have interverted the Fig. S13B and 5C in V2, as suggested, and changed the text and legends accordingly.

      Reviewer #2 (Recommendations for the authors):

      L30: vitreous sections?

      Corrected in V2.

      L32: as their main universal function --> as a universal function. To show it's the main universal function, you will need to look at this across various bacterial species.

      Changed to “possible universal function” in V2.

      L35: enabled the titration the actual --> titration of the actual?

      Corrected in V2.

      L34: consider breaking up this very long sentence.

      Done in V2.

      L37: may compensate the absence--> may compensate for the absence.

      Corrected in V2.

      L45: Using metabolic labeling and electrophoresis showed --> Metabolic labeling and...

      Corrected in V2.

      L46: This finding casts doubts on previous results, since most LTA were likely unknowingly discarded in these studies. This needs to be rephrased and is unnecessarily callous. While the current work casts doubts on any quantitative assessments of actual LTA levels measured in previous studies, it does not mean any qualitative assessments or conclusions drawn from these experiments are wrong. Better would be to say: These findings suggest that previously reported quantitative assessments of LTA levels are likely underestimating actual LTA levels, since much of the LTA would have been unknowingly discarded.

      If the authors do think that actual conclusions are wrong in previous work, then they need to be more explicit and explain why they were wrong.

      Yes indeed. The statement was toned down in V2.

      L55: Although generally non-essential. I would remove or rephrase this statement. I don't think any TA mutant will survive out in the wild and will be essential under a certain condition. So perhaps not essential for growth under ideal conditions, but for the rest pretty essential.

      The paragraph was amended by qualifying the essentiality to laboratory conditions and including selected references.

      L95: Note that the prevailing model until reference 20 (Gibson and Veening) was that the TA is polymerized intracellularly (see e.g. Figure 2 of PMID: 22432701, DOI: 10.1089/mdr.2012.0026). This intracellular polymerisation model seemed unlikely according to Gibson and Veening ('As TarP is classified by PFAM as a Wzy-type polymerase with predicted active site outside the cell, we speculate that TarP and TarQ polymerize the TA extracellularly in contrast to previous reports.'), but there is no experimental evidence as far as this referee knows of either model being correct.

      Despite the lack of experimental evidence, we think that Gibson and Veening are very likely correct, based on their argument, and also by analogy with the synthesis of other surface polysaccharides from undecaprenyl- or dolichol-linked precursors. It is unfortunate that Figure 2 of PMID: 22432701, DOI: 10.1089/mdr.2012.0026 was published in this way, since there was no evidence for a cytoplasmic polymerization, to our knowledge.

      L97: It is commonly believed, although I'm not sure it has ever been shown, that the capsule is covalently attached at the same position on the PG as WTA. Therefore, there must be some sort of regulation/competition between capsule biosynthesis and WTA biosynthesis (see also ref. 21). The presence of the capsule might thus also influence the characteristics of the periplasmic space. Considering that by far most pneumococcal strains are encapsulated, the authors should discuss this and why a capsule mutant was used in this study and how translatable their study using a capsule mutant is to S. pneumoniae in general.

      A paragraph was added in the Introduction of V2 to present the complication and a sentence was added at the end of the discussion to mention that this should be studied in the future.

      L102: Ref 29 should probably be cited here as well?

      Since in Ref 29 (Flores-Kim et al. 2019) there is a detectable amount of LTA (presumably precursors TA) in the ∆tacL stain, we prefer to cite only Hess et al. 2017 regarding the absence of LTA in the absence of TacL. However, we added in V2 a reference to Flores-Kim et al. 2019 in the following paragraph regarding the role of the LTA/WTA ratio.

      L106: dependent on the presence of the phosphotransferase LytR (21). --> dependent on the presence of the phosphotransferase LytR, whose expression is upregulated during competence (21).

      Corrected in V2.

      L119: I fail to see how the conclusions drawn by other groups (I assume the authors mean work from the Vollmer, Rudner, Bernhardt, Hammerschmidt, Havarstein, Veening groups?) are invalid if they compared WTA:LTA ratios between strains and conditions if they underestimated the LTA levels? Supposedly, the LTA levels were underestimated in all samples equally so the relative WTA/LTA ratio changes will qualitatively give the same outcome? I agree that these findings will allow for a reassessment of previous studies in which presumably too low LTA levels were reported, but I would not expect a difference in outcome when people compared WTA:LTA ratios between strains?

      The sentence was rephrased in V2 to be neutral regarding previous work and rather emphasize future possibilities.

      L131: Perhaps it would be good to highlight that such a conspicuous space has been noticed before by other EM methods (see e.g. Figs.4 and 5 or ref 19, or one of the most clear TEM S. pneumoniae images I have seen in Fig. 1F of Gallay et al, Nat. Micro 2021). However, always some sort of staining had previously been performed so it was never clear this was a real periplasmic space. CEMOVIS has this big advantage of being label free and imaging cells in their presumed native state.

      Thanks for pointing out these beautiful data that we had overlooked. We have added a few sentences and references in the Discussion of V2.

      L201: References are not numbered.

      Corrected in V2.

      L271/L892: Change section title. 'Evolution' can have multiple meanings. It would be more clear to write something like 'Increased TA chain length in stationary phase cells' or something like that.

      Changed in V2.

      L275: harvested

      Corrected in V2.

      L329: add, as suggested shown previously (I guess refs 24 and 29)

      Reference to Hess et al. 2017 has been added in V2. A sentence and further references to Flores-Kim, 2019, 2022 and Wu et al. 2014 were added at the end of the discussion with respect to the LTA-like signal observed in these studies of ∆tacL strains.

      L337: I think a concluding sentence is warranted here. These experiments demonstrate that membrane-bound TA precursors accumulate on the outside of the membrane, and are likely polymerized on the outside as well, in line with the model proposed in ref. 20.

      From the point of view of formal logic, the accumulation of membrane-bound TA precursors on the outer face of the membrane does not prove that they were assembled there. They could still be polymerized inside and translocated immediately. However, since this is extremely unlikely for the reasons discussed by Gibson and Veening, we have added a mild conclusion sentence and the reference in V2.

      L343: How accurate are these quantifications? Just by looking at the gel, it seems there is much less WTA in the lytR mutant than 50% of the wild type?

      Yes, the 51% value was a calculation error. This was changed to 41%. Likewise, the decrease of the WTA amount relative to LTA was corrected to 5- to 7-fold.

      Apart from the titration of TA in the WT strain, we haven’t yet carried out a careful quantification neither of TA nor of the LTA/WTA ratio in different strains and conditions, although we intend to do so in the near future using the method presented here.

      However, to better substantiate our statement regarding the ∆lytR strain, we have quantified two experiments of growth in C-medium with azido-choline, and two experiments of pulse labeling in BHI medium. The results are presented in the additional supplementary Fig. S14.

      L342: although WTA are less abundant and LTA appear to be longer (Fig. 6A). although WTA are less abundant and LTA appear to be longer (Fig. 6A), in line with a previous report showing that LytR the major enzyme mediating the final step in WTA formation (ref. 21). (or something like that). Perhaps better is to start this paragraph differently. For instance: Previous work showed that LytR is the major enzyme mediating the final step in WTA formation (ref. 21). As shown in Fig. 6A, the proportion of WTA significantly decreased in the lytR mutant. However, there was still significant WTA present indicating that perhaps another LCP protein can also produce WTA.

      Changed in V2.

      Of note, WTA levels would be a lot lower in encapsulated strains as used in Ref. 21 (assuming WTA and capsule compete for the same linkage on PG). So perhaps it would be hard to detect any residual WTA in a encapsulated lytR mutant?

      Investigation of the relationship between TA and capsule incorporation or O-acetylation is definitely a future area of study using this method of TA monitoring.

      L371: see my comments related to L131. Some TEM images clearly show the presence of a periplasmic space.

      Comments and references have been added in V2.

      L402: It would be really interesting to perform these experiments on a wild type encapsulated strain. Would these have much more LTA? (I understand you cannot do these experiments perhaps due to biosafety, but it might be interesting to discuss).

      Yes. It would be interesting to compare the TA in D39 and D39 ∆cps strains. We have added this perspective at the end of the discussion in V2.

      L418: ref lacks number

      Corrected in V2.

      L423: refs missing.

      References added in V2.

      L487: See my comments regarding L46. I do not see one valid point in the current paper why underestimating LTA levels would change any of the conclusions drawn in Ref. 21. I do not know the other papers cited well enough, but it seems highly unlikely that their conclusions would be wrong by systematically underestimating LTA levels. As far as I understand it, this current work basically confirms the major conclusions drawn by these 'doubtful' papers (that TacL makes LTA and LytR is the main WTA producer). As such, I find this sentence highly unfair without precisely specifying what the exact doubts are. Sure, this current paper now shows that probably people have discarded unknowingly LTA and therefore underestimated LTA levels, so any quantitative assessment of LTA levels are probably wrong. That is one thing. But to say this casts doubts on these studies is very serious and unfair (unless the authors provide good arguments to support these serious claims).

      Yes indeed. The sentence was rephrased to be strictly factual in V2.

      Table 2: I assume these strains are delta cps? Would be relevant to list this genotype.

      The Table 2 was completed in V2.

      The authors should comment on why the mutants have not been complemented, especially for lytR as it's the last gene in a complex operon. It would be great to see WTA levels being restored by ectopic expression of LytR.

      Yes. We think this could be part of an in-depth study of the attachment of WTA, together with the investigation of the other LCP phosphotransferases.

    1. eLife Assessment

      This important work investigates the mechanism that underlies the switch between feeding and mating behaviors in the oriental fruit fly, Bactrocera dorsalis. Using a variety of approaches, the authors show that this switch is mediated by the neuropeptide, sulfakinin, acting peripherally through the sulfakinin receptor 1 to regulate the expression of antennal odorant receptors. The evidence is solid in support of the hypothesis that sulfakinin signaling mediates changes in the periphery, although additional sites of action may also contribute to these changes.

    2. Joint Public Review:

      Summary:

      The behavioral switch between foraging and mating is important for resource allocation in insects. This study characterizes the role of sulfakinin and the sulfakinin receptor 1 in changes in olfactory responses associated with foraging versus mating behavior in the oriental fruit fly (Bactrocera dorsalis), a significant agricultural pest. This pathway regulates food consumption and mating receptivity in other species; here the authors use genetic disruption of sulfakinin and sulfakinin receptor 1 to provide strong evidence that changes in sulfakinin signaling modulate antennal responses to food versus pheromonal cues and alter the expression of ORs that detect relevant stimuli.

      Strengths:

      The authors utilize multiple complementary approaches including CRISPR/Cas9 mutagenesis, behavioral characterization, electroantennograms, RNA sequencing and heterologous expression to convincingly demonstrate the involvement of the sulfakinin pathway in the switch between foraging and mating behaviors. The use of both sulfakinin peptide and receptor mutants is a strength of the study and implicates specific signaling actors.

      Weaknesses:

      The authors demonstrate that SKR is expressed in olfactory neurons, however there are additional potential sites of action that may contribute to these results.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Joint Public Review:

      Summary:

      The behavioral switch between foraging and mating is important for resource allocation in insects. This study characterizes the role of sulfakinin and the sulfakinin receptor 1 in changes in olfactory responses associated with foraging versus mating behavior in the oriental fruit fly (Bactrocera dorsalis), a significant agricultural pest. This pathway regulates food consumption and mating receptivity in other species; here the authors use genetic disruption of sulfakinin and sulfakinin receptor 1 to provide strong evidence that changes in sulfakinin signaling modulate antennal responses to food versus pheromonal cues and alter the expression of ORs that detect relevant stimuli.

      Strengths:

      The authors utilize multiple complementary approaches including CRISPR/Cas9 mutagenesis, behavioral characterization, electroantennograms, RNA sequencing and heterologous expression to convincingly demonstrate the involvement of the sulfakinin pathway in the switch between foraging and mating behaviors. The use of both sulfakinin peptide and receptor mutants is a strength of the study and implicates specific signaling actors.

      Weaknesses:

      The authors demonstrate that SKR is expressed in olfactory neurons, however there are additional potential sites of action that may contribute to these results.

      Recommendations for the authors:

      The authors have addressed most of the issues raised by the reviewers. Below are a few outstanding issues.

      (1) Lines 68-69 describe "control of B. dorsalis include the use of the behavioral responses to semiochemicals" but does not describe what these responses are or how behavior is modulated.

      The sentence was revised as “Control of B. dorsalis include the use of the reproductive and feeding behavioral responses to semiochemicals” (lines 69 in the revision).

      (2) Statistical analysis for 9 hour starved females at 5 minutes is missing in Figure 1D and S1.

      We had added statistical analysis for 9 hour starved females at 5 minutes in the revised Figures 1D and S1, respectively (lines 578).

      (3) The legend in Figure S2 should be revised as it is not clear from the figure which of the odors are food associated odors.

      As suggested, we added food odor label in the revised Figure S2 (lines 666).

      (4) Line 167: "Therefore, the upregulated OR genes in starved WT flies, OR7a.4, OR7a.8 and OR10a, were activated by the pheromonal components, while down regulated genes, OR49a and OR63a, were activated by food volatiles." Based on the data, this sentence is incorrect - Therefore, the upregulated OR genes in starved WT flies, OR7a.4, OR7a.8 and OR10a, were activated by the food components, whereas downregulated genes, OR49a and OR63a, were activated by pheromonal components."

      We are sorry for our mistake. We had corrected it (lines 168-169).

      (5) Line 192: "The coordinated action of sulfakinin on mutiple downstreams,..." should be revised to "downstream pathways or tissues" or simply removing "multiple downstream".

      As suggested, we removed “multiple downstream”. See line 192.

      (6) Reference formatting is inconsistent: see line 207 vs line 208.

      We had corrected it as “(Wu et al., 2019)” (lines 207). 

      (7) Lines 241-244 The broad discussion regarding the evolution and ancestral function of CCK here and the phylogeny in Figure S6 are peripheral to the authors claims.

      As suggested, we removed the section and the Figure S6 in the revision.

    1. eLife Assessment

      This important study reports findings that Trpγ, a type of transient receptor potential (TRP) channel expressed in Dh44-releasing neuroendocrine cells, mediates starvation-dependent lipid catabolism. Overall, the claims of the authors are supported by solid evidence. The work should be of interest to both basic and medical biologists working on lipid metabolism.

    2. Reviewer #1 (Public review):

      Summary:

      This research article by Nath et al. from the Lee Lab addresses how lipolysis under starvation is achieved by a transient receptor potential channel, TRPγ, in the neuroendocrine neurons to help animals survive prolonged starvation. Through a series of genetic analyses, the authors identify that trpγ mutations specifically lead to a failure in lipolytic processes under starvation, thereby reducing animals' starvation resistance. The conclusion was confirmed through total triacylglycerol levels in the animals and lipid droplet staining in the fat bodies. This study highlights the importance of transient receptor potential (TRP) channels in the fly brain to modulate energy homeostasis and combat metabolic stress. However, the co-expression of trpγ and Dh44-R2 in the gut is not convincing, especially in the picture of the arrows pointing at the autofluorescence signals in the gut (Figure 7P). Therefore, the authors should either confirm the co-expression or acknowledge that trpγ and Dh44-R2 are not co-expressed in the gut and modify their model in Figure 8 accordingly, although clarifying their co-expression may not change the main conclusions of this study. Overall, the revised version includes the required clarifications on their important results that strengthen the interpretations of the research as well as the visibility of this study.

      Strengths:

      This study identifies the biological meaning of TRPγ in promoting lipolysis during starvation, advancing our knowledge about the TRPγ channel and the neural mechanisms to combat metabolic stress. Furthermore, this study demonstrates the potential of the TRPγ channel as a target to develop new therapeutic strategies for human metabolic disorders by showing that metformin and AMPK pathways are involved in its function in lipid metabolisms during starvation in Drosophila.

    3. Reviewer #2 (Public review):

      Summary

      In this paper, the function of trpγ in lipid metabolism was investigated. The authors found that lipid accumulation levels were increased in trpγ mutants and remained high during starvation; the increased TAG levels in trpγ mutants were restored by the expression of active AMPK in DH44 neurons and oral administration of the anti-diabetic drug metformin. Furthermore, oral administration of lipase, TAG and free fatty acids effectively restored survival of trpγ mutants under starvation conditions. These results indicate that TRPv plays an important role in the maintenance of systemic lipid levels through the proper expression of lipase. Furthermore, authors have shown that this function is mediated by DH44R2. This study provides an interesting finding in that the neuropeptide DH44 released from the brain regulates lipid metabolism through a brain-gut axis, acting on the receptor DH44R2 expressed in gut cells.

      Strengths

      Using Drosophila genetics, careful analysis of which cells express trpγ regulates lipid metabolism is performed in this study. The study supports its conclusions from various angles, including not only TAG levels, but also fat droplet staining and survival rate under starved conditions, and oral administration of substances involved in lipid metabolism.

      Weaknesses

      The function of lipases, as well as identification of cell types, in the DH44R2-expressing cells in the gut can be investigated.

    4. Reviewer #3 (Public review):

      In this manuscript, the authors demonstrated the significance of the TRPγ channel in regulating internal TAG levels. They found high TAG levels in TRPγ mutant, which was ascribed to a deficit in the lipolysis process due to the downregulation of brummer (bmm). It was notable that the expression of TRPγ in DH44+ PI neurons, but not dILP2+ neurons, in the brain restored the internal TAG levels and that the knockdown of TRPγ in DH44+ PI neurons resulted in an increase in TAG levels. These results suggested a non-cell autonomous effect of Dh44+PI neurons. Additionally, the expression of the TRPγ channel in Dh44 R2-expressing cells restored the internal TAG levels. The authors, however, did not provide an explanation of how TRPγ might function in both presynaptic and postsynaptic cells in the non-cell autonomous manner to regulate the TAG storage. The authors further determined the effect of TRPγ mutation on the size of lipid droplets (LD) and the lifespan and found that TRPγ mutation caused an increase in the size of LD and a decrease in the lifespan, which were reverted by feeding lipase and metformin. These were creative endeavors, I thought. The finding that DH44+ PI neurons have non-cell autonomous functions in regulating bodily metabolism (mainly sugar/lipid) in addition to directing sugar nutrient sensing and consumption is likely correct, but the paper has many loose ends.

      Comments on revisions:

      The authors have addressed nearly all of my concerns with additional experiments and explanations.

    5. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public Review):

      Summary:

      This research article by Nath et al. from the Lee Lab addresses how lipolysis under starvation is achieved by a transient receptor potential channel, TRPγ, in the neuroendocrine neurons to help animals survive prolonged starvation. Through a series of genetic analyses, the authors identify that TRPγ mutations specifically lead to a failure in lipolytic processes under starvation, thereby reducing animals' starvation resistance. The conclusion was confirmed through total triacylglycerol levels in the animals and lipid droplet staining in the fat bodies. This study highlights the importance of transient receptor potential (TRP) channels in the fly brain to modulate energy homeostasis and combat metabolic stress. While the data is compelling and the message is easy to follow, several aspects require further clarification to improve the interpretation of the research and its visibility in the field.

      Strengths:

      This study identifies the biological meaning of TRPγ in promoting lipolysis during starvation, advancing our knowledge about TRP channels and the neural mechanisms to combat metabolic stress. Furthermore, this study demonstrates the potential of the TRP channel as a target to develop new therapeutic strategies for human metabolic disorders by showing that metformin and AMPK pathways are involved in its function in lipid metabolisms during starvation in Drosophila.

      Weaknesses:

      Some key results that might strengthen their conclusions were left out for discussion or careful explanation (see below). If the authors could improve the writing to address their findings and connect their findings with conclusions, the research would be much more appreciated and have a higher impact in the field.

      Here, I listed the major issues and suggestions for the authors to improve their manuscript:

      (1) Are the increased lipid droplet size and the upregulated total TAG level measured in the starved or sated mutant in Figure 1? This information might be crucial for readers to understand the physiological function of TRP in lipid metabolism. In other words, clarifying whether the upregulated lipid storage is observed only in the starved trp mutant will advance our knowledge of TRPγ. If the increase of total TAG level is only observed in the starved animals, TRP in the Dh44 neurons might serve as a sensor for the starvation state required to promote lipolysis in starvation conditions. On the other hand, if the total TAG level increases in both starved and sated animals, activation of Dh44 through TRPγ might be involved in the lipid metabolism process after food ingestion.

      We measured total TAG level in Figure 1 and LD sizes in Figure 2 under sated condition. We inserted “under sated condition” to clarify it. lines 97 and 147-148.

      Thanks for your suggestions.

      (2) It is unclear how AMPK activation in Dh44 neurons reduces the total triacylglycerol (TAG) levels in the animals (Figure 3G). As AMPK is activated in response to metabolic stress, the result in Figure 3G might suggest that Dh44 neurons sense metabolic stress through AMPK activation to promote lipolysis in other tissues. Do Dh44 neurons become more active during starvation? Is activation of Dh44 neurons sufficient to activate AMPK in the Dh44 neurons without starvation? Is activation of AMPK in the Dh44 neurons required for Dh44 release and lipolysis during starvation? These answers would provide more insights into the conclusion in Lines 192-193.

      In our previous study, we demonstrated that trpγ mutants exhibited lower levels of glucose, trehalose and glycogen level (Dhakal et al. 2022), and in the current study, we observed excessive lipid storage in the trpγ mutant, indicating imbalanced energy homeostasis. Given the established role of AMPK in maintaining energy balance (Marzano et. al., 2021, Lin et al 2021), we employed the activated form of AMPK (UAS-AMPK<sup>TD</sup>) in our experiments. Our result showed that expression of activated AMPK in Dh44 neurons led to a reduction in total TAG levels, suggesting that AMPK activation in these neurons can promote lipolysis even in the absence of starvation. Regarding the activation of Dh44 neurons, Dus et al in 2015 reported that Dh44 cells in the brain are activated by nutritive sugars especially in starvation conditions. In addition, another report showed a role of Dh44 neuron in regulating starvation induced sleep suppression (Oh et. al., 2023) which may imply that these neurons become more active under starved conditions. We did not directly assess whether Dh44 neuron activity increases during starvation or whether AMPK activation in these neurons is required for DH44 release and subsequent lipolysis, our finding support the notion that AMPK activation in Dh44 neuron is sufficient to reduce TAG levels, potentially by metabolic stress response typically observed during starvation. We explained it like the following: “Dh44 neurons regulate starvation-induced sleep suppression (Oh et. al., 2023), which implies that these neurons become more active under starved conditions.” lines 190-191.

      (3) It is unclear how the lipolytic gene brummer is further downregulated in the trpγ mutant during starvation while brummer is upregulated in the control group (Figure 6A). This result implies that the trpγ mutant was able to sense the starvation state but responded abnormally by inhibiting the lipolytic process rather than promoting lipolysis, which makes it more susceptible to starvation (Figure 3B).

      Thanks for your suggestions. We explained it like the following: “The data indicates that the trpg mutant can sense the starvation state but responds abnormally by suppressing lipolysis instead of activating it. This dysregulated lipolytic response likely increases the mutant's vulnerability to starvation, as it cannot effectively mobilize lipid stores for energy during periods of nutrient deprivation.” lines 251-254.

      (4) There is an inconsistency of total TAG levels and the lipid droplet size observed in the Dh44 mutant but not in the Dh44-R2 mutant (Figures 7A and 7F). This inconsistency raises a possibility that the signaling pathway from Dh44 release to its receptor Dh44-R2 only accounts for part of the lipid metabolic process under starvation. Adding discussion to address this inconsistency may be helpful for readers to appreciate the finding.

      Thanks for your suggestion. We included the following in the Discussion: “There is an inconsistency of total TAG levels and the LD size observed in the Dh44 mutant. This inconsistency raises a possibility that the signaling pathway from DH44 release to its receptor DH44R2 only accounts for part of the lipid metabolic process under starvation. While Dh44 mutant flies displayed normal internal TAG levels, Dh44R2 mutant flies exhibited elevated TAG levels. This suggested that the lipolysis phenotype could be facilitated by a neuropeptide other than DH44. Alternatively, a DH44 neuropeptide-independent pathway could mediate the lipolysis.” lines 429-436.

      Reviewer #2 (Public Review):

      Summary:

      In this paper, the function of trpγ in lipid metabolism was investigated. The authors found that lipid accumulation levels were increased in trpγ mutants and remained high during starvation; the increased TAG levels in trpγ mutants were restored by the expression of active AMPK in DH44 neurons and oral administration of the anti-diabetic drug metformin. Furthermore, oral administration of lipase, TAG, and free fatty acids effectively restored the survival of trpγ mutants under starvation conditions. These results indicate that TRPv plays an important role in the maintenance of systemic lipid levels through the proper expression of lipase. Furthermore, authors have shown that this function is mediated by DH44R2. This study provides an interesting finding in that the neuropeptide DH44 released from the brain regulates lipid metabolism through a brain-gut axis, acting on the receptor DH44R2 presumably expressed in gut cells.

      Strengths:

      Using Drosophila genetics, careful analysis of which cells express trpγ regulates lipid metabolism is performed in this study. The study supports its conclusions from various angles, including not only TAG levels, but also fat droplet staining and survival rate under starved conditions, and oral administration of substances involved in lipid metabolism.

      Weaknesses:

      Lipid metabolism in the gut of DH44R2-expressing cells should be investigated for a better understanding of the mechanism. Fat accumulation in the gut is not mechanistically linked with fat accumulation in the fat body. The function of lipase in the gut (esp. R2 region) should be addressed, e.g. by manipulating gut-lipases such as magro or Lip3 in the gut in the contest of trpγ mutant. Also, it is not clarified which cell types in the gut DH44R2 is expressed. The study also mentioned only in the text that bmm expression in the gut cannot restore lipid droplet enlargement in the fat body, but this result might be presented as a figure.

      We appreciate the reviewer’s insightful suggestions. Unfortunately, due to the unviability of the reagent (UAS-Lip3), we were unable to manipulate gut lipase in trpy mutants as proposed. However, we additionally performed immunostaining to examine the co-expression of trpγ and Dh44R2 in the gut, and our results indicate that both trpγ and Dh44R2 are co-expressed in the R2 region of the gut (Figure 7O and P). Furthermore, we have updated our figures to address the point that bmm expression in the gut does not restore lipid droplet enlargement in the fat body, with the revised version (Figure 5I and J).

      Reviewer #3 (Public Review):

      In this manuscript, the authors demonstrated the significance of the TRPγ channel in regulating internal TAG levels. They found high TAG levels in TRPγ mutant, which was ascribed to a deficit in the lipolysis process due to the downregulation of brummer (bmm). It was notable that the expression of TRPγ in DH44+ PI neurons, but not dILP2+ neurons, in the brain restored the internal TAG levels and that the knockdown of TRPγ in DH44+ PI neurons resulted in an increase in TAG levels. These results suggested a non-cell autonomous effect of Dh44+PI neurons. Additionally, the expression of the TRPγ channel in Dh44 R2-expressing cells restored the internal TAG levels. The authors, however, did not provide an explanation of how TRPγ might function in both presynaptic and postsynaptic cells in the non-cell autonomous manner to regulate the TAG storage. The authors further determined the effect of TRPγ mutation on the size of lipid droplets (LD) and the lifespan and found that TRPγ mutation caused an increase in the size of LD and a decrease in the lifespan, which were reverted by feeding lipase and metformin. These were creative endeavors, I thought. The finding that DH44+ PI neurons have non-cell autonomous functions in regulating bodily metabolism (mainly sugar/lipid) in addition to directing sugar nutrient sensing and consumption is likely correct, but the paper has many loose ends. I would like to see a revision that includes more experiments to tighten up the findings and appropriate interpretations of the results.

      (1) The authors need to provide interpretations or speculations as to how DH44+ PI neurons have non-cell autonomous functions in regulating the internal TAG stores, and how both presynaptic DH44 neurons and postsynaptic DH44 R2 neurons require TRPγ for lipid homeostasis.

      In Discussion, we had mentioned our previous finding. “ We previously proposed that TRPg holds DH44 neurons in a state of afterdepolarization, thus reducing firing rates by inactivating voltage-gated Na+ channels (Dhakal et al., 2022). At the physiological level, this induces the consistent release of DH44 and depletion of DH44 stores, resulting in nutrient utilization and storage malfunctions.”

      We also included the following: “TRPg in DH44 neurons may influence the release of metabolic signals or hormones that act on postsynaptic DH44R2 cells. These postsynaptic cells could, in turn, modulate lipid storage and metabolism in a non-cell autonomous manner. However, the mechanism by which TRPg functions in DH44R2 cells remains unclear. One possible explanation is that TRPg in the gut may be activated by stretch or osmolarity (Akitake et al. 2015).” lines 439-440.

      This interaction between presynaptic and postsynaptic cells may ensure a coordinated response to metabolic changes and maintain lipid homeostasis. Thus, both Dh44-expressing and Dh44-R2-expressing cells are crucial for the proper functioning of TRPγ in regulating internal TAG levels and lipid storage.

      (2) The expression of TRPγ solely in DH44 R2 neurons of TRPγ mutant flies restored the TAG phenotype, suggesting an important function mediated by TRPγ in DH44 R2 neurons. However, the authors did not document the endogenous expression of TRPγ in the DH44R2+ gut cells. This needs to be shown.

      We appreciate the reviewer’s suggestion. To address this, we performed immunostaining to examine the expression of TRPγ in the DH44R2+ gut cells. Our results, as shown in Figure 7 O and P, confirm that TRPγ is co-expressed in the Dh44R2+ cells in the gut. We also found that Dh44R2 is expressed in the brain as well. We documented this part like the following: “Given that Dh44R2 is predominantly expressed in the intestine, we performed immunostaining to examine whether Dh44R2 co-localizes with trpg in gut cells. Our results confirmed that Dh44R2 and trpg are co-expressed in intestinal cells (Figure 7O and P). Additionally, we analyzed Dh44R2 expression in the brain and found that two Dh44R2-expressing cells are co-localized with Dh44-expressing cells in the PI region (Figure 7Q). To further delineate whether Dh44R2-mediated fat utilization is specific to the brain, gut, or fat body, we knocked down Dh44R2<sup>RNAi</sup> using Dh44-GAL4, myo1A-GAL4, and cg-GAL4, respectively (Figure 7–figure supplement 1E). Notably, knockdown of Dh44R2 with Myo1A-GAL4 resulted in elevated TAG levels, indicating that DH44R2 activity in lipid metabolism is specific to the gut.” lines 375-384.

      (3) While Dh44 mutant flies displayed normal internal TAG levels, Dh44R2 mutant flies exhibited elevated TAG levels (Figure 7A). This suggested that the lipolysis phenotype could be facilitated by a neuropeptide other than Dh44. Alternatively, a Dh44 neuropeptide-independent pathway could mediate the lipolysis. In either case, an additional result is needed to substantiate either one of the hypotheses.

      The Dh44 mutant flies exhibited normal TAG levels, whereas Dh44R2 mutant flies showed elevated TAG levels. However, when we examined the lipid droplets in the fat body, both Dh44 mutant and Dh44R2 mutant flies displayed larger lipid droplets, indicating a disruption in lipid metabolism. Additionally, we assessed starvation survival time and found that both Dh44 and Dh44R2 mutant flies exhibited reduced survival under starvation conditions compared to controls. Supplementation with lipase (Figure 7–figure supplement 1A), glycerol (Figure 7–figure supplement 1B), hexanoic acid (Figure 7–figure supplement 1C), and mixed TAGs (Figure 7–figure supplement 1D) improved starvation survival time, further supporting that the lipid metabolism pathway was impaired in both mutants. These observations highlight the role of Dh44 in regulating lipolysis. We included related Discussion: “There is an inconsistency of total TAG levels and the LD size observed in the Dh44 mutant. This inconsistency raises a possibility that the signaling pathway from DH44 release to its receptor DH44R2 only accounts for part of the lipid metabolic process under starvation. While Dh44 mutant flies displayed normal internal TAG levels, Dh44R2 mutant flies exhibited elevated TAG levels. This suggested that the lipolysis phenotype could be facilitated by a neuropeptide other than DH44. Alternatively, a DH44 neuropeptide-independent pathway could mediate the lipolysis.” lines 429-436.

      (4) While the authors observed an increased area of fat body lipid droplets (LD) in Dh44 mutant flies (Figure 7F), they did not specify the particular region of the fat body chosen for measuring the LD area.

      We have chosen the 2-3 segment in the abdomen for all fat body images, which we already mentioned in Nile red staining in the Method section line 630-631.

      (5) The LD area only accounts for TAG levels in the fat body, whereas TAG can be found in many other body parts, including the R2 area as demonstrated in Figure 5A-D using Nile red staining. As such, measuring the total internal TAG levels would provide a more accurate representation of TAG levels than the average fat body LD area.

      We have measured total internal TAG level in whole body throughout the experiments (Figure 1F, 2C, 2E, 3C, 3G, 4A, 4B, 7A, 7I, and many Supplementary Figures) except bmm expression using GAL4/UAS system. Now we include this new data in Figure 5–figure supplement 1) which is the same conclusion with LD analysis.

      (6) In Figure 5F-I, the authors should perform the similar experiment with Dh44, Dh44R1, and Dh44R2 mutant flies.

      We did the experiments with Dh44, Dh44R1, and Dh44R2 mutant flies and we found that Dh44 and Dh44R2 mutant flies showed reduced starvation survival time than control and which was increased after supplementation of lipase, glycerol, hexanoic acid and TAG (Figure 7– figure supplement 1A–D). lines 361-372.

      (7) The representative image in Figure 6B does not correspond to the GFP quantification results shown in Figure 6C. In trpr1;bmm::GFP flies, the GFP signal appears stronger in starved conditions than in satiated conditions.

      We updated it with new images. We quantified GFP intensity level using image J and found that GFP intensity level was significantly lower in starved condition in trpγ<sup>1</sup>;bmm::GFP flies than sated condition.

      (8) In Figure 6H-I, fat body-specific expression of bmm reversed the increased LD area in TRPγ mutants. The authors also showed that Dh44+PI neuron-specific expression of bmm yielded a similar result. The authors need to provide an interpretation as to how bmm acts in the fat body or DH44 neurons to regulate this.

      We first inserted the following in results: “Furthermore, the expression of bmm in the fat body, as well as Dh44 neurons in the PI region, can promote lipolysis at the systemic level.” lines 276-277.

      Additionally, we discussed it in the Discussion: “Brummer lipase is essential for regulating lipid levels in the insect fat body by mediating lipid mobilization and energy homeostasis. In Nilaparvata lugens, it facilitates triglyceride breakdown (Lu et al., 2018), while studies in Drosophila show that reduced Brummer lipase expression decreases fatty acids and increases diacylglycerol levels, highlighting its role in lipid metabolism (Nazario-Yepiz et al., 2021). Here, we additionally demonstrate that bmm expression in DH44 neurons within the PI region can systemically regulate TAG levels. Cell signaling or energy status in DH44 neurons may contribute to hormonal release that targets organs such as the fat body.” lines 451-459.

      (9) The authors should explain why the DH44 R1 mutant did not represent similar results as the wild type.

      We added “In addition, bmm levels in Dh44R1<sup>Mi</sup> under starved condition did not increase as significantly as in the control. This suggests a unique role of DH44 and its receptors in regulating lipid metabolism and response to nutritional status in Drosophila.” lines 358-360.

      (10) It would be good to have a schematic that represents the working model proposed in this manuscript.

      We updated the schematic model in revised version (Figure 8).

      Recommendations for the authors:

      Reviewing Editor (Recommendations For The Authors):

      This paper characterized the function of trpγ in Dh44-expressing PI neurons for lipid metabolism and lipolysis induced by prolonged starvation. The authors applied a series of lipolytic genetic manipulation and lipid/lipid metabolism supplements to rescue the trpγ deficits in lipolysis: the expression of active AMPK in the DH44-expressing PI neurons or brummer, a lipolytic gene, in the trpγ-expressing cells, and oral administration of the anti-diabetic drug metformin, lipase, TAG and free fatty acids. Despite this exhaustive characterization of the defective lipolysis in the trpγ mutants, there remain puzzles in inconsistent defects of Dh44 and DH44R2 in the total TAG levels and in the expression and functions of the receptor in the gut. Clarification of these points and other issues raised by the reviewers should improve the mechanisms of lipid metabolism through Dh44 signalling.

      Reviewer #1 (Recommendations For The Authors):

      (1) It might be worth introducing Dh44 in the introduction section as it is unclear to readers how the authors hypothesized the site-of-action of TRPγ in Dh44 neurons for lipid metabolism after reading the introduction.

      We introduced the following: “We found that TRPg expression in Dh44 neuroendocrine cells in the brain is critical for maintaining normal carbohydrate levels in tissues (Dhakal et al. 2022). Building on this, we hypothesized that TRPg in Dh44 cells also regulates lipid and protein homeostasis.” lines 69-71.

      (2) Providing a summary model in the end to integrate the present findings and their previous publication about TRPγ functions in Drosophila sugar selection would greatly help readers understand and appreciate the general role of TRPγ in balancing energy homeostasis.

      We made a schematic model in Figure 8.

      (3) Swapping the order of Figures 5 and 6 might be a better way to tell the story without logic gaps. The results addressing the mechanisms of metformin and TRPγ in promoting lipolysis under starvation are interrupted by the lipid storage data in the R2 cells in the current Figure 5A-5E. In addition, presenting Figure 5A-5E before or together with Figure 7 will help readers appreciate the expression of Dh44-R2 and its function in regulating lipid metabolism in Figure 7.

      We did.

      (4) It might be misleading to use the word "sated" for the condition of 5-hour mild starvation. The word "mild starvation" or the equivalents might be a better word choice.

      We appreciate the reviewer’s concern. As hemolymph sugar level does not drop down significantly in 5 hr starvation, the previous papers (Dus et al 2015, Dhakal et al 2022) indicated it as sated condition. To use the word consistently, we prefer using “sated” instead of “mild starvation”.

      (5) It is unclear what the white arrows are pointing at in Figures 7O and 7P. Some of those seem to be non-specific signals, so it is hard to connect the figure to the conclusion in Lines 351-353. It would be helpful to add some explanations to help readers interpret Figures 7O and 7P.

      In the previous version, Figure 7O and 7P white arrows represented the expression of Dh44R2 in the SEZ region of the brain and R2 region of the gut. In revised version, to make clear, we performed additional immunostaining for the co-expression of trpγ and Dh44R2 in the gut. We found that trpγ and Dh44R2 co-expressed at the R2 region of the gut specifically (Figure 7O and P). Similarly, we found that two cells of Dh44R2 co-expressed in Dh44 cells in the PI region of the brain (now Figure 7Q). We updated this part. lines 375-380.

      (6) The figure legend for the (G) panel in Figure 2-figure Supplement 1 was mislabeled as (F).

      We corrected it.

      (7) In Line 85, the authors might want to write "… among these mutants, only trpγ mutant displayed reduced carbohydrate levels, suggesting …". Please confirm the information for the sentence. lines 87-88.

      We clarified it.

      Reviewer #2 (Recommendations For The Authors):

      (1) The trpγ[G4] would be difficult for non-Drosophila researchers to understand; it would be better to use trpγ-Gal4.

      We got the mutant line from Dr. Craig Montell who named it. We explained it like the following in the main text: “controlled by GAL4 knocked into the trpg locus (trpg<sup>G4</sup> flies; +)” line 109.

      (2) The arrows in Figures 7O and 7P need to be explained in the figure legends.

      We did.

      Reviewer #3 (Recommendations For The Authors):

      (11) Lines 95-96 should have a reference.

      We did.

      (12) Lines 129-130: It should read "TRPγ expressed in DH44 cells is sufficient for the regulation of lipid levels."

      We changed it as suggested.

      (13) Figure 5E needs to be repeated with more trials.

      We increased the n numbers. Previously (Figure 5E) we included area of 10 LDs from 3 samples, and in revised figure (Figure 6I) we have included 28 LDs from 10 samples.

      (14) Figures 5F-I, bold lines are not too visible and therefore, dotted lines could be used.

      We changed it as suggested.

      (15) Line 356: It is not true that D-trehalose or D-fructose is commonly detected by DH44 neurons. These sugars at concentrations much higher than the physiological concentration range stimulate DH44 neurons (see Dus et al., 2015).

      We removed it.

      (16) Lines 362-363: It should read "Expression of TRPγ in DH44 neurons was necessary and sufficient to regulate the carbohydrate and lipid levels.".

      We changed it.

      (17) Lines 369-370: The authors need to consider removing the possible role of CRF in regulating lipid homeostasis. It could be considered to be far-fetched.

      We removed it.

      (18) Line 407-408: the sentence "Nevertheless, it is also known that DH44 neurons mediate the influence of dietary amino acids on promoting food intakes in flies (37)" needs to be removed. They used amino acid concentrations that were far greater than the physiological levels observed in the internal milieu of flies. Still, many laboratories cannot reproduce the result of using the high AA concentrations.

      We removed it.

    1. eLife Assessment

      This study provides an important computational tool for analyzing and deconvoluting a pool of plasmids sequenced without barcoding using nanopore long-read sequencing. The tool, which has been convincingly validated, is readily available to scientists interested in rapid and cost-effective verification of plasmid sequences as well as in scaling up analysis by pooling samples within barcodes.

    2. Reviewer #1 (Public review):

      This manuscript presents SAVEMONEY, a computational tool designed to enhance the utilization of Oxford Nanopore Technologies (ONT) long-read sequencing for the design and analysis of plasmid sequencing experiments. In the past few years, with the improvement in both sequencing length and accuracy, ONT sequencing is being rapidly extended to almost all omics analyses which are dominated by short-read sequencing (e.g., Illumina). However, relatively higher sequencing errors of long-read sequencing techniques including PacBio and ONT is still a major obstacle for plasmid/clone-based sequencing service that aims to achieve single base/nucleotide accuracy. This work provides a guideline for sequencing multiple plasmids together using the same ONT run without molecular barcoding, followed by data deconvolution. The whole algorithm framework is well-designed, and some real data and simulation data are utilized to support the conclusions. The tool SAVEMONEY is proposed to target users who have their own ONT sequencers and perform library preparation and sequencing by themselves, rather than relying on commercial services. As we know and discussed by the authors, in the real world, to ensure accuracy, the researchers will routinely pick up multiple colonies in the same plasmid construction and submit for Sanger sequencing. However, SAVEMONEY is not able to support the simultaneous analysis of multiple colonies in the same run, as compared to the barcoding-based approaches. This is a major limitation in the significance of this work. Encouraging computational efforts in ONT data debarcoding for mixed-plasmid or even single-cell sequencing would be more valuable in the field.

      Comments on revisions:

      My previous concerns have been addressed, and the revised manuscript has been significantly approved.

    3. Reviewer #2 (Public review):

      The authors developed an algorithm that allows to deconvolute plasmid sequences from a mixture of plasmids that have been sequenced by nanopore long read technology. As library preparations and barcoding of individual samples increases sequencing costs, the algorithm bypasses this need and thus decreases time on sample prep and sequencing costs. In a first step, the tool assesses which of the plasmid constructions can be mixed in a single library preparation by calculating a distance matrix between the reference plasmid and the constructions producing sequence clusters. The user is given groups of plasmids, from different clusters, to be pooled together for sequencing. After sequencing, the algorithm deconvolutes the reads by classifying them based on alignments to the reference sequence. A Bayesian analysis approach is used to obtain a consensus sequence and quality scores.

      Strengths

      The authors exploit one of the main advantages of long read sequencing that is to accurately resolve regions of high complexity, as regularly found in plasmids, and developed a tool that can validate plasmid constructions by reducing sequencing costs. Multiple plasmids (up to six) can be analyzed simultaneously in a single library without the need of sample barcoding, also reducing sample preparation time. Although inserts must be different, just 2 bases difference would be enough for correct assignation. Maximizes cost-efficiency for projects that require large amounts of plasmid constructions and high-throughput validation. The algorithm also allows for linear DNA analysis offering extra flexibility.

    4. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (public review): 

      This manuscript presents SAVEMONEY, a computational tool designed to enhance the utilization of Oxford Nanopore Technologies (ONT) long-read sequencing for the design and analysis of plasmid sequencing experiments. In the past few years, with the improvement in both sequencing length and accuracy, ONT sequencing is being rapidly extended to almost all omics analyses which are dominated by short-read sequencing (e.g., Illumina). However, relatively higher sequencing errors of long-read sequencing techniques including PacBio and ONT is still a major obstacle for plasmid/clone-based sequencing service that aims to achieve single base/nucleotide accuracy. This work provides a guideline for sequencing multiple plasmids together using the same ONT run without molecular barcoding, followed by data deconvolution. The whole algorithm framework is well-designed, and some real data and simulation data are utilized to support the conclusions. The tool SAVEMONEY is proposed to target users who have their own ONT sequencers and perform library preparation and sequencing by themselves, rather than relying on commercial services. As we know and discussed by the authors, in the real world, to ensure accuracy, the researchers will routinely pick up multiple colonies in the same plasmid construction and submit for Sanger sequencing. However, SAVEMONEY is not able to support the simultaneous analysis of multiple colonies in the same run, as compared to the barcoding-based approaches. This is a major limitation in the significance of this work. Encouraging computational ePorts in ONT data debarcoding for mixed-plasmid or even single-cell sequencing would be more valuable in the field. 

      We thank the reviewer for the positive response to our manuscript and the helpful comments.

      The tool SAVEMONEY is proposed to target users who have their own ONT sequencers and perform library preparation and sequencing by themselves, rather than relying on commercial services.

      We apologize that we were not clear enough in the manuscript. Our tool is designed for users who rely on commercial services (i.e., those who cannot include a barcode by themselves). However, it can also benefit those performing library preparation, as SAVEMONEY can be applied after standard barcode-based sequencing and de-multiplexing. The combination of standard barcodes with SAVEMONEY would significantly expands the scope of sequencing applications. For example, it would enable sequencing of more plasmid types than the number of available barcodes and, in some cases, it may even eliminate the need for barcode introduction. Because we do not own ONT equipment and because the primary target audience for the SAVEMONEY algorithm are users without ONT equipment, we were not able to conduct experiments using ONT. However, to clarify these possibilities, we added a dedicated paragraph describing these issues (3rd paragraph in the discussion section).

      However, SAVEMONEY is not able to support the simultaneous analysis of multiple colonies in the same run, as compared to the barcoding-based approaches.

      We agree with the reviewer about this limitation of SAVEMONEY, as it does not allow mixing of plasmids from multiple colonies in the same cloning run. However, that does not necessarily mean that SAVEMONEY cannot reduce sequencing costs in cloning. For example, when sequencing two colonies from each of three diPerent constructs (six plasmids in total), the standard approach would require sequencing costs for six samples. However, with SAVEMONEY, up to three plasmids can be mixed per sample, allowing them to be sequenced as just two samples. As a result, the sequencing cost per plasmid is reduced to one-third. The greatest benefits can be realized when SAVEMONEY is used at the laboratory level or by multiple researchers. To make this point clearer, we have added sentences in the 5th paragraph of the discussion section.

      (1) To provide more comprehensive information for users who care about the cost, the Introduction section should include a cost comparison between Sanger and ONT, with more details, such as diPerent ONT platforms (MinION, PromethION, FlongIe), chemistries (flow cells) and kits. This additional information will be more helpful and informative for the users who have their own sequencers and are the target audience for SAVEMONEY. 

      We thank the reviewer for pointing this out. Since we do not own ONT equipment, we are unable to provide a total cost for using the ONT platform. However, we have included the price per sample (~$15 per plasmid) for the commercial service we have used, as well as the equipment that they employ (V14 chemistry on a PromethION with an R10.4.1 flow cell) and the number of reads obtained per plasmid (~100–1000) in the 4th paragraph of the introduction section.     Though these costs will inevitably change over time, this information should still be helpful for those who own ONT sequencers in estimating the costs.

      (2) In "Overview of the algorithm" (Pages 3-4) under the Results section, instead of stating "However, coverage varies from ~100-1000 and is diPicult to predict because each nanopore flow cell has diPerent properties.", it will be beneficial to provide more detailed information, such as sequencing length, yield/read count per flow cell of diPerent platforms. This information will assist users in designing their own experiments ePectively. 

      We thank the reviewer for the comment. As mentioned in the previous response, we are unable to provide sequencing length, yield/read count per flow cell because we do not own ONT equipment. However, we apologize if it was not clear in "Overview of the algorithm" section that we are discussing the use of results obtained from commercial services, and therefore we need to provide more detailed information about the results from the commercial service. We have now clarified in the sentence pointed out by the reviewr that the numbers are derived from the information provided by commercial sequencing services. In addition, we have also added that typical examples of the result properties, i.e., read length and quality score distribution, can be found in Fig. 2 at the end of the same paragraph.

      (3) While this study optimized and evaluated the tool using a total of 14 plasmids, it may not provide suPicient power to represent the diversity of the plasmid world. Consideration should be given to expanding the dataset to include a broader range of plasmids in future studies to enhance the robustness and generalizability of the tool. 

      We are grateful to the reviewer for their valuable input. It is very reasonable that we had to expect that a larger number of plasmids should be used, even though the main target of SAVEMONEY is those who utilize commercial services. In the previous version of SAVEMONEY, it was not possible to process in a reasonable amount of time if too many plasmids were provided, though the algorithm itself does not have no restrictions based on the number of plasmids. Therefore, we have changed the underlying code to improve the algorithm, making it more than 20 times faster than the previous version (the benchmark time mentioned in the 3rd paragraph of the discussion section was improved to 3.1 minutes from the previous 65 minutes, using the same dataset and the same computer). Additionally, SAVEMONEY is now compatible with multiprocessing. The processing time is expected to decrease approximately inversely proportional to the number of CPU cores used. We have added these updates at the end of the 3rd paragraph in the discussion section.

      (4) If applicable and feasible, including a comparison or benchmark of SAVEMONEY against other similar tools would further strengthen the manuscript. This comparison would allow users to evaluate the advantages and disadvantages of diPerent tools for their specific needs. 

      We thank the reviewer for the suggestion. We have added the benchmark using the similar tool, On-Ramp, with the exact same set of plasmids and FASTQ data used for our benchmark (4th paragraph in the discussion section). Because the machine specifications used in the On-Ramp web server are unknown, a direct comparison is not possible. However, using only laptop-level computational resources, SAVEMONEY was able to process the data 38% faster than On-Ramp. When using mini-PC level computational resources, the processing time was 64% faster than on-RAMP.

      (5) The importance of pre-filtering raw sequencing reads should be emphasized as noisy reads can significantly impact the overall performance of the tool. It is essential to clarify whether any pre-filtering steps were performed in this study, such as filtering based on quality scores, read length, or other relevant factors. 

      We apologize for not being clear. Unfortunately, the commercial sequencing service we used did not provide the information regarding pre-filtering. However, the impact of the quality of pre-filtering based on quality score and read length on the quality of the final results is theoretically minimal in SAVEMONEY. First, during the initial step of the post-analysis, the classification step, short reads compared to the full plasmid length can be excluded based on the user-defined “score_threshold”. Simultaneously, low-quality reads with poor alignment to the plasmid can also be excluded, because “score_threshold” is related to the normalized alignment score. Even if there are low-quality reads that are not excluded at this stage, the ePect can be minimized during the final step of the post-analysis that generates consensus sequences. This is because our Bayesian analysis considers not only the base calling but also the q-scores to determine the consensus. Therefore, we believe the overall impact of pre-filtering on the final results is negligible.

      (6) The statement regarding the number of required reads per plasmid (20-30) and the maximum number of plasmids (up to six) that can be mixed in a single run may become outdated due to the rapid advancements in ONT technology. In the Discussion section, instead of assuming specific numbers, it would be more beneficial to provide information based on the current state of ONT sequencing, such as the number of reads per MinION flow cell that can be produced.

      We thank the reviewer for pointing this out. Because the number of required reads per plasmid depends on the accuracy of each read (i.e., the number of required reads can be reduced if the accuracy increases), we have added the description of these points to the last paragraph of the discussion section.

      Reviewer #2 (public review):  

      The authors developed an algorithm that allows for deconvoluting of plasmid sequences from a mixture of plasmids that have been sequenced by nanopore long read technology. As library preparations and barcoding of individual samples increase sequencing costs, the algorithm bypasses this need and thus decreases time on sample prep and sequencing costs. In the first step, the tool assesses which of the plasmid constructions can be mixed in a single library preparation by calculating a distance matrix between the reference plasmid and the constructions producing sequence clusters. The user is given groups of plasmids, from diPerent clusters, to be pooled together for sequencing. After sequencing, the algorithm deconvolutes the reads by classifying them based on alignments to the reference sequence. A Bayesian analysis approach is used to obtain a consensus sequence and quality scores. 

      Strengths 

      The authors exploit one of the main advantages of long-read sequencing which is to accurately resolve regions of high complexity, as regularly found in plasmids, and developed a tool that can validate plasmid constructions by reducing sequencing costs. Multiple plasmids (up to six) can be analyzed simultaneously in a single library without the need for sample barcoding, also reducing sample preparation time. Although inserts must be diPerent, just 2 bases diPerence would be enough for a correct assignation. It maximizes cost-ePiciency for projects that require large amounts of plasmid constructions and highthroughput validation. 

      We thank the reviewer for the positive response to our manuscript and the helpful comments.

      Weaknesses 

      The method proposed by the authors requires prior knowledge of plasmid sequences (i.e., blueprints or plasmid reference) and is not suitable for small experiments. The plasmid inserts or backbones must be diPerent e.g., multiple colonies from the same plasmid construction ePort cannot be submitted together. 

      As also discussed in the response to reviewer 1, we agree with the reviewer that SAVEMONEY does not allow you the analysis of plasmids from multiple colonies in the same cloning experiment. However, that does not necessarily mean that SAVEMONEY cannot reduce the sequencing cost. For example, when sequencing two colonies from each of three diPerent constructs (six plasmids in total), the standard approach would require sequencing costs for six samples. However, with SAVEMONEY, up to three plasmids can be mixed per sample, allowing them to be sequenced as just two samples. As a result, the sequencing cost per plasmid is reduced to one-third. The greatest benefits can be realized when SAVEMONEY is used at the laboratory level or by multiple researchers. To make this point clearer, we have added sentences in the 5th paragraph of the discussion section.

      The reviewer also expressed concern that SAVEMONEY is not suitable for experiments at a small scale. To put it more precisely, SAVEMONEY cannot be used when the experiment size is minimal, such as in a lab that consistently constructs only a single plasmid at a time. That said, the strength of SAVEMONEY lies in its scalability. Even in labs where plasmid construction is typically limited to one at a time, there may be occasional instances where two or more plasmids are created simultaneously. In such cases, SAVEMONEY can be used to reduce sequencing costs. Moreover, in a typical molecular biology lab where multiple plasmids are constructed every week, SAVEMONEY can be particularly ePective. Given its adaptability and cost-saving potential and widespread use since its initial publication on bioRxiv and on Google Colab, we are confident that SAVEMONEY will continue to be a valuable tool for a wide range of researchers.

      Recommendations For The Authors:

      Reviewer #2 (Recommendations For The Authors): 

      The manucript assumes all samples are sent out for sequencing at a specific company. This could be generalized for a much broader use since many labs now own nanopore sequencers. In turn, the advantage of reducing hands-on sample prep becomes more evident. 

      We thank the reviewer for pointing this out. We agree that SAVEMONEY can also benefit those performing library preparation. Combination of standard barcodes with SAVEMONEY significantly expands the scope of sequencing applications. For example, it enables sequencing of more plasmid types than the number of available barcodes and, in some cases, may even eliminate the need for the sample prep step to introduce barcode. Because we do not own ONT equipment, we could not conduct experiments using ONT. However, to clarify these possibilities, we added a dedicated paragraph (3rd paragraph in the discussion section).

      The base calling model (high accuracy, super accuracy) used by Plasmidsaurus and tested here should be mentioned.  

      We thank the reviewer for the suggestion. The description about the base calling model (HAC) was added in Materials and Methods section.

      Other modifications to the revised manuscript 

      Beyond changes made in response to reviewer comments above, we have also through our continued use and improvement of SAVEMONEY, made additional changes to the algorithm and therefore to the manuscript. Those changes are outlined below. Improvements in the pre-survey step

      (1) The pre-survey algorithm was reduced to a Zero-One Integer Linear Programming Problem to guarantee the optimal combinations, as previous versions did not ensure an optimal solution. Relatedly, the explanation of the algorithm in the main manuscript was updated.

      (2) The algorithm was modified to ensure that the number of plasmids distributed to each group is balanced. A new feature was also added to allow users to specify the number of groups, which is beneficial when balancing between cost and quality.

      (3) An error was corrected in Fig. 2, where the distance calculation method for the hierarchical clustering step for group formation was Farthest Point Algorithm, which calculates distance between two clusters based on the farthest pair of plasmids. The correct method is the Nearest Point Algorithm. This error was present only in Fig. 2, while other implementations, including source code of SAVEMONEY and Google Colab page, were correct from the beginning. We have corrected the error in Fig. 2.

      Modifications in figures, manuscripts, and other aspects

      (1) Fig. 3 was updated to reflect the update of SAVEMONEY, although it did not show any important diPerences.

      (2) Parameter names were updated as follows:

      “threshold (pre)” -> “distance_threshold”

      “threshold (post)” -> “score_threshold” Added “number_of_groups”

      (3) The order of elements was rearranged in Fig. 4.

      (4) Incorrect calculations were fixed in Fig. 4g, h, and i (old Fig. 4d, h, and l). Related to that, Fig. 4j, k, and l and Table 1 were added, in addition to the explanation in the main manuscript.

      (5) SAVEMONEY was packaged and was released on PyPI to facilitate easy installation and integration by other developers.

      (6) SAVEMONEY was updated and expanded to accommodate linear DNA fragments, such as PCR amplicons and long synthetic DNA. Users can select the topology of DNA by specifying that as an option. A description of this new capability was added at the end of “Overview of the algorithm” section.

    1. eLife Assessment

      This useful study reveals that as C. elegans, a poikilothermic ("cold-blooded") animal, adapt to cold (4ºC), they display a drastic reduction in translation (assessed by polysome profiling and SUNSET). The remaining translation (by ribo-seq) correlates with mRNA levels (by RNA-seq), and the changes in gene expression at least partially require IRE-1, an established endoplasmic reticulum stress sensor. The reviewers consider the data assessing global translation and RNA expression upon cold exposure and the data demonstrating the requirement of ire-1 to be solid, but the conclusion that "transcription" is the major regulatory step and "lipid changes" can be a signal for IRE-1 activation in cold adapted worms needs substantially more evidence. Overall, this study demonstrated a good correlation between translation and RNA levels and yielded an inventory of gene changes as C. elegans adapt to cold, and will be of general interest to researchers interested in stress response and cold adaptation.

    2. Reviewer #2 (Public review):

      Summary:

      This study investigates cold induced states in C. elegans, using polysome profiling and RNA seq to identify genes that are differentially regulated and concluding that cold-specific gene regulation occurs at the transcriptional level. This study also includes analysis of one gene from the differentially regulated set, lips-11 (a lipase), and finds that it is regulated in response to a specific set of ER stress factors.

      Strengths:

      (1) Understanding how environmental conditions are linked to stress pathways is generally interesting.<br /> (2) The study used well-established genetic tools to analyze ER stress pathways.

      Weaknesses:

      (1) The conclusions regarding a general transcriptional response are based on a few genes, with much of the emphasis on lips-11, which does not affect survival in response to cold.

      (2) Definitive conclusions regarding transcription vs translational effects would require the use of blockers such as alpha-amanitin or cyclohexamide. Although this may be beyond the scope of the study, it does affect the breadth of the conclusions that can be made.

      (3) Conclusions regarding the role of lipids are based on supplementation with oleic acid or choline, yet there is no lipid analysis of the cold animals, or after lips-1 knockdown. Although choline is important for PC production, adding choline in normal PC could have many other metabolic impacts and doesn't necessarily implicate PC without lipidomic or genetic evidence. Although they note the caveats, their evidence falls short of proving a role in PC production.

    3. Reviewer #3 (Public review):

      Summary:

      The authors sought to understand the molecular mechanisms that cells use to survive cold temperatures by studying gene expression regulation in response to cold in C. elegans. They determined whether gene expression changes during cold adaptation occur primarily at the transcriptional level and identified specific pathways, such as the unfolded protein response pathway, that are activated to possibly promote survival under cold conditions.

      Strengths:

      Effective use of bulk RNA sequencing (RNA-seq) to measure transcript abundance and ribosome profiling (ribo-seq) to assess translation rates, providing a comprehensive view of gene expression regulation during cold adaptation. This combined approach allows for correlation between mRNA levels and their translation, thereby offering evidence for the authors' conclusion that transcriptional regulation is the primary mechanism of cold-specific gene expression changes.

      Weaknesses:

      Many aspects of the weakness have been addressed by the revision. Still, the weak cold sensitivity phenotype observed in ire-1 mutants suggests the ER-UPR pathway's role is likely minor, modulatory or there is an unknown compensatory mechanism responsible for surviving cold.

    4. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1:

      (…) some concerns with interpretations and technical issues make several major conclusions in this manuscript less rigorous, as explained in detail in comments below. In particular, the two major concerns I have: 1) the contradiction between the strong reduction of global translation, with puromycin incorporation gel showing no detectable protein synthesis in cold, and an apparently large fraction of transcripts whose abundance and translation in Fig. 2A are both strongly increased. 2) The fact that no transcripts were examined for dependance on IRE-1/XBP1 for their induction by cold, except for one transcriptional reporter, and some weaknesses (see below) in data showing activation of IRE-1/XBP-1 pathway. The conclusion for induction of UPR by cold via specific activation of IRE-1/XBP-1 pathway, in my opinion, requires additional experiments.

      Relating to the first point, the results of puromycin incorporation and ribosome profiling are not contradictory. The former shows absolute changes in translation, i.e. changes in how much protein the cell is producing, while the latter shows relative changes between the produced proteins, i.e. how the cell prioritizes its protein production. An observed up-regulation in ribosome profiling does not necessarily mean (but could) that the corresponding protein goes up in absolute terms (units produced per time). Instead, it implies that out of the population of all translating ribosomes, a larger fraction is translating (prioritizing) this particular mRNA relative to other mRNAs. The second point is addressed later in the response.

      Major concerns:

      (1) Fig. 1B shows polysomes still present on day 1 of 4ºC exposure, but the gel in Fig. 1C suggests a complete lack of protein synthesis. Why?

      We realized that the selected gel exposure may give the false impression of a complete lack of puromycin incorporation at 4ºC. To avoid confusion, we now show in Figure 1 – figure supplement 1 the original gel image next to its longer exposure. The quantification of puromycin incorporation remains in Fig. 1C (it is based on 3 biological replicates and only one replicate is shown in the corresponding supplement). We hope it is now clear that there is an ongoing puromycin incorporation/translation at 4ºC, albeit much reduced compared with 20ºC.

      What is then the evidence that ribosomal footprints used in much of the paper as evidence of ongoing active translation are from actual translating rather than still bound to transcripts but stationary ribosomes, considering that cooling to 4ºC is often used to 'freeze' protein complexes and prevent separation of their subunits? The authors should explain whether ribosome profiling as a measure of active translation has been evaluated specifically at 4ºC, or test this experimentally.

      While the ribosomal profiling alone might not prove ongoing translation, the residual puromycin incorporation does (see the longer gel exposure in Figure 1 – figure supplement 1). To strengthen this argument, we selected two additional genes (cebp-1 and numr-1) whose ribosomal footprints increase in the cold, and whose GFP-fusions were available from the CGC. Monitoring their expression, we observed the expected increase in the cold (see Figure 2 – figure supplement 3 A-B). The ongoing translation in the cold is also in line with our previous study (Peke et al., 2022), where we observed de novo protein synthesis of other proteins under the same cooling conditions as in this study.

      They should also provide some evidence (like Western blots) of increases in protein levels for at least some of the strongly cold-upregulated transcripts, like lips-11.

      As explained above, we addressed it by additionally examining two strains expressing GFP-fused proteins, whose translation in the cold is predicted to increase according to our ribosomal profiling data. See the new Figure 2 – figure supplement 3 A-B.

      As puromycin incorporation seems to be the one direct measure of global protein synthesis here, it conflicts with much of the translation data, especially considering that quite a large fraction of transcripts have increased both mRNA levels and ribosome footprints, and thus presumably increased translation at 4ºC, in Fig. 2A.

      We hope the above explanations put this concern to rest.

      Also, it is not clear how quantitation in Fig. 1C relates to the gel shown, the quantitation seems to indicate about 50-60% reduction of the signal, while the gel shows no discernable signal.

      A above, see a longer western blot exposure in Figure 1 – figure supplement 1 and note that the quantification is based on three biological replicates.

      (2) It is striking that plips-11::GFP reporter is induced in day 1 of 4ºC exposure, apparently to the extent that is similar to its induction by a large dose of tunicamycin (Fig. 3 supplement),

      We did not intend to compare the extend of induction between cold and tunicamycin treatment. The tunicamycin experiment was meant to confirm that, as suggested by expression data from Shen et al. 2005, lips-11 is upregulated upon UPR activation.

      …but the three IRE-1 dependent UPR transcripts from Shen 2005 list were not induced at all on day 1 (Fig. 4 supplement). Moreover, the accumulation of the misfolded CPL-1 reporter, that was interpreted as evidence that misfolding may be triggering UPR at 4ºC, was only observed on day 1, when the induction of the three IRE-1 targets is absent, but not on day 3, when it is stronger. How does this agree with the conclusion of UPR activation by cold via IRE-1/XBP-1 pathway?

      In the originally submitted supplemental figure, we compared mRNA levels between day 1 animals at 20ºC versus 4ºC. However, as argued later by this reviewer, it may be better to use day 0 animals at 20ºC as the reference (since at 20ºC the animals will continue producing embryos). Thus, we repeated the RT-qPCR analysis with additional time points (and genes relevant to other comments). This analysis, now in Figure 4 – figure supplement 2, shows that these mRNAs (dnj-27, srp-7, and C36B7.6) increased already at day 1 in the cold compared with the reference 20ºC animals on day 0, and their levels increased further on day 3.

      It is true that the authors do note very little overlap between IRE-1/XBP-1-dependent genes induced by different stress conditions, but for most of this paper, they draw parallels between tunicamycin-induced and cold induced IRE-1/XBP-1 activation.

      We carefully re-examined the manuscript to ensure that we do not draw parallels between cold and tunicamycin treatment. The three genes (dnj-27, srp-7, and C36B7.6) were taken from Shen et al. because that study reported lips-11 as an IRE-1-responsive gene, which we realized thanks to the Wormbase annotation of lips-11. Examining the three genes in our expression data, srp-7 (like lips-11) is also upregulated more than 2-fold, while the other two genes go up but less than 2-fold. As mentioned by the reviewer, we note little overlap between the different stress conditions suggesting that the response is context dependent. Additional differences may arise if, as we hypothesize, UPR is activated in the cold in response to both protein and lipid stress. Note that the 2-fold cutoff used in the previous Figure 7 – figure supplement 1 was (erroneously) on the log2 scale, so showed genes upregulated at least 4-fold. We now corrected it to 2-fold. While there are now a few more overlapping genes, the overall conclusion, that there is little overlap between different conditions, did not change. We now list the shared genes in the new Supplementary file 5.

      The conclusion that "the transcription of some cold-induced genes reflects the activation of unfolded protein response (UPR)..." is based on analysis of only one gene, lips-11. No other genes were examined for IRE-1 dependence of their induction by cold, neither the other 8 genes that are common between the cold-induced genes here and the ER stress/IRE-1- induced in Shen 2005 (Venn diagram in Figure 7 supplement), nor the hsp-4 reporter. What is the evidence that lips-11 is not the only gene whose induction by cold in this paper's dataset depends on IRE-1? This is a major weakness and needs to be addressed.

      Furthermore, whether induction by cold of lips-11 itself is due to IRE1 activation was not tested, only a partial decrease of reporter fluorescence by ire-1 RNAi is shown. A quantitative measure of the change of lips-11 transcript in ire-1 and xbp-1 mutants is needed to establish if it depends on IRE-1/XBP-1 pathway.

      We now examined by RT-qPCR if the induction of the three genes from Shen at al. (dnj-27, srp-7, and C36B7.6), as well as lips-11 and hsp-4 depends on IRE-1. In the new Figure 4 – figure supplement 2, we show that the upregulation of all these genes is reduced in the cold in the ire1 mutant (although in the wild type, the increase of hsp-4 mRNA appeared to be non-significant, despite the observed upregulation of the hsp-4 GFP reporter).

      The authors could provide more information and the additional data for the transcripts upregulated by both ER stress and cold, including the endogenous lips-11 and hsp-4 transcripts: their identity, fold induction by both cold and ER stress, how their induction is ranked in the corresponding datasets (all of these are from existing data), and do they depend on IRE-1/XBP-1 for induction by cold?

      As above, the dependence of endogenous lips-11 and hsp-4 on IRE-1 is now shown in the new Figure 4 – figure supplement 2, and the shared genes from Figure 7 – figure supplement 1 are listed in the new Supplementary file 5. We did not perform additional analysis comparing various data sets, as we felt that understanding the differences between IRE-1-mediated transcription outputs across different conditions goes well beyond this study.

      Without these additional data and considering that the authors did not directly measure the splicing of xbp-1 transcript (see comment for Fig. 3 below), the conclusion that cold induces UPR by specific activation of IRE-1/XBP-1 pathway is premature.

      To address the splicing of endogenous xbp-1, we examined our ribosome profiling data for the translation of spliced xbp-1, and found that the spliced variant is more abundant in the cold. This data is now shown in Figure 3 – figure supplement 2B.

      There are also technical issues that are making it difficult to interpret some of the results, and missing controls that decrease the rigor of conclusions:

      (1) For RNAseq and ribosome occupancy, were the 20ºC day 1 adult animals collected at the same time as the other set was moved to 4ºC, or were they additionally grown at 20ºC for the same length of time as the 4ºC incubations, which would make them day 2 adults or older at the time of analysis? This information is only given for SUnSET: "animals were cultivated for 1 or 3 additional days at 4ºC or 20ºC".

      In the RNAseq experiments, the 20ºC animals were collected at the same time as the others were moved to 10ºC (and then 4ºC), so they were not additionally grown at 20ºC. We make it now clear in Methods.

      This could be a major concern in interpreting translation data: First, the inducibility of both UPR and HSR in worms is lost at exactly this transition, from day 1 to day 2 or 3 adults, depending on the reporting lab (for example Taylor and Dillin 2013, Labbadia and Morimoto, 2015, De-Souza et al 2022).

      As explained above, the 20ºC animals were collected at the same time as the others were moved to 4ºC. Then, we reported before that ageing appears to be suppressed in animals incubated at 4ºC (Habacher et al., 2016; Figure S1C). Thus, it terms of their biological age, cold-incubated animals appear to be closer to the 20ºC animals at the time they are moved to the cold (day 0). Thus, the ageing-associated deterioration in UPR inducibility mentioned above presumably does not apply to cold-incubated animals, which is in line with the observed IRE-1-dependent upregulation of several genes in day 3 animals at 4ºC.

      How do authors account for this? Would results with reporter induction, or induction of IRE-1 target genes in Fig. 4, change if day 1 adults were used for 20ºC?

      Our analysis in Figure 4 – figure supplement 2 now includes 20ºC animals at day 0, 1, and 3.

      Second, if animals at the time of shift to 4ºC were only beginning their reproduction, they will presumably not develop further during hibernation, while an additional day at 20ºC will bring them to the full reproductive capacity. Did 4ºC and 20ºC animals used for RNAseq and ribosome occupancy have similar numbers of embryos, and were the embryos at similar stages?

      As explained above, the reference animals at 20ºC were young adults containing few embryos. Indeed, at 4ºC the animals do not accumulate embryos. Although we cannot say that for all genes, note that the genes analysed in Figure 4 – figure supplement 2 increase in abundance also when compared with the day 3 animals kept at 20ºC.

      (2) Second, no population density is given for most of the experiments, despite the known strong effects of crowding (high pheromone) on C. elegans growth. From the only two specifics that are given, it seems that very different population sizes were used: for example, 150 L1s were used in survival assay, while 12,000 L1s in SUnSET. Have the authors compared results they got at high population densities with what would happen when animals are grown in uncrowded plates? At least a baseline comparison in the beginning should have been done.

      None of the experiments involved crowded populations. In the SUnSET experiments, we just used larger and more plates to obtain sufficient material.

      (3) Fig. 3: it is unclear why the accepted and well characterized quantitative measure of IRE1 activation, the splicing of xbp-1transcript, is not determined directly by RT-PCR. The fluorescent XBP-1spliced reporter, to my knowledge, has not been tested for its quantitative nature and thus its use here is insufficient. Furthermore, the image of this fluorescent reporter in Fig. 3b shows only one anterior-most row of cells of intestine, and quantitation was done with 2 to 5 nuclei per animal, while lips-11 is induced in entire intestine. Was there spliced XBP-1 in the rest of the intestinal nuclei? Could the authors show/quantify the entire animal (20 intestinal cells) rather than one or two rows of cells?

      As explained above, we now included the analysis of xbp-1 splicing in Figure 3 – figure supplement 2B. As for the fluorescent reporter, it is difficult to measure all gut nuclei since part of the gut is occluded by the gonad. Nonetheless, we do see induction of the reporter in other gut nuclei and show now additional examples from midgut in Figure 3 – figure supplement 2A.  

      (4) The differences in the outcomes from this study and the previous one (Dudkevich 2022) that used 15ºC to 2ºC cooling approach are puzzling, as they would suggest two quite different IRE-1 dependent programs of cold tolerance. It would be good if authors commented on overlapping/non-overlapping genes, and provided their thoughts on the origin of these differences considering the small difference in temperatures.

      Indeed, there seem to be substantial differences between different temperatures and cooling paradigms. While understanding the C. elegans responses to cold is still in its infancy, one possible explanation for the observed differences is that we used different starting growth temperatures. While the initial populations in our study were grown at 20ºC, Dudkevich et al. used 15ºC. Worms display profound physiological differences between these two temperatures. For example, Xiao et al. (2013) showed that the cold-sensitive TRPA-1 channel is important at 15ºC but not 20ºC. Thus, the trajectories along which worms adapt to near freezing temperature may vary depending on their initial physiological state (and perhaps the target temperature, as we used 4ºC and they 2ºC). We now expanded argumentation on this topic in Discussion. I should also say that we planned on testing NLP-3 function in our paradigm, but our request for strains remained unanswered.

      Second, have the authors performed a control where they reproduced the rescue by FA supplementation of poor survival of ire-1 mutants after the 15ºC to 2ºC shift? Without this or another positive control, and without measuring change in lipid composition in their own experiments, it is unclear whether the different outcomes with respect to FAs are due to a real difference in adaptive programs at these temperatures, or to failure in supplementation?

      While we did not re-examine the findings by Dudkevich et al., we did include now another positive control. As reporter by Hou et al. (2014), supplementing unsaturated FAs rescues the induction of the hsp-4 reporter in fat-6 RNAi-ed animals. Although we were able to reproduce that result (Figure 6 – figure supplement 1), the same supplementation procedure did not suppress the lips11 reporter (Figure 6 – figure supplement 2).

      (5) Have the authors tested whether and by how much ire-1(ok799) mutation shortens the lifespan at 20ºC? This needs to be done before the defect in survival of ire-1 mutants in Fig. 7a can be interpreted.

      The lifespan at standard cultivation temperature was examined by others (Henis-Korenblit et al., 2010; Hourihan et al., 2016), showing that ire-1(ok799) mutants live shorter. However, while some mechanism that prolong lifespan may also improve cold survival, the two phenomena are not identical and whether IRE-1 facilitates longevity and cold survival in the same or different way remains to be seen.

      Reviewer #2:

      (1) The conclusions regarding a general transcriptional response are based on one gene, lips-11, which does not affect survival in response to cold. We would suggest altering the title, to replace "Reprograming gene expression: with" Regulation of the lipase lips-11".

      We now examined IRE-1 dependent induction of additional genes – see Figure 4 – figure supplement 2. While we do not know what fraction of cold-induced genes depends on IRE-1, we feel that our findings justify the statement that that gene expression in the cold involves the IRE1/XBP-1 pathway (title) or that that the transcription of some/a subset of cold-induced genes depend on this pathway (in abstract, model, and discussion).

      (2) There is no gene ontology with the gene expression data.

      We now included the top 10 most enriched and suppressed gene categories between 10ºC and 4ºC (since the biggest change happens between these conditions, as shown in Figure 2 – figure supplement 1A). This is now included in the Figure 2 – figure supplement 2.

      (3) Definitive conclusions regarding transcription vs translational effects would require use of blockers such as alpha amanatin or cyclohexamide.

      As explained also for reviewer 1, we confirmed now that at least some genes, whose translation is upregulated based on the ribosome profiling, are indeed upregulated in the cold at the protein level (Figure 2 – figure supplement 3A-B). Thus, the increase in ribosomal occupancy seems to accurately reflect increased translation. Since mRNA levels correlate overall with the ribosomal occupancy, it appears that the mRNA levels are the main determinants of the translation output. Because the lips-11 promoter is sufficient to upregulate the GFP reporter in the cold, it further suggests that the regulation happens at the transcription level. It is true that at this point we cannot completely rule out the effects of mRNA stability, which we clearly acknowledge in the discussion.

      (4) Conclusions regarding the role of lipids are based on supplementation with oleic acid or choline, yet there is no lipid analysis of the cold animals, or after lips-1 knockdown.

      We agree that this is an important direction for future studies but feel that lipidomic analysis goes beyond the scope of current work.

      Although choline is important for PC production, adding choline in normal PC could have many other metabolic impacts and doesn't necessarily implicate PC without lipidomic or genetic evidence.

      We agree and acknowledge it now in Discussion: “However, choline also plays other roles, including in neurotransmitter synthesis and methylation metabolism. Thus, we cannot yet rule out the possibility that the protective effects of choline supplementation stem from functions outside PC synthesis.”

      Reviewer #3:

      The study has several weaknesses: it provides limited novel insights into pathways mediating transcriptional regulation of cold-inducible genes, as IRE-1 and XBP-1are already well-known responders to endoplasmic reticulum stress, including that induced by cold.

      We presume the reviewer refers to the study by Dudkevich et al. (2022). As explained in our manuscript, there are important differences between that study and ours in how the IRE-1 signalling is utilized and to what ends.

      Additionally, the weak cold sensitivity phenotype observed in ire-1 mutants casts doubt on the pathway's key role in cold adaptation. The study also overlooks previous research (e.g.PMID: 27540856) that links IRE-1 to SKN-1, another major stress-responsive pathway, potentially missing important interactions and mechanisms involved in cold adaptation.

      We state in the manuscript that the IRE-1 pathway plays a modest but significant role in cold adaptation and state in the Fig. 7 model and Discussion that additional pathways work alongside IRE-1 to drive cold-specific gene expression.

      Recommendations for the authors:

      Reviewer #1:

      Minor comments:

      (1) Fig. 2B - reporter expression seems to be already present in the intestine of 20ºC animals. What is the turnover rate of GFP in the intestine and how is it affected by the temperature shift? If GFP degradation is inhibited, could it explain the increase in signal in 4ºC animals, rather than increased transcription? This seems to be true for the hsp-4 transcriptional reporter, as the GFP fluorescence appears to increase during 4ºC incubation (Fig. 4a), but the hsp-4 message levels are only increased after 1 day but not in later days at 4ºC, based on the RNAseq in provided dataset. How well do changes in lips-11 reporter fluorescence correspond to the changes in the endogenous lips-11 transcript?

      Note that increased GFP fluorescence is accompanied by increased mRNA levels. In addition to the RNAseq data, we now also examined changes of the endogenous lips-11 transcript by RTqPCR and observed its strong (and IRE-1 dependent) upregulation in the cold– see Figure 4 – figure supplement 2. Moreover, we now included two other examples of GFP-tagged proteins whose fluorescence increases in the cold, concomitant with increased mRNA levels and ribosomal occupancy (Figure 2 – figure supplement 2A-B).

      (2) Descriptions of methods to measure different aspects of translation are very abbreviated and in some places make it difficult to understand the paper. One example - what is RFP in Fig. 2a?

      We replaced now “RFP” with “RPF” (ribosome protected fragment) and the abbreviation is explained firsts time it is used.

      (3) How was the effectiveness of RNAi at 4ºC validated?

      As explained in Methods, we subjected animals to RNAi long before they were transferred to 4ºC, so the corresponding protein is depleted prior to cooling.

      (4) Several of the conclusions on translation and ribosomal occupancy are written in a somewhat confusing way. For example, the authors state that "shift from 10ºC to 4ºC had a strong effect" when describing "impact on translation (ribosomal occupancy)" (page 4), but in the next sentence, they state "a good correlation between mRNA levels and translation (Figure 2A)". Was ribosomal occupancy normalized to the transcript abundance?

      We do not perceive any discrepancy between the two statements. The former refers to the difference between time points, where we observed the largest change in both the transcriptome and ribosomal occupancy from 10ºC to 4ºC (as can be inferred in the PCA plot in Figure 2 - figure supplement 1). The latter refers to the observation that changes in mRNA levels mirrored, in most of cases, similar changes in the ribosomal occupancy.

      The ribosomal occupancy was not normalized, as that would essentially normalize the y-axis (ribosomal occupancy) with the x-axis (mRNA), and so express changes in “translational efficiency” as a function of changes in mRNA abundance. While this type of analysis can also reveal interesting biological phenomena, it would explore a different question.

      (5) "For most transcripts ... increased the abundance of a particular protein appears to correlate depend primarily on the abundance of its mRNA" (page 5). This is an overstatement, the protein levels were not quantified.

      As explained above, we now additionally monitored the expression of two GFP-tagged proteins (CEBP-1 and NUMR-1). Monitoring their expression, we observed the expected increase in GFP fluorescence in the cold (see Figure 2 – figure supplement 3 A-B). While we did not examine them also by western blot, these observations are in line with our conclusions.

      (6) The statement "Since transcription is the main determinant of mRNA levels, these results suggest that cold-specific gene expression primarily depends on transcription activation" seems to assume that message degradation doesn't have much of an impact at 4ºC. What is the evidence here? The authors themselves later suggest either transcription or mRNA stability in Discussion.

      While we cannot exclude that mRNA stability of some genes may be affected, this concern is more valid for the messages that go down in the cold. Although we have done it for only selected genes, each time we observed an increase in the mRNA levels, we also observed the corresponding increase in the protein; this study and Pekec et al. (2022). Then, the lips-11 reporter was designed to monitor the activity of its promoter, which we showed in sufficient to upregulate reporter GFP in the cold. We have now expanded the corresponding paragraph in Discussion, which will hopefully come across as more balanced.  

      Reviewer #2:

      (1) Alter title, conclusions to better reflect specific nature of the work.

      We now provided additional data and feel that it justifies our conclusions and title.

      (2) Use Gene Ontology searches to look at patterns of gene expression in RNA seq data.

      We now show it in Figure 2 – figure supplement 2.

      (3) Use genetic or lipidomic tools rather than solely adding exogenous lipids.

      We agree that lipidomic analysis is an important direction for future research, but feel that lipidomic analysis and further genetic experiments go beyond the scope of current manuscript.

      Reviewer #3:

      To strengthen the evidence for the role of IRE-1 in cold adaptation, the authors might consider performing additional functional assays, such as testing the effects of IRE-1 and XBP-1 mutations under varying cold conditions and testing the genetic interaction of ire-1 with xbp-1, skn-1, and hsf-1 in cold sensitivities. It is also worth using alternative approaches such as independent alleles of ire-1, knockdowns or tissue-specific knockouts (without potential developmental compensation in global constitutive mutants) to better characterize the contribution of IRE-1 to cold adaptation. Additionally, studies that examine tissue-specific responses to cold exposure could provide important insights, as different tissues may utilize distinct molecular pathways to adapt to cold stress.

      We also tested ire-1 and xbp-1 functions by RNAi-mediated depletion. SKN-1 is a good candidate for future studies, but Horikawa at al. (2024) showed that HSF-1 is not required for cold dormancy (at 4ºC); we also show now that HSF-1::GFP does not increase in the cold (Figure 2 – figure supplement 3C).

      This reviewer also recommends clarifying the novelty of your findings in the context of existing literature, particularly regarding the established roles of IRE-1 and XBP-1 in responding to endoplasmic reticulum stress.

      The entry point of this study was to clarify a long-standing problem in hibernation research, i.e., the apparent discrepancy between a global translation repression and de novo gene expression observed in the cold. By connecting cold-mediated expression of some genes to the IRE-1/XBP1 pathway, we strengthen the argumentation for transcription-mediated gene regulation in hibernating animals. We did go the extra mile to test the possible reason behind the activation of UPR<sup>ER</sup> in the cold but feel that a deeper analysis deserves a separate study.

      The term "hibernation" should be avoided or reworded since the study does not provide direct behavioral or physiological evidence for hibernation-like states; instead, the manuscript could refer to "cold-induced responses" or "adaptations to cold temperatures."

      The term “hibernation” was used before even in the context of the C. elegans dauer state, which, arguably, is even less appropriate. In addition to a global suppression of translation shown here, we reported before that the same cooling regime suppresses ageing (Habacher et al., 2016; Figure S1C). Incubating at 4ºC also arrests C. elegans development (Horikawa et al., 2024). Thus, while the worm and mammalian hibernation are certainly not equivalent – which we clearly spell out – we like to use “hibernation” interchangeably with “cold dormancy” to draw attention to a fascinating aspect of C. elegans biology. Still, we use now quotation marks in the title to avoid misunderstanding.

      The discussion could be strengthened by addressing the relevance of prior studies, such as those linking IRE-1 to SKN-1 (PMID: 27540856), TRPA-1 (PMID: 23415228), ZIP-10 (PMID: 29664006), HSF-1 (PMID: 38987256) in cold adaptation and elaborating on how your findings provide new

      The IRE-1/SKN-1 and ZIP-10 papers are now mentioned when describing the model in Figure 7. The TRP-1 and HSF-1 papers are cited when discussing physiological differences between different cold temperatures. Consistent with our studies, the HSF-1 paper shows that nematodes enter a dormant state at 4ºC (but at 9ºC and higher temperatures continue developing). Importantly, HSF-1 promotes the development at 9ºC but is not important for the arrest at 4ºC. We also shown now in Figure 2 – figure supplement 3C that HSF-1 does not go up at 4ºC.

    1. eLife Assessment

      This manuscript provides important findings for understanding the mechanisms of a major gene causing the gonad of fish and other vertebrates, including mammals, to become an ovary rather than a testis. Evidence is solid, but alternative explanations for a number of the claims must be considered and discussed. The impact of the work would benefit by placing it in a richer historical context.

    2. Reviewer #1 (Public Review):

      The mechanisms that regulate establishment of the germline stem cells and germline progenitors during zebrafish reproductive development are not understood. Prior single cell analysis characterized the cell types of the early zebrafish ovary during and at stages after sexual differentiation. In this work Hsu et al. took a single approach to analyze the cell types present in the early gonad during early sex determination. As expected, they identified germline stem cells (GSCs) that express canonical GSC markers and distinct populations of progenitors. Unexpectedly, they found multiple populations of transcriptionally distinct progenitor populations that the authors termed early (those lacking the differentiation marker foxl2l), committed (those expressing fox2l2 and S-phase genes) and late (those expressing fox2l2 and meiotic genes) progenitors. Comparisons of their dataset to the published zebrafish ovary datasets confirmed the presence of these distinct progenitor populations in the ovary. Further, they convincingly validated the presence of these progenitor subtypes using fluorescent in situ hybridization. To investigate the relationship between progenitor subsets and known regulators of ovary differentiation, the authors conducted single cell analysis of gonads lacking the transcription factor, Foxl2l. As previously reported, Foxl2l absence blocks ovary differentiation and all foxl2l mutants develop testes. The single cell analysis here indicates that foxl2l is inappropriately expressed in GSCs and early progenitors and that germ cell differentiation is blocked at the committed progenitor stage since few committed progenitors and no late progenitors or meiotic transcripts were detected in the single cell analysis of foxl2l mutants. Based on the coexpression of genes that are not typically expressed together in normally developing germ cells, specifically nanos2 and foxl2l, and dmrt1 and foxl2l, the authors conclude that Foxl2l is required for the committed progenitor program and that it prevents committed progenitors from returning to the GSC state.

      Overall, the data provide new insights into the cell populations of the early differentiating gonad, define distinct progenitor states, pinpoint a requirement for the ovary differentiation factor Foxl2l at a specific stage of progenitor differentiation, and generate new hypotheses to be tested. Many but not all of the conclusions are supported by compelling data, and some findings and conclusions need to be clarified in the context of the published literature.

      (1) The authors conclude that the committed progenitors revert to GSCs based on the coexpression of nanos2 and foxl2l nanos2 and based on expression of id1 in mutants but not in WT. Without functional data demonstrating that the progenitors revert to an earlier state, alternative interpretations should be considered. For example, it is possible that the cells initiate the committed progenitor program but continue to express the GSC program and that the coexpression of both programs blocks differentiation. Consistent with this possibility, some Fox family members, FoxL2 and FoxPs for example, are known to be both activators and repressors of transcription or act primarily as repressors. Potentially relevant to this work, repressive activity of FoxL2 has been previously reported in the mammalian ovary (Pisarska et al Endocrinology 2004, Pisarska Am J. Phys Endo. Metabolism 2010, Kuo Reproduction 2012, Kuo Endocrinology 2011, as well as more recent publications). In that context interfering with FoxL2 was proposed to cause upregulated expression of genes normally repressed by FoxL2, accelerated follicle recruitment, and premature ovarian failure.

      (2) The authors conclude that the committed progenitor stage is "the gate toward female determination" and that the cells "stay at S-Phase temporarily before differentiation". This conclusion seems to be based solely on single cell RNAseq expression. In several species, including zebrafish, meiotic entry occurs earlier in females and has been correlated with ovary development. The possibility that the late progenitor stage, the stage when meiotic genes are detected in this study and a stage missing in foxl2l mutants, is actually the key stage for female determination cannot be excluded by the data provided.

      (3) The authors discuss prior working showing that loss of germ cells leads to male development and that germ cells are required for female development and claim to extend that work by showing here that some progenitors are already sexually differentiated. First, the stages compared are completely different. The earlier work looks at the primordial germ cells and their loss in the first few days of development before a gonad forms. In contrast, this work examines stages well after the gonad has formed and during sex determination. The second concern is that the conclusion that the progenitors are differentiated is based solely on the expression of foxl2l, which is initially expressed in the juvenile ovary state that lab strains have been shown to develop through (Wilson et al Front Cell Dev Bio 2024). While it is fair to state that some cells express ovary markers at this stage, it is unclear that this is sufficient evidence that the cells are differentiated. For example, in the context of the foxl2l mutant, the authors observe that GSCs and early progenitors inappropriately express foxl2l, but the mutants develop as males. Thus, expression of foxl2l transcripts alone is insufficient evidence to claim that the cells are already differentiated as female.

      (4) The comparison between medaka and zebrafish foxl2l mutants seems to suggest that Foxl2l is required for meiosis in medaka but has a different role in zebrafish. However, if foxl2l represses the earlier developmental programs of GSCs and early progenitors, it is possible that continued expression of these early programs interferes with activation of meiotic genes. This could account for the absence of the late progenitor stage in foxl2l mutants since the late progenitor stage is defined by and distinguished from the earlier stages by expression of foxl2l and meiotic genes. If so, foxl2l may be similarly required in both systems.

      (5) The authors state that "Foxl2l may ensure female differentiation by preventing stemness and antagonizing male development." It is unclear why suppressing stemness would be necessary for female differentiation since female zebrafish have stem cells as do male zebrafish. It seems likely that turning off the GSC and early differentiation programs is important for allowing expression of meiosis and oocyte differentiation genes, and that a gene other than Foxl2l is required for differentiation from GSCs to spermatocytes.

      (6) Based on its expression in mutant progenitors, p53 is proposed to assist with alternative differentiation of mutant germ cells. Although p53 transcripts are expressed, no evidence is provided that p53 is involved in differentiation of germ cells, and sex bias has not been associated with the published p53 mutants in zebrafish. Furthermore, while p53 has been shown to be important for ovary to testis transformation in mutant contexts in adults, it appears dispensable for testis development in mutants that disrupt ovary differentiation in earlier stages (Rodriguez-Mari et al PLoS Gen 2010, Shive PNAS 2010, Hartung et al Mol. Reprod. Dev 2014, Miao Development 2017, Kaufman et al PLoSGen 2018, Bertho et al Development 2021. It is possible that p53 eliminates foxl2l mutant germ cells that are simultaneously expressing multiple developmental programs, but this possibility would need to be tested.

    3. Reviewer #2 (Public Review):

      In this manuscript, Hsu et al. used scRNA-seq to profile germ cells isolated from zebrafish ovaries. They identified the transcriptional profile of germ cells representing the early stages of oogenesis, from germline stem cells to newly formed follicle stage oocytes. They identified foxl2l as a gene expressed in probable oocyte progenitor cells, one of the least understood germ cell stages in the ovary. To understand to role of Foxl2l in oogenesis, they produced loss-of-function mutations in foxl2l using CRISPR/Cas9. They found that all foxl2l mutants are males as adults, suggesting that Foxl2l is required for oogenesis. To gain more insights, they performed scRNA-seq on cells isolated from 28 dpf foxl2l mutant ovaries and found that in the absence of foxl2l, germ cells appear to arrest as early progenitors. These results argue that Foxl2l, like its medaka homolog Foxl3, is necessary for promoting oocyte vs. spermatocyte differentiation during the oocyte progenitor stage.

    4. Reviewer #3 (Public Review):

      This is the first report to show a transcriptional factor, foxl2l, is essential for the development of female germs. Without foxl2l, germ cells will be developed into sperms. The report also clearly defined the arrested stage of early germ cells in foxl2l mutants, or stages that is critical for foxl2l to play a role for the further development of female germ cells. Due to lack of cell lineage tracing, the claim of foxl2l suppression of dedifferentiate of progenitor cells to GSC based on the gene expression and cell number changes is weak. In addition, separation of early germ cell types in foxl2l mutant using marker genes from WT may not be optimal.

    5. Author response:

      Reviewer #1 (Public Review):

      (1) The authors conclude that the committed progenitors revert to GSCs based on the coexpression of nanos2 and foxl2l nanos2 and based on expression of id1 in mutants but not in WT. Without functional data demonstrating that the progenitors revert to an earlier state, alternative interpretations should be considered. For example, it is possible that the cells initiate the committed progenitor program but continue to express the GSC program and that the coexpression of both programs blocks differentiation.

      Thanks for your insightful comment. We have explored possible alternative interpretations of our data. Regarding the suggested possibility of a continued GSC program in the mutant, we have examined the expression of GSC markers including nanos2 in the mutant at different stages. We found that in the mutant, nanos2 or other GSC markers were not significantly upregulated in GSC-to progenitor transition (G-P) and early progenitors (Prog-E) (Fig. 4B). The expression of these GSC markers was also low in the integrated clusters I4-I6 when G-P and Prog-E stages were prominent (Fig. 3D and Fig. 3E). GSC marker nanos2 was high only in mutant Prog-C. These results argue against continued GSC programs in the foxl2l mutants. Another possible explanation is that perhaps some mutant Prog-C acquires some GSC property with the upregulation of nanos2 instead of a continuous GSC program. We have now clarified our rationale about mutant cells gaining new GSC properties and included both interpretations in the Result.

      Consistent with this possibility, some Fox family members, FoxL2 and FoxPs for example, are known to be both activators and repressors of transcription or act primarily as repressors. Potentially relevant to this work, repressive activity of FoxL2 has been previously reported in the mammalian ovary (Pisarska et al Endocrinology 2004, Pisarska Am J. Phys Endo. Metabolism 2010, Kuo Reproduction 2012, Kuo Endocrinology 2011, as well as more recent publications). In that context interfering with FoxL2 was proposed to cause upregulated expression of genes normally repressed by FoxL2, accelerated follicle recruitment, and premature ovarian failure.

      FoxL2 exerts both activating and repressive activities. We believe that Foxl2l can also activate and repress its target gene expression. Although its target genes have not been clearly identified, Foxl2l may activate genes involved such process as oogenic meiosis, and may also repress other genes involved in other processes, say perhaps nanos2.

      (2) The authors conclude that the committed progenitor stage is "the gate toward female determination" and that the cells "stay at S-Phase temporarily before differentiation". This conclusion seems to be based solely on single cell RNAseq expression. In several species, including zebrafish, meiotic entry occurs earlier in females and has been correlated with ovary development. The possibility that the late progenitor stage, the stage when meiotic genes are detected in this study and a stage missing in foxl2l mutants, is actually the key stage for female determination cannot be excluded by the data provided.

      We agree that Prog-L is important for the initiation of female meiosis. We have made revision in the text to point out the importance of Prog-L in female differentiation.

      (3) The authors discuss prior working showing that loss of germ cells leads to male development and that germ cells are required for female development and claim to extend that work by showing here that some progenitors are already sexually differentiated. First, the stages compared are completely different. The earlier work looks at the primordial germ cells and their loss in the first few days of development before a gonad forms. In contrast, this work examines stages well after the gonad has formed and during sex determination.

      Both previous studies and our study indicate the important role of germ cells in zebrafish sex differentiation during gonadal development. The earlier works show that the abundance of primordial germ cells contributes to sex differentiation. Our current finding further suggests the existence of female identify in some germ cells at the juvenile stage and discusses the importance of cell in sexual differentiation. We have added the developmental age in our study to emphasize the age difference.

      The second concern is that the conclusion that the progenitors are differentiated is based solely on the expression of foxl2l, which is initially expressed in the juvenile ovary state that lab strains have been shown to develop through (Wilson et al Front Cell Dev Bio 2024). While it is fair to state that some cells express ovary markers at this stage, it is unclear that this is sufficient evidence that the cells are differentiated.

      The conclusion about the differentiation of progenitors is not based solely on foxl2l expression; rather, it is according to the whole transcriptomic profiles of both WT (Figure 1B) and foxl2l mutant cells (Figure 3A) as well as the foxl2l mutant phenotype (Figure 2C). Three types of progenitors, Prog-E, Prog-C and Prog-L were identified by whole transcriptomic analysis in WT. In foxl2l mutants, the transcriptomic profile further shows that Prog-L and meiotic cells are completely lost, and all germ cells undergo male differentiation eventually. These results together indicate that the differentiation of Prog-C to Prog-L guides the progenitor toward female differentiation. Our result also showed that in the juvenile gonad, foxl2l expression is high in two types of progenitors, Prog-C and Prog-L, and become low after meiotic entry.

      For example, in the context of the foxl2l mutant, the authors observe that GSCs and early progenitors inappropriately express foxl2l, but the mutants develop as males. Thus, expression of foxl2l transcripts alone is insufficient evidence to claim that the cells are already differentiated as female.

      The foxl2l mutants develop into males because they lack functional Foxl2l. Although the mutated foxl2l transcript is present in mutant cells, these transcripts are not functional. These mutants develop into males eventually. This result is consistent with our claim that functional Foxl2l is important for the development of Prog-L and female differentiation.

      (4) The comparison between medaka and zebrafish foxl2l mutants seems to suggest that Foxl2l is required for meiosis in medaka but has a different role in zebrafish. However, if foxl2l represses the earlier developmental programs of GSCs and early progenitors, it is possible that continued expression of these early programs interferes with activation of meiotic genes. This could account for the absence of the late progenitor stage in foxl2l mutants since the late progenitor stage is defined by and distinguished from the earlier stages by expression of foxl2l and meiotic genes. If so, foxl2l may be similarly required in both systems.

      Medaka and zebrafish Foxl2l may share similar functions such as the stimulation of meiotic gene expression and promotion of oogenesis in the female germ cells preparing for meiotic entry. In addition, we also detected aberrant upregulation of nanos2 in some foxl2l mutant cells. The idea of “continued expression of these early programs interferes with activation of meiotic genes” is conceivable, but for now we have no evidence for it. We do not know whether the absence of meiotic genes is due to an interference caused by the activation of nanos2 or due to the complete loss of Prog-L and meiotic cells. It will also be interesting to find out whether medaka Foxl2l has a role in early progenitors

      (5) The authors state that "Foxl2l may ensure female differentiation by preventing stemness and antagonizing male development." It is unclear why suppressing stemness would be necessary for female differentiation since female zebrafish have stem cells as do male zebrafish. It seems likely that turning off the GSC and early differentiation programs is important for allowing expression of meiosis and oocyte differentiation genes, and that a gene other than Foxl2l is required for differentiation from GSCs to spermatocytes.

      It is true that we have not proved whether suppression of stemness is required for female differentiation. Maybe our earlier statement is a bit misleading. We agree that it is likely that turning off the GSC and early differentiation programs is important for allowing expression of meiotic and oocyte differentiation genes, and that a gene other than Foxl2l is required for differentiation from GSCs to spermatocytes. To avoid confusion, we have modified our statement in the text.

      (6) Based on its expression in mutant progenitors, p53 is proposed to assist with alternative differentiation of mutant germ cells. Although p53 transcripts are expressed, no evidence is provided that p53 is involved in differentiation of germ cells, and sex bias has not been associated with the published p53 mutants in zebrafish. Furthermore, while p53 has been shown to be important for ovary to testis transformation in mutant contexts in adults, it appears dispensable for testis development in mutants that disrupt ovary differentiation in earlier stages (Rodriguez-Mari et al PLoS Gen 2010, Shive PNAS 2010, Hartung et al Mol. Reprod. Dev 2014, Miao Development 2017, Kaufman et al PLoSGen 2018, Bertho et al Development 2021. It is possible that p53 eliminates foxl2l mutant germ cells that are simultaneously expressing multiple developmental programs, but this possibility would need to be tested.

      The tp53<sup>-/-</sup>foxl2l<sup>-/-</sup> double mutant cannot alleviate the all-male phenotype of foxl2l<sup>-/-</sup> mutant (Dev Biol, 517, 91-99, 2024), indicating that the male development is not due to p53-mediated germ cell apoptosis. We have cited the suggested papers and compared relation of tp53 between these mutants (fancl, zar1, etc.) mentioned in the cited papers. Since tp53 was enriched in certain foxl2l<sup>-/-</sup> mutant cell clusters, and tp53 mutation fails to rescue the all-male phenotype, it is possible that p53 expressed in these mutant cell clusters has roles other than inducing apoptosis. One assumption is that p53 may be involved in the germ cell differentiation, especially p53 is known to promote differentiation of airway epithelial progenitors, adipogenesis and embryonic stem cells. We have emphasized that the suggested role of p53 in germ cell differentiation is our assumption in the Discussion.

      Reviewer #3 (Public Review):

      This is the first report to show a transcriptional factor, foxl2l, is essential for the development of female germs. Without foxl2l, germ cells will be developed into sperms. The report also clearly defined the arrested stage of early germ cells in foxl2l mutants, or stages that is critical for foxl2l to play a role for the further development of female germ cells.

      (1) Due to lack of cell lineage tracing, the claim of foxl2l suppression of dedifferentiate of progenitor cells to GSC based on the gene expression and cell number changes is weak.

      Thanks for your comments pointing out our contribution and also weakness. We acknowledge the lack of direct evidence on the reversion of mutant Prog-C to GSC in our data. We now removed the claim about the repression of stemness by Foxl2l.

      (2) In addition, separation of early germ cell types in foxl2l mutant using marker genes from WT may not be optimal.

      The cell type of mutant cell is determined by two independent analyses. First is inferring the developmental stage of mutant cells. This approach assumes that mutant cells can indeed be mapped to specific WT stages through their transcriptomic profiles. However, as indicated by this reviewer’s comments, mutant cells exhibited heterogeneity and can be distinct from WT cells. Defining cell types in mutants by WT markers may not be optimal. To address this, we conducted another analysis, co-clustering. Mutant cells and WT cells at early stages (GSC , G-P, Prog-E, Prog-C(S) and Prog-C) were co-clustered. This approach does not assume a direct correspondence between mutant and WT developmental stages. Instead, it facilitates the identification of novel germ cell types in mutants while characterizing the relationship between WT and mutant cells. In some clusters, both WT and mutant cells were present, indicating high transcriptomic similarity. In other clusters, most cells are only mutant cells, indicating distinct mutant cell types (Figure 3C). We can, therefore, assign developmental properties to these mutant cells with confidence.

    1. eLife Assessment

      This useful study provides convincing evidence that Drosophila can taste cholesterol through a subset of bitter-sensing gustatory receptor neurons, and that flies avoid high-cholesterol food. However, the same receptors have been previously found to be involved in the detection of multiple seemingly unrelated chemicals, and the reported expression patterns of these receptors contradict past reports. These caveats are not mentioned in the paper, raising critical concerns about the study's conclusions.

    2. Reviewer #1 (Public review):

      Summary:

      Pradhan et al investigated the potential gustatory mechanisms that allow flies to detect cholesterol. They found that flies are indifferent to low cholesterol and avoid high cholesterol. They further showed that the ionotropic receptors Ir7g, Ir51b, and Ir56d are important for the cholesterol sensitivity in bitter neurons. The figures are clear and the behavior result is interesting. However, I have several major comments, especially on the discrepancy of the expression of these Irs with other lab published results, and the confusing finding that the same receptors (Ir7g, Ir51b) have been implicated in the detection of various seemingly unrelated compounds.

      Strengths:

      The results are very well presented, the figures are clear and well-made, text is easy to follow.

      Weaknesses:

      (1) Regarding the expression of Ir56d. The reported Ir56d expression pattern contradicts multiple previous studies (Brown et al., 2021 eLife, Figure 6a-c; Sanchez-Alcaniz et al., 2017 Nature Communications, Figure 4e-h; Koh et al., 2014 Neuron, Figure 3b). These studies, using three different driver lines, consistently showed Ir56d expression in sweet-sensing neurons and taste peg neurons. Importantly, Sanchez-Alcaniz et al. demonstrated that Ir56d is not expressed in Gr66a-expressing (bitter) neurons. This discrepancy is critical since Ir56d is identified as the key subunit for cholesterol detection in bitter neurons, and misexpression of Ir7g and Ir51b together is insufficient to confer cholesterol sensitivity (Fig.4b,d). Which Ir56d-GAL4 (and Gr66a-I-GFP) line was used in this study? Is there additional evidence (scRNA sequencing, in-situ hybridization, or immunostaining) supporting Ir56d expression in bitter neurons?

      (2) Ir51b has previously been implicated in detecting nitrogenous waste (Dhakal 2021), lactic acid (Pradhan 2024), and amino acids (Aryal 2022), all by the same lab. Additionally, both Ir7g and Ir51b have been implicated in detecting cantharidin, an insect-secreted compound that flies may or may not encounter in the wild, by the same lab. Is Ir51b proposed to be a specific receptor for these chemically distinct compounds or a general multimodal receptor for aversive stimuli? Unlike other multimodal bitter receptors, the expression level of Ir51b is rather low and it's unclear which subset of GRNs express this receptor. The chemical diversity among nitrogenous waste, amino acids, lactic acid, cantharidin, and cholesterol raises questions about the specificity of these receptors and warrants further investigation and at a minimum discussion in this paper. Given the wide and seemingly unrelated sensitivity of Ir51b and Ir7g to these compounds I'm leaning towards the hypothesis that at least some of these is non-specific and ecologically irrelevant without further supporting evidence from the authors.

      (3) The Benton lab Ir7g-GAL4 reporter shows no expression in adults. Additionally, two independent labellar RNA sequencing studies (Dweck, 2021 eLife; Bontonou et al., 2024 Nature Communications) failed to detect Ir7g expression in the labellum. This contradicts the authors' previous RT-PCR results (Pradhan 2024 Fig. S4, Journal of Hazardous Materials) showing Ir7g expression in the labellum. Additionally the Benton and Carlson lab Ir51b-GAL4 reporters show no expression in adults as well. Please address these inconsistencies.

      (4) The premise that high cholesterol intake is harmful to flies, which makes sensory mechanisms for cholesterol avoidance necessary, is interesting but underdeveloped. Animal sensory systems typically evolve to detect ecologically relevant stimuli with dynamic ranges matching environmental conditions. Given that Drosophila primarily consume fruits and plant matter (which contain minimal cholesterol) rather than animal-derived foods (which contain higher cholesterol), the ecological relevance of cholesterol detection requires more thorough discussion. Furthermore, at high concentrations, chemicals often activate multiple receptors beyond those specifically evolved for their detection. If the cholesterol concentrations used in this study substantially exceed those encountered in the fly's natural diet, the observed responses may represent an epiphenomenon rather than an ecologically and ethologically relevant sensory mechanism. What is the cholesterol content in flies' diet and how does that compare to the concentrations used in this paper?

    3. Reviewer #2 (Public review):

      Summary:

      In Cholesterol Taste Avoidance in Drosophila melanogaster, Pradhan et al. used behavioral and electrophysiological assays to demonstrate that flies can: (1) detect cholesterol through a subset of bitter-sensing gustatory receptor neurons (GRNs) and (2) avoid consuming food with high cholesterol levels. Mechanistically, they identified five members of the IR family as necessary for cholesterol detection in GRNs and for the corresponding avoidance behavior. Ectopic expression experiments further suggested that Ir7g + Ir56d or Ir51b + Ir56d may function as tuning receptors for cholesterol detection, together with the Ir25a and Ir76b co-receptors.

      Strengths:

      The experimental design of this study was logical and straightforward. Leveraging their expertise in the Drosophila taste system, the research team identified the molecular and cellular basis of a previously unrecognized taste category, expanding our understanding of gustation. A key strength of the study was its combination of electrophysiological recordings with behavioral genetic experiments.

      Weaknesses:

      My primary concern with this study is the lack of a systematic survey of the IRs of interest in the labellum GRNs. Consequently, there is no direct evidence linking the expression of putative cholesterol IRs to the B GRNs in the S6 and S7 sensilla.

      Specifically, the authors need to demonstrate that the IR expression pattern explains cholesterol sensitivity in the B GRNs of S6 and S7 sensilla, but not in other sensilla. Instead of providing direct IR expression data for all candidate IRs (as shown for Ir56d in Figure 2-figure supplement 1F), the authors rely on citations from several studies (Lee, Poudel et al. 2018; Dhakal, Sang et al. 2021; Pradhan, Shrestha et al. 2024) to support their claim that Ir7g, Ir25a, Ir51b, and Ir76b are expressed in B GRNs (Lines 192-194). However, none of these studies provide GAL4 expression or in situ hybridization data to substantiate this claim.

      Without a comprehensive IR expression profile for GRNs across all taste sensilla, it is difficult to interpret the ectopic expression results observed in the B GRN of the I9 sensillum or the A GRN of the L-sensillum (Figure 4). It remains equally plausible that other tuning IRs-beyond the co-receptor Ir25a and Ir76b-could interact with the ectopically expressed IRs to confer cholesterol sensitivity, rather than the proposed Ir7g + Ir56d or Ir51b + Ir56d combinations.

    4. Reviewer #3 (Public review):

      Summary:

      Whether and how animals can taste cholesterol is not well understood. The study provides evidence that 1) cholesterol activates a subset of bitter-sensing gustatory receptor neurons (GRNs) in the fly labellum, but not other types of GRNs, 2) flies show aversion to high concentrations of cholesterol, and this is mediated by bitter GRNs, and 3) cholesterol avoidance depends on a specific set of ionotropic receptor (IR) subunits acting in bitter GRNs. The claims of the study are supported by electrophysiological recordings, genetic manipulations, and behavioral readouts.

      Strengths:

      Cholesterol taste has not been well studied, and the paper provides new insight into this question. The authors took a comprehensive and rigorous approach in several different parts of the paper, including screening the responses of all 31 labellar sensilla, screening a large panel of receptor mutants, and performing misexpression experiments with nearly every combination of the 5 IRs identified. The effects of the genetic manipulations are very clear and the results of electrophysiological and behavioral studies match nicely, for the most part. The appropriate controls are performed for all genetic manipulations.

      Weaknesses:

      The weaknesses of the study, described below, are relatively minor and do not detract from the main conclusions of the paper.

      (1) The paper does not state what concentrations of cholesterol are present in Drosophila's natural food sources. Are the authors testing concentrations that are ethologically relevant?

      (2) The paper does not state or show whether the expression of IR7g, IR51b, and IR56d is confined to bitter GRNs. Bitter-specific expression of at least some of these receptors would be necessary to explain why bitter GRNs but not sugar GRNs (or other GRN types) normally show cholesterol responses.

      (3) The authors only investigated the responses of GRNs in the labellum, but GRN responses in the leg may also contribute to the avoidance of cholesterol feeding. Alternatively, leg GRNs might contribute to cholesterol attraction that is unmasked when bitter GRNs are silenced. In support of this possibility, Ahn et al. (2017) showed that Ir56d functions in sugar GRNs of the leg to promote appetitive responses to fatty acids.

      (4) The authors might consider using proboscis extension as an additional readout of taste attraction or aversion, which would help them more directly link the labellar GRN responses to a behavioral readout. Using food ingestion as a readout can conflate the contribution of taste with post-ingestive effects, and the regulation of food ingestion also may involve contributions from GRNs on multiple organs, whereas organ-specific contributions can be dissociated using proboscis extension. For example, does presenting cholesterol on the proboscis lead to aversive responses in the proboscis extension assay (e.g., suppression of responses to sugar)? Does this aversion switch to attraction when bitter GRNs are silenced, as with the feeding assay?

      (5) The authors claim that the cholesterol receptor is composed of IR25a, IR76b, IR56d, and either IR7g or IR51b. While the authors have shown that IR25a and IR76b are each required for cholesterol sensing, they did not show that both are required components of the same receptor complex. If the authors are relying on previous studies to make this assumption, they should state this more clearly. Otherwise, I think further misexpression experiments may be needed where only IR25a or IR76b, but not both, are expressed in GRNs.

    5. Author response:

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      Pradhan et al investigated the potential gustatory mechanisms that allow flies to detect cholesterol. They found that flies are indifferent to low cholesterol and avoid high cholesterol. They further showed that the ionotropic receptors Ir7g, Ir51b, and Ir56d are important for the cholesterol sensitivity in bitter neurons. The figures are clear and the behavior result is interesting. However, I have several major comments, especially on the discrepancy of the expression of these Irs with other lab published results, and the confusing finding that the same receptors (Ir7g, Ir51b) have been implicated in the detection of various seemingly unrelated compounds.

      Strengths:

      The results are very well presented, the figures are clear and well-made, text is easy to follow.

      Weaknesses:

      (1) Regarding the expression of Ir56d. The reported Ir56d expression pattern contradicts multiple previous studies (Brown et al., 2021 eLife, Figure 6a-c; Sanchez-Alcaniz et al., 2017 Nature Communications, Figure 4e-h; Koh et al., 2014 Neuron, Figure 3b). These studies, using three different driver lines, consistently showed Ir56d expression in sweet-sensing neurons and taste peg neurons. Importantly, Sanchez-Alcaniz et al. demonstrated that Ir56d is not expressed in Gr66a-expressing (bitter) neurons. This discrepancy is critical since Ir56d is identified as the key subunit for cholesterol detection in bitter neurons, and misexpression of Ir7g and Ir51b together is insufficient to confer cholesterol sensitivity (Fig.4b,d). Which Ir56d-GAL4 (and Gr66a-I-GFP) line was used in this study? Is there additional evidence (scRNA sequencing, in-situ hybridization, or immunostaining) supporting Ir56d expression in bitter neurons?

      We agree that the expression pattern of Ir56d diverges from two prior reports . The studies by Brown et al. and Koh et al. employed the same Ir56d-GAL4 driver line, which exhibited expression in sweet-sensing gustatory receptor neurons (GRNs) and taste peg neurons, but not bitter GRNs (the Sanchez-Alcaniz et al. paper did not use an Ir56d-Gal4).

      In our study, we used a Ir56d-GAL4 driver line (KDRC:2307) and the Gr66a-I-GFP reporter line (Weiss et al., 2011 Neuron). This is a crucial distinction, as differences in the regulatory regions used to generate different driver lines are well known to underlie differences in expression patterns. Our double-labeling experiments revealed co-expression of Ir56d with Gr66a-positive bitter GRNs specifically within the S6 and S7 sensilla—types previously shown to exhibit strong electrophysiological responses to cholesterol (Figure 2—figure supplement 1F).

      We believe this observation is biologically significant and consistent with our functional data. Specifically, targeted expression of Ir56d in bitter neurons using the Gr33a-GAL4 was sufficient to rescue cholesterol avoidance behavior in Ir56d<sup>1</sup> mutants (Figure 3G). These results demonstrate that Ir56d plays a functional role in bitter GRNs for cholesterol detection. The convergence of genetic, behavioral, and electrophysiological data presented in our study provides compelling support for this previously unappreciated expression pattern and function of Ir56d.

      (2) Ir51b has previously been implicated in detecting nitrogenous waste (Dhakal 2021), lactic acid (Pradhan 2024), and amino acids (Aryal 2022), all by the same lab. Additionally, both Ir7g and Ir51b have been implicated in detecting cantharidin, an insect-secreted compound that flies may or may not encounter in the wild, by the same lab. Is Ir51b proposed to be a specific receptor for these chemically distinct compounds or a general multimodal receptor for aversive stimuli? Unlike other multimodal bitter receptors, the expression level of Ir51b is rather low and it's unclear which subset of GRNs express this receptor. The chemical diversity among nitrogenous waste, amino acids, lactic acid, cantharidin, and cholesterol raises questions about the specificity of these receptors and warrants further investigation and at a minimum discussion in this paper. Given the wide and seemingly unrelated sensitivity of Ir51b and Ir7g to these compounds I'm leaning towards the hypothesis that at least some of these is non-specific and ecologically irrelevant without further supporting evidence from the authors.

      While it is true that IR51b and IR7g are responsive to a range of compounds, they share chemical features such as nitrogen-containing groups, hydrophobicity, or amphipathic structures suggesting that recognition of these chemicals may be mediated by the same or overlapping domains within the receptor complexes. These features could facilitate binding to a structurally diverse yet chemically related groups of aversive ligands.

      In the case of cholesterol, while its sterol ring system is distinct from the other compounds, it shares hydrophobic and amphipathic properties that may enable interaction with these receptors via similar structural motifs. Importantly, our data demonstrates that Ir51b and Ir7g are necessary but not sufficient on their own to confer cholesterol sensitivity, indicating that additional co-factors or receptor subunits are required for full functionality (Figure 4B, D). Furthermore, our dose-response analysis (Figure 3F) shows that Ir7g is particularly important at higher cholesterol concentrations, supporting the idea of graded sensitivity rather than indiscriminate activation. This suggests that these receptors may have evolved to recognize cholesterol and its analogs (e.g., phytosterols such as stigmasterol, yet to be tested), which are naturally found in the fly’s diet (e.g., yeast and plant-derived matter), as ecologically relevant cues signaling microbial contamination, lipid imbalance, or dietary overconsumption.

      We acknowledge the reviewer’s concern regarding the relatively low expression levels of Ir51b and Ir7g. However, we note that low transcript abundance does not necessarily equate to diminished physiological relevance. Finally, we agree that the chemical diversity of ligands associated with Ir51b and Ir7g warrants deeper investigation, particularly through structure-function studies aimed at identifying ligand-binding domains and receptor-ligand interactions at atomic resolution.

      (3) The Benton lab Ir7g-GAL4 reporter shows no expression in adults. Additionally, two independent labellar RNA sequencing studies (Dweck, 2021 eLife; Bontonou et al., 2024 Nature Communications) failed to detect Ir7g expression in the labellum. This contradicts the authors' previous RT-PCR results (Pradhan 2024 Fig. S4, Journal of Hazardous Materials) showing Ir7g expression in the labellum. Additionally the Benton and Carlson lab Ir51b-GAL4 reporters show no expression in adults as well. Please address these inconsistencies.

      With respect to Ir7g, we acknowledge that the Ir7g-GAL4 reporter line from the Benton lab does not exhibit detectable expression in adult labella. Furthermore, two independent transcriptomic studies—Dweck et al., 2021 (eLife) and Bontonou et al., 2024 (Nature Communications) also did not detect Ir7g transcripts in bulk RNA-seq datasets derived from adult labella. However, our previously published RT-PCR data (Pradhan et al., 2024, Journal of Hazardous Materials, Fig. S4) revealed Ir7g expression in labellar tissue, albeit at low levels. Our RT-PCR includes an internal control (tubulin) with the same reaction tube with control and the Ir7g mutant as a negative control. Therefore, we stand behind the findings that Ir7g is expressed in the labellum.

      We would like to point out that RT-PCR is more sensitive and better-suited to detect low-abundance transcripts than bulk RNA-seq, which may fail to capture transcripts due to limitations in depth of coverage. Moreover, immunohistochemistry can have limitations in detecting very low expression levels. Costa et al. 2013 (Translational lung cancer research) states that “RNA-Seq technique will not likely replace current RT-PCR methods, but will be complementary depending on the needs and the resources as the results of the RNA-Seq will identify those genes that need to then be examined using RT-PCR methods”.

      Similarly, regarding Ir51b, while the GAL4 reporter lines from the Benton and Carlson labs do not show robust adult expression, our RT-PCR and functional data strongly support a role for Ir51b in labellar bitter GRNs. Specifically, Ir51b<sup>1</sup> mutants display electrophysiological deficits in response to cholesterol (Figure 2A–B), and these defects are rescued by expressing Ir51b in Gr33a-positive bitter neurons (Figure 3G), providing functional validation of the RT-PCR expression.

      (4) The premise that high cholesterol intake is harmful to flies, which makes sensory mechanisms for cholesterol avoidance necessary, is interesting but underdeveloped. Animal sensory systems typically evolve to detect ecologically relevant stimuli with dynamic ranges matching environmental conditions. Given that Drosophila primarily consume fruits and plant matter (which contain minimal cholesterol) rather than animal-derived foods (which contain higher cholesterol), the ecological relevance of cholesterol detection requires more thorough discussion. Furthermore, at high concentrations, chemicals often activate multiple receptors beyond those specifically evolved for their detection. If the cholesterol concentrations used in this study substantially exceed those encountered in the fly's natural diet, the observed responses may represent an epiphenomenon rather than an ecologically and ethologically relevant sensory mechanism. What is the cholesterol content in flies' diet and how does that compare to the concentrations used in this paper?

      Drosophila melanogaster cannot synthesize sterols de novo, and must acquire them from its diet. In natural environments, flies acquire sterols from fermenting fruit, decaying plant matter, and yeast, which contain trace amounts of phytosterols (e.g., stigmasterol, β-sitosterol) and ergosterol. While the exact sterol concentrations in these sources remain uncharacterized, our behavioral assays used concentrations (0.001–0.01% by weight) that align with the low levels expected in such nutrient-limited ecological niches.

      In our study, the cholesterol concentrations tested ranged from 0.001% to 0.1%, thereby spanning both the physiologically relevant and slightly elevated range. Importantly, avoidance behaviors and receptor activation were most prominent at 0.1% cholesterol. While it is true that high chemical concentrations may elicit off-target effects via broad receptor activation, our genetic and electrophysiological data indicate that the observed responses are mediated by specific ionotropic receptors (Ir51b, Ir7g, Ir56d) and not merely generalized chemical stress.

      Ecologically, elevated sterol levels may also signal conditions unsuitable for egg-laying or larval development. For example, high levels of cholesterol or other sterols may occur in substrates colonized by pathogenic microbes, decaying animal tissue, or in cases of abnormal microbial fermentation, which could represent a nutritional or microbial hazard. The avoidance of cholesterol may help signal the flies to avoid consuming decaying animal tissue. In this context, sensory detection of excessive cholesterol might serve as a protective function.

      Reviewer #2 (Public review):

      Summary:

      In Cholesterol Taste Avoidance in Drosophila melanogaster, Pradhan et al. used behavioral and electrophysiological assays to demonstrate that flies can: (1) detect cholesterol through a subset of bitter-sensing gustatory receptor neurons (GRNs) and (2) avoid consuming food with high cholesterol levels. Mechanistically, they identified five members of the IR family as necessary for cholesterol detection in GRNs and for the corresponding avoidance behavior. Ectopic expression experiments further suggested that Ir7g + Ir56d or Ir51b + Ir56d may function as tuning receptors for cholesterol detection, together with the Ir25a and Ir76b co-receptors.

      Strengths:

      The experimental design of this study was logical and straightforward. Leveraging their expertise in the Drosophila taste system, the research team identified the molecular and cellular basis of a previously unrecognized taste category, expanding our understanding of gustation. A key strength of the study was its combination of electrophysiological recordings with behavioral genetic experiments.

      Weaknesses:

      My primary concern with this study is the lack of a systematic survey of the IRs of interest in the labellum GRNs. Consequently, there is no direct evidence linking the expression of putative cholesterol IRs to the B GRNs in the S6 and S7 sensilla.

      Specifically, the authors need to demonstrate that the IR expression pattern explains cholesterol sensitivity in the B GRNs of S6 and S7 sensilla, but not in other sensilla. Instead of providing direct IR expression data for all candidate IRs (as shown for Ir56d in Figure 2-figure supplement 1F), the authors rely on citations from several studies (Lee, Poudel et al. 2018; Dhakal, Sang et al. 2021; Pradhan, Shrestha et al. 2024) to support their claim that Ir7g, Ir25a, Ir51b, and Ir76b are expressed in B GRNs (Lines 192-194). However, none of these studies provide GAL4 expression or in situ hybridization data to substantiate this claim.

      Without a comprehensive IR expression profile for GRNs across all taste sensilla, it is difficult to interpret the ectopic expression results observed in the B GRN of the I9 sensillum or the A GRN of the L-sensillum (Figure 4). It remains equally plausible that other tuning IRs-beyond the co-receptor Ir25a and Ir76b-could interact with the ectopically expressed IRs to confer cholesterol sensitivity, rather than the proposed Ir7g + Ir56d or Ir51b + Ir56d combinations.

      We provide electrophysiological data demonstrating that the S6 and S7 sensilla respond to cholesterol (Figure 1D). This finding is consistent with the hypothesis that these sensilla harbor the complete receptor complexes necessary for cholesterol detection. In our electrophysiological recordings, only those bitter GRNs that co-express Ir56d along with either Ir7g or Ir51b generate action potentials in response to cholesterol. Other S-type sensilla lacking one or more of these subunits remain unresponsive, reinforcing the idea that these components are necessary for receptor function and sensory coding of cholesterol. Moreover, in the cholesterol-insensitive I9 sensillum (based on our mapping results using electrophysiology), co-expression of either Ir7g + Ir56d or Ir51b + Ir56d conferred de novo cholesterol sensitivity (Figure 4B). Importantly, no cholesterol response was observed when any of these IRs was expressed alone or when Ir7g + Ir51b were co-expressed without Ir56d. These findings strongly argue against the possibility that endogenous tuning IRs in I9 sensilla (e.g., Ir25a, Ir76b) are sufficient to generate cholesterol responsiveness.

      Furthermore, based on the literature, Ir25a and Ir76b are endogenously expressed in I- and L-type sensilla. Thus, their presence alone is insufficient for cholesterol responsiveness. These data support the model that cholesterol sensitivity depends on a specific, multi-subunit receptor complex (e.g., Ir7g + Ir25a + Ir56d + Ir76b or Ir51b + Ir25a + Ir56d + Ir76b).

      In conclusion, while we acknowledge that our data do not provide a full anatomical map of IR expression across all sensilla, our results strongly support the idea that cholesterol sensitivity in S6 and S7 sensilla arises from specific combinations of IRs expressed in the B GRNs.

      Reviewer #3 (Public review):

      Summary:

      Whether and how animals can taste cholesterol is not well understood. The study provides evidence that 1) cholesterol activates a subset of bitter-sensing gustatory receptor neurons (GRNs) in the fly labellum, but not other types of GRNs, 2) flies show aversion to high concentrations of cholesterol, and this is mediated by bitter GRNs, and 3) cholesterol avoidance depends on a specific set of ionotropic receptor (IR) subunits acting in bitter GRNs. The claims of the study are supported by electrophysiological recordings, genetic manipulations, and behavioral readouts.

      Strengths:

      Cholesterol taste has not been well studied, and the paper provides new insight into this question. The authors took a comprehensive and rigorous approach in several different parts of the paper, including screening the responses of all 31 labellar sensilla, screening a large panel of receptor mutants, and performing misexpression experiments with nearly every combination of the 5 IRs identified. The effects of the genetic manipulations are very clear and the results of electrophysiological and behavioral studies match nicely, for the most part. The appropriate controls are performed for all genetic manipulations.

      Weaknesses:

      The weaknesses of the study, described below, are relatively minor and do not detract from the main conclusions of the paper.

      (1) The paper does not state what concentrations of cholesterol are present in Drosophila's natural food sources. Are the authors testing concentrations that are ethologically Drosophila melanogaster primarily feeds on fermenting fruits and associated microbial communities, especially yeast, which serve as major sources of dietary sterols. These natural food sources are known to contain phytosterols such as stigmasterol and β-sitosterol. One study quantified phytosterols (e.g., stigmasterol, sitosterol) in fruits, reporting concentrations between 1.6–32.6 mg/100 g edible portion (~0.0016–0.0326% wet weight) (Han et al 2008). The range we tested falls within this range. Additionally, ergosterol, the principal sterol in yeast and a structural analog of cholesterol, is present at levels of about 0.005% to 0.02% in yeast-rich environments.

      To ensure physiological relevance, we designed our behavioral assays to include a broad concentration range of cholesterol, from 10<sup>-5</sup>% to 10<sup>-1</sup>%. This spans both physiological levels (0.001–0.01%), which are comparable to those found in the natural diet, and supra-physiological levels (e.g., 0.1%), which exceed natural exposure but help define the threshold for aversive behavior.

      Our results demonstrate that flies begin to avoid cholesterol at concentrations ≥10<sup>-3</sup>% more (Figure 3A), which falls within the upper physiological range and may reflect the threshold beyond which cholesterol or related sterols become deleterious. At these higher concentrations, excess sterols may disrupt membrane fluidity, interfere with hormone signaling, or promote microbial overgrowth—all of which could compromise fly health.

      (2) The paper does not state or show whether the expression of IR7g, IR51b, and IR56d is confined to bitter GRNs. Bitter-specific expression of at least some of these receptors would be necessary to explain why bitter GRNs but not sugar GRNs (or other GRN types) normally show cholesterol responses.

      We show the Ir56d-Gal4 is co-expressed with Gr66a-GFP in S6/S7 sensilla, indicating that it is expressed in bitter GRNs (Figure 2—figure supplement 1F). In the case of Ir7g and Ir51b, there are no reporters or antibodies to address expression. However, previously they have been shown to be expressed in bitter GRNs using RT-PCR (Dhakal et al. 2021, Communications Biology; Pradhan et al. 2024, Journal of Hazardous Materials). In addition, we provide functional evidence that bitter GRNs are required for the cholesterol response since silencing bitter GRNs abolishes cholesterol-induced action potentials (Figure 1E–F). Moreover, we showed that we could rescue the Ir7g<sup>1</sup>, Ir51b<sup>1</sup> and Ir56d<sup>1</sup> mutant phenotypes only when we expressed the cognate transgenes in bitter GRNs using the Gr33a-GAL4 (Figure 3G). Thus, while Ir7g/Ir51b are not exclusive to bitter GRNs, their functional role in cholesterol detection is bitter-GRN-specific.

      (3) The authors only investigated the responses of GRNs in the labellum, but GRN responses in the leg may also contribute to the avoidance of cholesterol feeding. Alternatively, leg GRNs might contribute to cholesterol attraction that is unmasked when bitter GRNs are silenced. In support of this possibility, Ahn et al. (2017) showed that Ir56d functions in sugar GRNs of the leg to promote appetitive responses to fatty acids.

      This is an interesting idea. Indeed, when bitter GRNs are hyperpolarized, the flies exhibit a strong attraction to cholesterol. Nevertheless, the cellular basis for cholesterol attraction and whether it is mediated by GRNs in the legs will require a future investigation.

      (4) The authors might consider using proboscis extension as an additional readout of taste attraction or aversion, which would help them more directly link the labellar GRN responses to a behavioral readout. Using food ingestion as a readout can conflate the contribution of taste with post-ingestive effects, and the regulation of food ingestion also may involve contributions from GRNs on multiple organs, whereas organ-specific contributions can be dissociated using proboscis extension. For example, does presenting cholesterol on the proboscis lead to aversive responses in the proboscis extension assay (e.g., suppression of responses to sugar)? Does this aversion switch to attraction when bitter GRNs are silenced, as with the feeding assay?

      We thank the reviewer for the suggestion regarding the use of the proboscis extension reflex (PER) assay to strengthen the link between labellar GRN activity and behavioral responses to cholesterol.

      Author response image 1.

      Our PER assay results shown above indicate that cholesterol presentation on the labellum or forelegs leads to an aversive response, as evidenced by a significant reduction in proboscis extension when compared to control stimuli (Author response image 1A. 2% sucrose or 2% sucrose with 10<sup>-1</sup>% cholesterol was applied to labellum or forelegs and the percent PER was recorded. n=6. Data were compared using single-factor ANOVA coupled with Scheffe’s post-hoc test. Statistical significance was compared with the control. Means ± SEMs. **p<0.01). This finding supports the idea that cholesterol is detected by labellar and leg GRNs and elicits behavioral avoidance. In contrast, sucrose stimulation robustly induces proboscis extension, as expected for an appetitive stimulus. We confirmed the defects of due to each Ir mutant by presenting the stimuli to the labellum (Author response image 1B). Together, these PER results provide a more direct behavioral correlate of labellar and leg GRN activation and reinforce our conclusion that cholesterol is sensed as an aversive tastant through the labellar bitter GRNs.

      (5) The authors claim that the cholesterol receptor is composed of IR25a, IR76b, IR56d, and either IR7g or IR51b. While the authors have shown that IR25a and IR76b are each required for cholesterol sensing, they did not show that both are required components of the same receptor complex. If the authors are relying on previous studies to make this assumption, they should state this more clearly. Otherwise, I think further misexpression experiments may be needed where only IR25a or IR76b, but not both, are expressed in GRNs.

      In our study, we relied on prior work demonstrating that Ir25a and Ir76b function as broadly required co-receptors in most IR-dependent chemosensory pathways (Ganguly et al., 2017; Lee et al., 2018). These studies showed that Ir25a and Ir76b are co-expressed in many GRNs across multiple taste modalities. Functional IR complexes often fail to form or signal properly in the absence of these co-receptors. Thus, it is widely accepted in the field that Ir25a and Ir76b function together as a core heteromeric scaffold for diverse IR complexes, akin to co-receptors in other ionotropic glutamate receptor families. We state that while Ir25a and Ir76b are presumed co-receptors in the cholesterol receptor complex based on their conserved roles, their direct physical interaction with Ir7g, Ir51b, and Ir56d remains to be demonstrated.

      In support of this model, we note that in our ectopic expression experiments using I9 sensilla, which endogenously express Ir25a and Ir76b, introduction of either Ir7g + Ir56d or Ir51b + Ir56d was sufficient to confer cholesterol sensitivity (Figure 4B). We obtained a similar result in L6 sensilla (Figure 4D), which also endogenously express Ir25a and Ir76b. These findings imply that both co-receptors are already present in these sensilla and are likely part of the functional complex. However, we agree that we have not directly tested the requirement for both co-receptors in a minimal reconstitution context, such as expressing only Ir25a or Ir76b alongside tuning IRs in an otherwise null background. Such an experiment would indeed provide more direct evidence of their joint requirement in the receptor complex. Future studies, including heterologous expression experiments, will be necessary to define the cholesterol-receptor complexes.

    1. eLife Assessment

      This valuable study advances understanding of how corticotrophin releasing factor in the bed nucleus of the stria terminalis regulates sustained and phasic fear and how this differs between sexes. The evidence is convincing and based on state-of-the-art techniques. The work will be of interest to neuroscientists studying the biological basis of fear processing.

    2. Reviewer #1 (Public review):

      The aim of this study is to test the overarching hypothesis that plasticity in BNST CRF neurons drives distinct behavioral responses to unpredictable threat in males and females. The manuscript provides solid evidence for a sex-specific role for CRF-expressing neurons in the BNST in unpredictable aversive conditioning and subsequent hypervigilance across sexes. As the authors note, this is an important question given the high prevalence of sex differences in stress-related disorders, like PTSD, and the role of hypervigilance and avoidance behaviors in these conditions. The study includes in vivo manipulation, bulk calcium imaging, and cellular resolution calcium imaging, which yield important insights into cell-type specific activity patterns. A major strength of this manuscript is the inclusion of both males and females and attention to possible behavioral and neurobiological differences between them throughout.

    3. Reviewer #2 (Public review):

      This study examined the role of CRF neurons in the BNST in both phasic and sustained fear in males and females. The authors first established a differential fear paradigm whereby shocks were consistently paired with tones (Full) or only paired with tones 50% of the time (Part), or controls who were exposed to only tones with no shocks. Recall tests established that both Full and Part conditioned male and female mice froze to the tones, with no difference between the paradigms. Additional studies using the NSF and startle test, established that neither fear paradigm produced behavioral changes in the NSF test, suggesting that these fear paradigms do not result in an increase in anxiety-like behavior. Part fear conditioning, but not Full, did enhance startle responses in males but not females, suggesting that this fear paradigm did produce sustained increases in hypervigilance in males exclusively. Photometry studies found that while undifferentiated BNST neurons all responded to shock itself, only Full conditioning in males lead to a progressive enhancement of the magnitude of this response. BNST neurons in males, but not females, were also responsive to tone onset in both fear paradigms, but only in Full fear did the magnitude of this response increase across training. Knockdown of CRF from the BNST had no effect on fear learning in males or females, nor any effect in males on fear recall in either paradigm, but in females enhanced both baseline and tone-induced freezing only in Part fear group. When looking at anxiety following fear training, it was found in males that CRF knockdown modulated anxiety in Part fear trained animals and amplified startle in Full trained males but had no effect in either test in females. Using 1P imaging, it was found that CRF neurons in the BNST generally decline in activity across both conditioning and recall trials, with some subtle sex differences emerging in the Part fear trained animals in that in females BNST CRF neurons were inhibited after both shock and omission trials but in males this only occurred after shock and not omission trials. In recall trials, CRF BNST neuron activity remained higher in Part conditioned mice relative to Full conditioned mice.

      Overall, this is a very detailed and complex study that incorporates both differing fear training paradigms and males and females, as well as a suite of both state-of-the-art imaging techniques and gene knockdown approaches to isolate the role and contributions of CRF neurons in the BNST to these behavioral phenomena. The strengths of this study come from the thorough approach that the authors have taken, which in turn helped to elucidate nuanced and sex specific roles of these neurons in the BNST to differing aspects of phasic and sustained fear. More so, the methods employed provide a strong degree of cellular resolution for CRF neurons in the BNST. In general, the conclusions appropriately follow the data, although the authors do tend to minimize some of the inconsistencies across studies, although this has now been addressed to some degree. The discussion has also been improved to now address some of the inconsistencies in the data head on. Discussion of a few other points is below:

      - Given the focus on CRF neurons in the BNST, it was unclear why the photometry studies were performed in undifferentiated BNST neurons as opposed to CRF neurons specifically, although the authors have now explained this in better depth making this clearer to the reader.

      - The CRF KD studies are interesting, but it remains speculative as to whether these effects are mediated locally in the BNST or due to CRF signaling at downstream targets. As the literature on local pharmacological manipulation of CRF signaling within the BNST seems to be largely performed in males, the addition of pharmacological studies here would benefit this to help to resolve if these changes are indeed mediated by local impairments in CRF release within the BNST or not. While it is not essential to add these experiments, the authors have addressed this point in the discussion and highlighted studies like this as necessary in future work.

      - The authors have addressed the difference between arousal and anxiety by expanding the discussion to include more focus on the behavioral measures. The CRF KD data are still somewhat confusing but better contextualized now. Overall, the manuscript has been improved by the revisions and edits the authors have made.

    4. Reviewer #3 (Public review):

      Hon et al. investigated the role of BNST CRF signaling in modulating phasic and sustained fear in male and female mice. They found that partial and full fear conditioning had similar effects in both sexes during conditioning and during recall. However, males in the partially reinforced fear conditioning group showed enhanced acoustic startle, compared to the fully reinforced fear conditioning group, an effect not seen in females. Using fiber photometry to record calcium activity in all BNST neurons, the authors show that the BNST was responsive to foot shock in both sexes and both conditioning groups. Shock response increased over the session in males in the fully conditioned fear group, an effect not observed in the partially conditioned fear group. This effect was not observed in females. Additionally, tone onset resulted in increased BNST activity in both male groups, with the tone response increasing over time in the fully conditioned fear group. This effect was less pronounced in females, with partially conditioned females exhibiting a larger BNST response. During recall in males, BNST activity was suppressed below baseline during tone presentations and was significantly greater in the partially conditioned fear group. Both female groups showed an enhanced BNST response to the tone that slowly decayed over time. Next, they knocked CRF in the BNST to examine its effect on fear conditioning, recall and anxiety-like behavior after fear. They found no effect of the knockdown in either sex or group during fear conditioning. During fear recall, BNST CRF knockdown lead to an increase in freezing in only the partially conditioned females. In the anxiety-like behavior tasks, BNST CRF knockdown lead to increased anxiolysis in the partially reinforced fear male, but not in females. Surprisingly, BNST CRF knockdown increased startle response in fully conditioned, but not partially conditioned males. An effect not observed in either female group. In a final set of experiments, the authors single photon calcium imaging to record BNST CRF cell activity during fear conditioning and recall. Approximately, 1/3 of BNST CRF cells were excited by shock in both sexes, with the rest inhibited and no differences were observed between sexes or group during fear conditioning. During recall, BNST CRF activity decreased in both sexes, an effect pronounced in male and female fully conditioned fear groups.

      Overall, these data provide novel, intriguing evidence in how BNST CRF neurons may encode phasic and sustained fear differentially in males and females. The experiments were rigorous. My biggest concerns I have regard the interpretations and some conclusions from this data set, which I have stated below.

      (1) It was surprising to see minimal and somewhat conflicting behavioral effects due to BNST CRF knockdown. The authors provide a representative image and address this in the conclusion. They mention the role of local vs projection CRF circuits as well as the role of GABA. I don't think those experiments are necessary for this manuscript. However, it may be worthwhile to see through in situ hybridization or IHC, to see BNST CRF levels after both full and partial conditioned fear paradigms. Additionally, it would help to see a quantification of the knockdown of the animals. The authors can add a figure showing deltaF/F changes from control.

      (2) Related to the previous point, it was surprising to see an effect of the CRF deletion in the full fear group compared to the partial fear in the acoustic startle task. To strengthen the conclusion about differential recruitment of CRF during phasic and sustained fear, the experiment in my previous point could help elucidate that. Conversely, intra-BNST administration of a CRF antagonist into the BNST before the acoustic startle after both conditioning tasks could also help. Or patch from BNST CRF neurons after the conditioning tasks to measure intrinsic excitability. Not all these experiments are needed to support the conclusion, it's some examples.

      (3) In Figure 5 F and K, the authors report data combined for both part and full fear conditioning. Were there any differences between the number of excited or inhibited neurons b/t the conditioning groups? Also, can the authors separate male and female traces in Fig 5 E and P?

      (4) Also, regarding the calcium imaging data, what was the average length of a transient induced by shock? Were there any differences between the sexes?

    5. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public Review):

      The aim of this study is to test the overarching hypothesis that plasticity in BNST CRF neurons drives distinct behavioral responses to unpredictable threat in males and females. The manuscript provides evidence for a possible sex-specific role for CRF-expressing neurons in the BNST in unpredictable aversive conditioning and subsequent hypervigilance across sexes. As the authors note, this is an important question given the high prevalence of sex differences in stress-related disorders, like PTSD, and the role of hypervigilance and avoidance behaviors in these conditions. The study includes in vivo manipulation, bulk calcium imaging, and cellular resolution calcium imaging, which yield important insights into cell-type specific activity patterns. However, it is difficult to generate an overall conclusion from this manuscript, given that many of the results are inconsistent across sexes and across tests and there is an overall lack of converging evidence. For example, partial conditioning yields increased startle in males but not females, yet, CRF KO only increases startle response in males after full conditioning, not partial, and CRF neurons show similar activity patterns between partial and full conditioning across sexes. Further, while the study includes a KO of CRF, it does not directly address the stated aim of assessing whether plasticity in CRF neurons drives the subsequent behavioral effects unpredictable threat.

      We appreciate the reviewer’s summary and agree that there is a large amount of complexity to the results, and that it was difficult to generate a simple model/conclusion to summarize our work. This is the unfortunate side effect of looking across both sexes at different conditioning paradigms, however, we believe that it is important to convey this information to the field even without a simple answer.  Our data reinforces the very important findings from the Maren and Holmes groups that partial fear is a different process than full fear, and that the BNST plays a differential role here. We have reworded the manuscript to better convey this complexity.

      A major strength of this manuscript is the inclusion of both males and females and attention to possible behavioral and neurobiological differences between them throughout. However, to properly assess sex-differences, sex should be included as a factor in ANOVA (e.g. for freezing, startle, and feeding data in Figure 1) to assess whether there is a significant main effect or interaction with sex. If sex is not a statistically significant factor, both sexes should be combined for subsequent analyses. See, Garcia-Sifuentes and Maney, eLife 2021 https://elifesciences.org/articles/70817. There are additional cases where t-tests are used to compare groups when repeated measures ANOVAs would be more appropriate and rigorous.

      We agree with the reviewer that this is the more appropriate analysis and have changed the analysis and figures throughout the revised manuscript to better assess sex differences as well as differences between fear conditions.

      Additionally, it's unclear whether the two sexes are equally responsive to the shock during conditioning and if this is underlying some of the differences in behavioral and neuronal effects observed. There are some reports that suggest shock sensitivity differs across sexes in rodents, and thus, using a standard shock intensity for both males and females may be confounding effects in this study.

      This is a great point. We have conducted appropriate analysis (Sex by Tone Repeated measures two-way ANOVAS for each of the groups: Ctrl, Full, Part) and there are no sex differences in freezing between males and females. The extent of conditioning is not different between the groups suggesting that if there was a difference in shock sensitivity, it is not driving any discernible differences in behavioral performance. However, it is possible that the experience of the shock differs for the animals even in the absence of any measurable behavior.

      The data does not rule out that BNST CRF activity is not purely tracking the mobility state of the animal, given that the differences in activity also track with differences in freezing behavior. The data shows an inverse relationship between activity and freezing. This may explain a paradox in the data which is why males show a greater suppression of BNST activity after partial conditioning than full conditioning, if that activity is suspected to drive the increased anxiety-like response. Perhaps it reflects that activity is significantly suppressed at the end of the conditioning session because animals are likely to be continuously freezing after repeated shock presentations in that context. It would also explain why there is less of a suppression in activity over the course of the recall session, because there is less freezing as well during recall compared with conditioning.

      While it is possible that the BNST may be tracking activity, we believe it is not purely tracking mobility state. For instance, while freezing increases across tone exposures in Part fear regardless of sex, males show an increase while females show a reduction in BNST response during tone 5 (Fig 2K). The data the reviewer refers to showing the inverse relationship with BNST activity and freezing would have suggested the opposite response if it were purely tracking the mobility state of the animal. This is also the case with BNST<sup>CRF</sup> activity to first and last tone during recall. Despite the suppression of activity over the course of recall (Fig 5K), we see an increase in BNST<sup>CRF</sup> tone response when comparing tone 1 and 6 in males and a decrease in females (Fig 6M), again suggesting the BNST is responding to more than just activity.

      A mechanistic hypothesis linking BNST CRF neurons, the behavioral effects observed after fear conditioning, and manipulation of CRF itself are not clearly addressed here.

      We disagree with this assertion. The data suggests a model in which males respond with increased arousal and Part fear males show persistent activation of the BNST and BNST<sup>CRF</sup> neurons during fear conditioning and recall while female Part fear mice show the opposite response. This female response differs from what the field believes to be the role of the BNST in sustained fear. Additionally, we show that CRF knockdown is not involved in fear differentiation or fear expression in males, while it enhances fear learning and recall in females. We have reworded the manuscript to highlight these novel findings.

      Reviewer #2 (Public Review):

      This study examined the role of CRF neurons in the BNST in both phasic and sustained fear in males and females. The authors first established a differential fear paradigm whereby shocks were consistently paired with tones (Full) or only paired with tones 50% of the time (Part), or controls who were exposed to only tones with no shocks. Recall tests established that both Full and Part conditioned male and female mice froze to the tones, with no difference between the paradigms. Additional studies using the NSF and startle test, established that neither fear paradigm produced behavioral changes in the NSF test, suggesting that these fear paradigms do not result in an increase in anxiety-like behavior. Part fear conditioning, but not Full, did enhance startle responses in males but not females, suggesting that this fear paradigm did produce sustained increases in hypervigilance in males exclusively.

      Thank you for this clear summary of the behavioral work.

      Photometry studies found that while undifferentiated BNST neurons all responded to shock itself, only Full conditioning in males lead to a progressive enhancement of the magnitude of this response. BNST neurons in males, but not females, were also responsive to tone onset in both fear paradigms, but only in Full fear did the magnitude of this response increase across training. Knockdown of CRF from the BNST had no effect on fear learning in males or females, nor any effect in males on fear recall in either paradigm, but in females enhanced both baseline and tone-induced freezing only in Part fear group. When looking at anxiety following fear training, it was found in males that CRF knockdown modulated anxiety in Part fear trained animals and amplified startle in Fully trained males but had no effect in either test in females. Using 1P imaging, it was found that CRF neurons in the BNST generally decline in activity across both conditioning and recall trials, with some subtle sex differences emerging in the Part fear trained animals in that in females BNST CRF neurons were inhibited after both shock and omission trials but in males this only occurred after shock and not omission trials. In recall trials, CRF BNST neuron activity remained higher in Part conditioned mice relative to Full conditioned mice.

      Overall, this is a very detailed and complex study that incorporates both differing fear training paradigms and males and females, as well as a suite of both state of the art imaging techniques and gene knockdown approaches to isolate the role and contributions of CRF neurons in the BNST to these behavioral phenomena. The strengths of this study come from the thorough approach that the authors have taken, which in turn helped to elucidate nuanced and sex specific roles of these neurons in the BNST to differing aspects of phasic and sustained fear. More so, the methods employed provide a strong degree of cellular resolution for CRF neurons in the BNST. In general, the conclusions appropriately follow the data, although the authors do tend to minimize some of the inconsistencies across studies (discussed in more depth below), which could be better addressed through discussion of these in greater depth. As such, the primary weakness of this manuscript comes largely from the discussion and interpretation of mixed findings without a level of detail and nuance that reflects the complexity, and somewhat inconsistency, across the studies. These points are detailed below:

      - Given the focus on CRF neurons in the BNST, it is unclear why the photometry studies were performed in undifferentiated BNST neurons as opposed to CRF neurons specifically (although this is addressed, to some degree, subsequently with the 1P studies in CRF neurons directly). This does limit the continuity of the data from the photometry studies to the subsequent knockdown and 1P imaging studies. The authors should address the rationale for this approach so it is clear why they have moved from broader to more refined approaches.

      The reviewer raises a good point.  We did some preliminary photometry studies with BNST CRF neurons and found that there was poor time locked signal. We reasoned that this was due to the heterogeneity of the cell activity, as we saw in our previous publication (Yu et al). Because of this, we moved to the 1p imaging work in place of continued BNST CRF photometry. We have also reworded the manuscript to better discuss the complexities and inconsistencies in findings across the studies.

      - The CRF KD studies are interesting, but it remains speculative as to whether these effects are mediated locally in the BNST or due to CRF signaling at downstream targets. As the literature on local pharmacological manipulation of CRF signaling within the BNST seems to be largely performed in males, the addition of pharmacological studies here would benefit this to help to resolve if these changes are indeed mediated by local impairments in CRF release within the BNST or not. While it is not essential to add these experiments, the manuscript would benefit from a more clear description of what pharmacological studies could be performed to resolve this issue.

      We agree with the reviewer that the addition of this experiment would be highly informative for differentiating the role of CRF in the BNST. This is something that will need to be considered moving forward and we have added this as a point of discussion.

      - While I can appreciate the authors perspective, I think it is more appropriate to state that startle correlates with anxiety as opposed to outright stating that startle IS anxiety. Anxiety by definition is a behavioral cluster involving many outputs, of which avoidance behavior is key. Startle, like autonomic activation, correlates with anxiety but is not the same thing as a behavioral state of anxiety (particularly when the startle response dissociates from behavior in the NSF test, which more directly tests avoidance and apprehension). Throughout the manuscript the use of anxiety or vigilance to describe startle becomes interchangeable, but then the authors also dissociate these two, such as in the first paragraph of the discussion when stating that the Part fear paradigm produces hypervigilance in males without influencing fear or anxiety-like behaviors. The manuscript would benefit from harmonization of the language used to operationally define these behaviors and my recommendation would be to remain consistent with the description that startle represents hypervigilance and not anxiety, per se.

      The reviewer raises an excellent point, we have clarified in the revised manuscript.

      - The interpretation of the anxiety data following CRF KD is somewhat confusing. First, while the authors found no effect of fear training on behavior in the NSF test in the initial studies, now they do, however somewhat contradictory to what one would expect they found that Full fear trained males had reduced latency to feed (indicative of an anxiolytic response), which was unaltered by CRF KD, but in Part fear (which appeared to have no effect on its own in the NSF test), KD of CRF in these animals produced an anxiolytic effect. Given that the Part fear group was no different from control here it is difficult to interpret these data as now CRF KD does reduce latency to feed in this group, suggesting that removal of CRF now somehow conveys an anxiolytic response for Part fear animals. In the discussion the authors refer to this outcome as CRF KD "normalizing" the behavior in the NSF test of Part fear conditioned animals as now it parallels what is seen after Full fear, but given that the Part fear animals with GFP were no different then controls (and neither of these fear training paradigms produced any effect in the NSF test in the first arm of studies), it seems inappropriate to refer to this as "normalization" as it is unclear how this is now normalized. Given the complexity of these behavioral data, some greater depth in the discussion is required to put these data in context and describe the nuance of these outcomes, in particular a discussion of possible experimental factors between the initial behavioral studies and those in the CRF KD arm that could explain the discrepancy in the NSF test would be good (such as the inclusion of surgery, or other factors that may have differed between these experiments). These behavioral outcomes are even more complex given that the opposite effect was found in startle whereby CRF KD amplified startle in Full trained animals. As such, this portion of the discussion requires some reworking to more adequately address the complexity of these behavioral findings.

      The reviewer raises a good point, and we agree that there are many inconsistencies in the behaviors. We believe it is still good to show these results but have expanded the manuscript on potential reasons for these behavioral inconsistencies.

      Reviewer #3 (Public Review):

      Hon et al. investigated the role of BNST CRF signaling in modulating phasic and sustained fear in male and female mice. They found that partial and full fear conditioning had similar effects in both sexes during conditioning and during recall. However, males in the partially reinforced fear conditioning group showed enhanced acoustic startle, compared to the fully reinforced fear conditioning group, an effect not seen in females. Using fiber photometry to record calcium activity in all BNST neurons, the authors show that the BNST was responsive to foot shock in both sexes and both conditioning groups. Shock response increased over the session in males in the fully conditioned fear group, an effect not observed in the partially conditioned fear group. This effect was not observed in females. Additionally, tone onset resulted in increased BNST activity in both male groups, with the tone response increasing over time in the fully conditioned fear group. This effect was less pronounced in females, with partially conditioned females exhibiting a larger BNST response. During recall in males, BNST activity was suppressed below baseline during tone presentations and was significantly greater in the partially conditioned fear group. Both female groups showed an enhanced BNST response to the tone that slowly decayed over time. Next, they knocked CRF in the BNST to examine its effect on fear conditioning, recall and anxiety-like behavior after fear. They found no effect of the knockdown in either sex or group during fear conditioning. During fear recall, BNST CRF knockdown lead to an increase in freezing in only the partially conditioned females. In the anxiety-like behavior tasks, BNST CRF knockdown lead to increased anxiolysis in the partially reinforced fear male, but not in females. Surprisingly, BNST CRF knockdown increased startle response in fully conditioned, but not partially conditioned males. An effect not observed in either female group. In a final set of experiments, the authors single photon calcium imaging to record BNST CRF cell activity during fear conditioning and recall. Approximately, 1/3 of BNST CRF cells were excited by shock in both sexes, with the rest inhibited and no differences were observed between sexes or group during fear conditioning. During recall, BNST CRF activity decreased in both sexes, an effect pronounced in male and female fully conditioned fear groups.

      Overall, these data provide novel, intriguing evidence in how BNST CRF neurons may encode phasic and sustained fear differentially in males and females. The experiments were rigorous.

      We thank you for this positive review of our manuscript.

      Recommendations for the authors:

      Reviewer #1 (Recommendations For The Authors):

      There are several graphs representing different analyses of (presumably) the same group of subjects, but which have different N/group. For example, in Figure 2:

      (1) Fig 2P seems to have n=10 in Part Male group (Peak), but 2Q only has n=9 in Part Male group (AUC)

      (2) Fig 2S seems to have n=10 in Part Female group (Peak), but 2T only has n=7 in Part Female group (AUC)

      (3) Fig 2G (Tone Resp) has n=6 Full Males but 2F (Tone Resp), 2H (Shock Resp), and 2I (Shock Resp) have n=7 Full Males

      (4) Fig 2K (Tone Resp) has n=7 Full Females but 2L (Tone Resp), 2M (Shock Resp), and 2N (Shock Resp) have n=8 Full Females

      (5) Fig 2L (Tone Resp) has n=9 Part Females but 2K (Tone Resp), 2M (Shock Resp), and 2N (Shock Resp) have n=10 Part Females

      It's possible that this is just due to overlapping individual data points which are made harder to see due to the low resolution of the figures. If so, this can be easily rectified. However, there may also be subjects missing from some analyses which must be clarified or corrected.

      We thank you for catching these. We have gone through and fixed any issues with data points and have added statistics and exclusions in datasets to figure legends to further explain inconsistencies.

      Regarding statistical tests:

      (2) Data in Figs 2G and 2I should be analyzed using a two-way RM ANOVA.

      We have now included sex as a factor in most of our analysis and are now using appropriate statistical tests.

      (3) Data in Fig 3K should be analyzed using a two-way RM ANOVA.

      We are now using appropriate statistical tests.

      Calcium activity in response to the shock during conditioning and in response to the tone during recall should be included in Figure 5. Given partial and full animals also receive unequal presentations of the cue, it would be useful to see the effects trial by trial or normalized to the first 3 presentations only.

      The reviewer raises a great point. We have changed this figure and have now added the response to shock and tones. Since we are most interested in the difference between sustained and phasic fear, we decided to compare tone 3 in Full fear and tone 4 in Part fear, which differ in the ambiguity of their cue and only have one tone difference.

      Histology maps should be included for all experiments depicting viral spread and implant location for all animals, in addition to the included representative histology images. These can be placed in the supplement.

      We agree this is helpful. While we have confirmed all of the experiments are hits, the tissue is no longer in condition for this analysis.

      Referring to the quantification of peaks in fiber photometry and cellular resolution calcium imaging data as "spikes" is a bit misleading given the inexact relationship between GCAMP sensor dynamics/calcium binding and neuronal action potentials, perhaps calling it "event" frequency would be more clear.

      We have changed the references of spikes to events as suggested.

      The legend for Figure 2S is mislabeled as A.

      Thank you for catching this mistake, it has been fixed.

      The methods refer to CRFR1 fl/fl animals but it seems no experiments used these animals, only CRF fl/fl.

      We have fixed this, thank you.

      Reviewer #2 (Recommendations For The Authors):

      As stated in the public review, while I think the addition of local pharmacological studies blocking CRF1 and 2 receptors in the BNST in both males and females, done under the same conditions as all of the other testing herein, would help to resolve some of the speculation of interpreting the CRF KD data, I dont think these studies are essential to do, but it would be good for the authors to more explicitly state what studies could be done and how they could facilitate interpretation of these data.

      Thank you for this suggestion. We have added this discussion into the manuscript.

      Asides from this, my other recommendations for the authors are to more clearly address the discrepancies in behavioral outcomes across studies and explicitly describe their rationale for the sequence of experiments performed and to harmonize their operationalization of how they define anxiety.

      Again, we appreciate these great suggestions. We have added more discussion on the behavioral discrepancies as well as rationale for the experiments. We have also changed the wording to remain consistent that the NSF test relates to anxiety and the Startle test relates to vigilance.

      - In Figure 2, Panel S is listed as Panel A in the caption and should be corrected.

      Thank you for catching this mistake, we have fixed it.

      Reviewer #3 (Recommendations For The Authors):

      My biggest concerns I have regard the interpretations and some conclusions from this data set, which I have stated below.

      (1) It was surprising to see minimal and somewhat conflicting behavioral effects due to BNST CRF knockdown. The authors provide a representative image and address this in the conclusion. They mention the role of local vs projection CRF circuits as well as the role of GABA. I don't think those experiments are necessary for this manuscript. However, it may be worthwhile to see through in situ hybridization or IHC, to see BNST CRF levels after both full and partial conditioned fear paradigms. Additionally, it would help to see a quantification of the knockdown of the animals.

      Thank you for these great suggestions. We will consider these for future experiments. We piloted out some CRF sensor experiments to probe this, but it was unclear if the signal to noise for the sensor was sufficient. We hope to do more of this in the future if we ever manage to get funding for this work.

      The authors can add a figure showing deltaF/F changes from control.

      We did not have control mice in these in-vivo experiments Our main interests lie in understanding the differences in Full and Part Fear conditioning paradigms specifically.

      (2) Related to the previous point, it was surprising to see an effect of the CRF deletion in the full fear group compared to the partial fear in the acoustic startle task. To strengthen the conclusion about differential recruitment of CRF during phasic and sustained fear, the experiment in my previous point could help elucidate that. Conversely, intra-BNST administration of a CRF antagonist into the BNST before the acoustic startle after both conditioning tasks could also help. Or patch from BNST CRF neurons after the conditioning tasks to measure intrinsic excitability. Not all these experiments are needed to support the conclusion, it's some examples.

      We thank the reviewer for these suggestions and agree that these are important experiments. We will consider this in future experiments exploring the role of BNST CRF in fear conditioning.

      (3) In Figure 5 F and K, the authors report data combined for both part and full fear conditioning. Were there any differences between the number of excited or inhibited neurons b/t the conditioning groups?

      We are only looking at the first shock exposure in these figures. These were combined because the first tone and shock exposure is identical in Full and Part fear conditioning. Differences in these behavioral paradigms emerge after Tone 3 exposure, where Part fear does not receive a shock while Full fear does.

      Also, can the authors separate male and female traces in Fig 5 E and P?

      Traces in Fig E are from females only. We did not include male traces because males and females had identical responses to first shock, and we felt only one trace was needed as an example. Traces in Figure P are from males. We did not show female traces because females did not show differential effects from baseline to end.

      (4) Also, regarding the calcium imaging data, what was the average length of a transient induced by shock? Were there any differences between the sexes?

      We have many cells in each condition, and the length of traces after shock were all different and hard to quantify, as for example, sometimes cells were active before shock and thus trace length would be difficult to quantify. Therefore, to keep consistency and reduce ambiguity regarding trace lengths, we focused on keeping the time consistent across mice and focused on the 10 second window post shock to be consistent across conditions.

    1. eLife Assessment

      This valuable study used functional MRI experiments to identify the involvement of a left parietal area (PF) in reasoning about the physical properties of actions, objects, and events. Solid evidence was shown regarding the commonalities and differences across different types of reasoning tasks, yet the methodological and theoretical interpretations require further scrutiny. The study would be of interest to researchers studying the cognitive and neural mechanisms of reasoning and problem solving.

    2. Reviewer #1 (Public review):

      In this study, Osiurak and colleagues investigate the neurocognitive basis of technical reasoning. They use multiple tasks from two neuroimaging studies to show that the area PF is central to technical reasoning and plays an essential role in tool-use and non-tool-use physical problem-solving, as well as both conditions of mentalizing tasks. They also demonstrate the specificity of technical reasoning, finding that area PF is not involved in the fluid-cognition task or the mentalizing network (INT+PHYS vs. PHYS-only). This work enhances our understanding of the neurocognitive basis of technical reasoning that supports advanced technologies.

      Strengths:

      - The topic this study focuses on is intriguing and can help us understand the neurocognitive processes involved in technical reasoning and advanced technologies.<br /> - The researchers collected fMRI data from multiple tasks. The data is rich and encompasses the mechanical problem-solving task, psychotechnical task, fluid-cognition task, and mentalizing tasks.<br /> - The article is well written.

      The authors have addressed many of the reviewers' concerns in their response. They utilized both correlation analysis and coordinate analysis to tackle alternative hypotheses, namely the same-region-but-different-function interpretation and the adjacency interpretation. Additionally, ROI analysis was conducted to validate the negative results. These additional analyses have enhanced the reliability of the findings. This study offers valuable insights into the neurocognitive mechanisms underlying technical reasoning.

      Weaknesses:

      While the authors attempted to address the limitations of overlap analysis by correlating activation across different tasks within subjects, this issue could not be entirely resolved due to the constraints of the current experimental design. The mechanical problem-solving task was not included since the sample of subjects differed from that of other tasks. Furthermore, the fluid-cognition task was not scanned in the same run as the psychotechnical and mentalizing tasks, which may have contributed to a lack of correlation between them, thereby affecting result interpretation. Moreover, the core cognitive focus of this study, technical reasoning, may be influenced by assumptions about motion-related information. While this issue has been discussed in the discussion section, further evidence is needed to substantiate this interpretation.

    3. Reviewer #2 (Public review):

      Strengths:

      The authors have done a nice job providing additional data in response to reviewer feedback. I appreciate that accuracy plots are now included, as well as a separate analysis where differences in parameter estimates are performed for participants whose accuracy data were above chance levels. I also appreciate the new figure with the sphere ROIs for each participant, as they help us appreciate anatomical variability in the peak response separately for each task.

      I have four concerns related to the weaknesses of the study:

      (1) Although the results still hold when removing participants whose accuracy was 50% or less, a major limitation of this study is that participants made a button press response only to the last trial in a block. This is problematic because a participant could get all trials in a block correct except for the last one, or a participant could get all trials in a block wrong, and performance would be considered equivalent-as a consequence, it is not possible for one to know if participants who are at chance are performing differently from participants who are not at chance, and it is not possible to control for variance in reaction time (a concern also raised by reviewer 3).

      (2) My second concern relates to the way in which the data are interpreted based on thresholding. There is above-threshold activation in the left SMG for all tasks except the fluid cognition task. The z-scores associated with significant voxels in Figure 3 are very strong (minimum z is 6). If one were to relax the threshold of the group level maps to, e.g., p < .001, uncorrected, FDR q < .05, or FWER of .10, there will be overlapping voxels outside the SMG. The discussion of the left SMG in the manuscript is prominent and narrowly construed-the left SMG is discussed as if it were 'the' region: "This confirms that the technical-reasoning network depends upon the recruitment of the left area PF, even if additional cognitive processes involving other peripheral brain areas can be engaged depending on the task" (pp. 9). My intuition is there will be numerous other areas of overlap when using a threshold that is still highly significant (e.g., z = 3 or 4). So, for proponents of the technical reasoning hypothesis, is there a counterfactual or alternative brain area/network/system not in the left SMG?

      (3) I like the new Figure 6 because it shows variability in the location of the peak coordinate at the level of single participants. And, indeed, there's considerable variability that is typical when localizing ROIs in single participants. My concern is the level at which hypothesis testing is performed. An independent SMG ROI is used to extract parameter estimates and correlate responses between tasks to show a pattern of correlation that comports with a technical reasoning model of left SMG function. This is a fine approach but it does not rule out the so-called 'same region different function' interpretation because it relies on correlation-one cannot reverse infer that the left SMG is carrying out the same function across different tasks because the response in that area is more strongly correlated between certain tasks. This finding points to that possibility and makes interesting predictions for future studies to pursue, but it cannot tell us whether common functions in the left SMG are involved in each task. E.g., one interesting prediction for future studies is to test if patients with lesions to this site are disproportionately more inaccurate in the experimental condition of the mechanical problem solving task, the psychotechnical task, the mentalizing task, but not the fluid cognition task.

      (4) I appreciated the approach to testing the adjacency interpretation by showing the sphere and peak Y coordinate across the tasks. It is interesting that across the groups, there is no difference in the peak Y coordinate of the psychotechnical task and both conditions of the mentalizing task, whereas the peak Y coordinate in the fluid intelligence task is more anterior in the post-central gyrus across participants (why is that?). But why restrict the analysis to just the Y coordinate? A rigorous way to test the adjacency hypothesis is to compute Euclidean distance among X, Y, and Z coordinates between any two tasks collected in the same participant. One can then test if the Euclidean distance between, e.g., the psychotechnical task and one condition of the mentalizing task is smaller than the Euclidean distance between the psychotechnical task and the fluid cognition task. Similarly, one can test whether Euclidean distance between the INT and PHY conditions of the mentalizing task is smaller than the Euclidean distance between the INT and psychotechnical task or PHY and psychotechnical task. There is no justification to restrict this analysis to the anterior-posterior dimension only.

    4. Reviewer #3 (Public review):

      The authors have responded very thoughtfully to many of the points raised, and the revised manuscript will make a useful contribution to our understanding of some of the computations performed by area PF. In particular, the investigators' addition of analyses of peak activations, their additional clarifications that area PF is likely to be part of a larger network concerned with technical reasoning, and their responses to the reviewers' concerns about differential task difficulty have strengthened the conclusions that can be drawn from the study.

      The authors' response does not completely mitigate the concern noted by all 3 reviewers that the control tasks were easier than their corresponding experimental tasks (for everything but the fluid cognition task). The specific trouble with this issue can be appreciated by looking at Figure 4A, for example, which shows that area PF was activated for many individuals in both the control task and the experimental mechanical problem-solving task, but more so for the latter. Since the experimental task was harder (and more trial time was likely spent on task trying to solve it), the concern remains that area PF was driven harder by the experimental task in part due to the more challenging nature of that task.

      The revised manuscript counters that the fluid cognition task was also harder than its control condition, yet did not activate PF more than its control condition. But this response seems to sidestep the central point of the reviewers' concerns: the fundamental computations that underlie the technical reasoning tasks may also be present in the respective (non technical-reasoning-based) control tasks and drive area PF activations to greater or lesser degrees based on how much they tax those computations. The fact that the fluid cognition experimental task and control task are not differentially difficult does not mitigate this concern, it just suggests that neither of those tasks tap the same fundamental computations, whatever they may be. (As an added note, Figures 2 and 4 show that both the PHYS-only and INT+PHYS mentalizing tasks only weakly activated PF, and both of these tasks were easier than the other technical cognition tasks).

      The new ROI analysis with removal of subjects who performed below 50% in the revised manuscript is somewhat helpful, but there are two remaining issues: 1) chance performance is defined by a binomial test in this case, so scores somewhat above 50% may still be at chance depending on the number of items, and thus there may have been subjects who were not removed who could not perform the tasks; 2) it would have been convincing to include accuracy as a covariate in the modeling of BOLD parameter estimates for the remaining above-chance subjects to ensure that all reported effects remain once differential task difficulty is taken into account. It also appears that the legend for Figure S2, which indicates that the figure includes just subjects who performed at or below 50%, may not be correct; does the figure instead show data from subjects who performed at or above 50%?

      Despite these remaining concerns, there are many aspects of this revised study that render it a useful contribution that will likely spur further research in this very interesting area.

    5. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1 (Public Review):

      Summary:

      In this study, Osiurak and colleagues investigate the neurocognitive basis of technical reasoning. They use multiple tasks from two neuroimaging studies and overlap analysis to show that the area PF is central for reasoning, and plays an essential role in tool-use and non-tool-use physical problem-solving, as well as both conditions of mentalizing task. They also demonstrate the specificity of the technical reasoning and find that the area PF is not involved in the fluid-cognition task or the mentalizing network (INT+PHYS vs. PHYS-only). This work suggests an understanding of the neurocognitive basis of technical reasoning that supports advanced technologies.

      Strengths:

      -The topic this study focuses on is intriguing and can help us understand the neurocognitive processes involved in technical reasoning and advanced technologies.

      -The researchers obtained fMRI data from multiple tasks. The data is rich and encompasses the mechanical problem-solving task, psychotechnical task, fluid-cognition task, and mentalizing task.

      -The article is well written.

      We sincerely thank Reviewer 1 for their positive and very helpful comments, which helped us improve the MS. Thank you.

      Weaknesses:

      - Limitations of the overlap analysis method: there are multiple reasons why two tasks might activate the same brain regions. For instance, the two tasks might share cognitive mechanisms, the activated regions of the two tasks might be adjacent but not overlapping at finer resolutions, or the tasks might recruit the same regions for different cognition functions.

      Thus, although overlap analysis can provide valuable information, it also has limitations.

      Further analyses that capture the common cognitive components of activation across different

      tasks are warranted, such as correlating the activation across different tasks within subjects for a region of interest (i.e. the PF).

      We thank Reviewer 1 for this comment. We added new analyses to address the two alternative interpretations stressed here by Reviewer 1, namely, the same-region-but-differentfonction interpretation and the adjacency interpretation. The new analyses ruled out both alternative interpretations, thereby reinforcing our interpretation.

      “The conjunction analysis reported was subject to at least two key limitations that needed to be overcome to assure a correct interpretation of our findings. The first was that the tasks could recruit the same regions for different cognition functions (same-region-but-different-function interpretation). The second was that the activated regions of the different tasks could be adjacent but did not overlap at finer resolutions (adjacency interpretation). We tested the same-region-but-different-function interpretation by conducting additional ROI analyses, which consisted of correlating the specific activation of the left area PF (i.e., difference in terms of mean Blood-Oxygen Level Dependent [BOLD] parameter estimates between the experimental condition minus the control condition) in the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. This analysis did not include the mechanical problem-solving task because the sample of participants was not the same for this task. As shown in Fig. 5, we found significant correlations between all the tasks that were hypothesized as recruiting technical reasoning, i.e., the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .05). By contrast, no significant correlation was obtained between these three tasks and the fluid-cognition task (all p > .15). This finding invalidates the same-region-but-different-function interpretation by revealing a coherent pattern in the activation of the left area PF in situations in which participants were supposed to reason technically. We examined the adjacency interpretation by analysing the specific locations of individual peak activations within the left area PF ROI for the mechanical problemsolving task, the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. These peaks, which corresponded to the maximum value of activation obtained for each participant within the left area PF ROI, are reported in Fig. 6. As can be seen, the peaks of the fluid-cognition task were located more anteriorly, in the left area PFt (Parietal Ft) and the postcentral cortex, compared to the peaks of the other four tasks, which were more posterior, in the left area PF. Statistical analyses based on the y coordinates of the individual activation peaks confirmed this description (Fig. 6). Indeed, the y coordinates of the peaks of the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task were posterior to the y coordinates of the peaks of the fluid-cognition task (all p < .05), whereas no significant differences were reported between the four tasks (all p > .05). These findings speak against the adjacency interpretation by revealing that participants recruited the same part of the left area PF to perform tasks involving technical reasoning.” (p. 11-13)

      Control tasks may be inadequate: the tasks may involve other factors, such as motor/ actionrelated information. For the psychotechnical task, fluid-cognition task, and mentalizing task, the experiment tasks need not only care about technical-cognition information but also motor-related information, whereas the control tasks do not need to consider motor-related information (mainly visual shape information). Additionally, there may be no difference in motor-related information between the conditions of the fluid-cognition task. Therefore, the regions of interest may be sensitive to motor-related information, affecting the research conclusion.

      We thank Reviewer 1 for this comment. We added a specific section in the discussion that addresses this limitation.

      “The second limitation concerns the alternative interpretation that the left area PF is not central to technical reasoning but to the storage of sensorimotor programs about the prototypical manipulation of common tools. Here we show that the left area PF is recruited even in situations in which participants do not have to process common manipulable tools. For instance, some items of the psychotechnical task consisted of pictures of tractor, boat, pulley, or cannon. The fact that we found a common activation of the left area PF in such tasks as well as in the mechanical problem-solving task, in which participants could nevertheless simulate the motor actions of manipulating novel tools, indicates that this brain area is not central to tool manipulation but to physical understanding. That being said, some may suggest that viewing a boat or a cannon is enough to incite the simulation of motor actions, so our tasks were not equipped to distinguish between the manipulation-based approach and the reasoning-based approach. We have already shown that the left area PF is more involved in tasks that focus on the mechanical dimension of the tool-use action (e.g., the mechanical interaction between a tool and an object) than its motor dimension (i.e., the interaction between the tool and the effector [e.g., 24, 40]). Nevertheless, we recognize that future research is still needed to test the predictions derived from these two approaches.” (p. 18-19)

      -Negative results require further validation: the cognitive results for the fluid-cognition task in the study may need more refinement. For instance, when performing ROI analysis, are there any differences between the conditions? Bayesian statistics might also be helpful to account for the negative results.

      We agree that our negative results required further validation. We conducted the ROI analyses suggested by Reviewer 1, which confirmed the initial whole-brain analyses.

      “Region of interest (ROI) results. We conducted additional analyses to test the robustness of our findings. One of our results was that we did not report any specific activation of the left area PF in the fluid-cognition task contrary to the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. However, this negative result needed exploration at the ROI level. Therefore, we created a spherical ROI of the left area PF with a radius of 12 mm in the MNI standard space (–59; –31; 40). This ROI was literature-defined to ensure the independence of its selection (40). ROI results are shown in Fig. 4. The analyses confirmed the results obtained with the whole-brain analyses by indicating a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .35).” (p. 10-11)

      Reviewer #1 (Recommendations For The Authors):

      (1) I may not fully grasp some of the arguments. In the abstract, what does the term "intermediate-level" mean, and why is it an intermediate-level state? In the sentence "the existence of a specific cognitive module in the human brain dedicated to materiality", I cannot see a clear link between technical cognition and the word "materiality".

      We used the term materiality to refer to a potential human trait that allows us to shape the physical world according to our ends, by using, making tools and transmiting them to others. This is a reference to Allen et al. (2020; PNAS): “We hope this empirical domain and modeling framework can provide the foundations for future research on this quintessentially human trait: using, making, and reasoning about tools and more generally shaping the physical world to our ends” (p. 29309). Scientists (including archaeologists, economists, psychologists, neuroscientists) interested in human materiality have tended to focus on how we manipulate things according to our thought (motor cognition) or how we conceptualize our behaviour to transmit it to others (language, social cognition). However, little has been said on the intermediate level, that is, technical cognition. We added the term “technical cognition” here, which should help to make the connection more quickly.

      “Yet, little has been said about the intermediate-level cognitive processes that are directly involved in mastering this materiality, that is, technical cognition.” (p. 2)

      (2) The introduction could provide more details on why the issue of "generalizability and specificity" is important to address, to clarify the significance of the research question.

      We followed this comment and added a sentence to explain why it is important to address this research question. Again, we thank Reviewer 1 for their helpful comments.

      “Here we focus on two key aspects of the technical-reasoning hypothesis that remain to be addressed: Generalizability and specificity. If technical reasoning is a specific form of reasoning oriented towards the physical world, then it should be implicated in all (the generalizability question) and only (the specificity question) the situations in which we need to think about the physical properties of our world.” (p. 5)

      Reviewer #2 (Public Review):

      Summary:

      The goal of this project was to test the hypothesis that a common neuroanatomic substrate in the left inferior parietal lobule (area PF) underlies reasoning about the physical properties of actions and objects. Four functional MRI (fMRI) experiments were created to test this hypothesis. Group contrast maps were then obtained for each task, and overlap among the tasks was computed at the voxel level. The principal finding is that the left PF exhibited differentially greater BOLD response in tasks requiring participants to reason about the physical properties of actions and objects (referred to as technical reasoning). In contrast, there was no differential BOLD response in the left PF when participants engaged in fMRI variant of the Raven's progressive matrices to assess fluid cognition.

      Strengths:

      This is a well-written manuscript that builds from extensive prior work from this group mapping the brain areas and cognitive mechanisms underlying object manipulation, technical reasoning, and problem-solving. Major strengths of this manuscript include the use of control conditions to demonstrate there are differentially greater BOLD responses in area PF over and above the baseline condition of each task. Another strength is the demonstration that area PF is not responsive in tasks assessing fluid cognition - e.g., it may just be that PF responds to a greater extent in a harder condition relative to an easy condition of a task. The analysis of data from Task 3 rules out this alternative interpretation. The methods and analysis are sufficiently written for others to replicate the study, and the materials and code for data analysis are publicly available.

      We sincerely thank Reviewer 2 for their precious comments, which helped us improve the MS. 

      Weaknesses:

      The first weakness is that the conclusions of the manuscript rely on there being overlap among group-level contrast maps presented in Figure 2. The problem with this conclusion is that different participants engaged in different tasks. Never is an analysis performed to demonstrate that the PF region identified in e.g., participant 1 in Task 2 is the same PF region identified in Participant 1 in Task 4.

      We added new analyses that demonstrated that “the PF region identified in e.g., participant 1 in Task 2 is the same PF region identified in Participant 1 in Task 4”. We thank Reviewer 2 for this comment, because these new analyses reinforced our interpretation.

      “The conjunction analysis reported was subject to at least two key limitations that needed to be overcome to assure a correct interpretation of our findings. The first was that the tasks could recruit the same regions for different cognition functions (same-region-but-different-function interpretation). The second was that the activated regions of the different tasks could be adjacent but did not overlap at finer resolutions (adjacency interpretation). We tested the same-region-but-different-function interpretation by conducting additional ROI analyses, which consisted of correlating the specific activation of the left area PF (i.e., difference in terms of mean Blood-Oxygen Level Dependent [BOLD] parameter estimates between the experimental condition minus the control condition) in the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. This analysis did not include the mechanical problem-solving task because the sample of participants was not the same for this task. As shown in Fig. 5, we found significant correlations between all the tasks that were hypothesized as recruiting technical reasoning, i.e., the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .05). By contrast, no significant correlation was obtained between these three tasks and the fluid-cognition task (all p > .15). This finding invalidates the same-region-but-different-function interpretation by revealing a coherent pattern in the activation of the left area PF in situations in which participants were supposed to reason technically. We examined the adjacency interpretation by analysing the specific locations of individual peak activations within the left area PF ROI for the mechanical problemsolving task, the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. These peaks, which corresponded to the maximum value of activation obtained for each participant within the left area PF ROI, are reported in Fig. 6. As can be seen, the peaks of the fluid-cognition task were located more anteriorly, in the left area PFt (Parietal Ft) and the postcentral cortex, compared to the peaks of the other four tasks, which were more posterior, in the left area PF. Statistical analyses based on the y coordinates of the individual activation peaks confirmed this description (Fig. 6). Indeed, the y coordinates of the peaks of the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task were posterior to the y coordinates of the peaks of the fluid-cognition task (all p < .05), whereas no significant differences were reported between the four tasks (all p > .05). These findings speak against the adjacency interpretation by revealing that participants recruited the same part of the left area PF to perform tasks involving technical reasoning.” (p. 11-13)

      A second weakness is that there is a variance in accuracy between tasks that are not addressed. It is clear from the plots in the supplemental materials that some participants score below chance (~ 50%). This means that half (or more) of the fMRI trials of some participants are incorrect. The methods section does not mention how inaccurate trials were handled. Moreover, if 50% is chance, it suggests that some participants did not understand task instructions and were systematically selecting the incorrect item.

      It is true that the experimental conditions were more difficult than the control conditions, with some participants who performed at or below 50% in the experimental conditions. We added a section in the MS to stress this aspect. To examine whether this potential difficulty effect biased our interpretation, we conducted new ROI analyses by removing all the participants who performed at or below the chance level. These analyses revealed the same results as when no participant was excluded, suggesting that this did not bias our interpretation.

      “As mentioned above, the experimental conditions of all the tasks were more difficult than their control conditions. As a result, the specific activation of the left area PF documented above could simply reflect that this area responds to a greater extent in a harder condition relative to an easy condition of a task. This interpretation is nevertheless ruled out by the results obtained with the fluid-cognition task. We did not report a specific activation of the left area PF in this task while its experimental condition was more difficult than its control condition. To test more directly this effect of difficulty, we conducted new ROI analyses by removing all the participants who performed at or below 50% (Fig. S2). These new analyses replicated the initial analyses by showing a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .48). In sum, the ROI analyses corroborated the wholebrain analyses and ruled out the potential effect of difficulty.” (p. 11)

      A third weakness is related to the fluid cognition task. In the fMRI task developed here, the participant must press a left or right button to select between 2 rows of 3 stimuli while only one of the 3 stimuli is the correct target. This means that within a 10-second window, the participant must identify the pattern in the 3x3 grid and then separately discriminate among 6 possible shapes to find the matching stimulus. This is a hard task that is qualitatively different from the other tasks in terms of the content being manipulated and the time constraints.

      We acknowledge that the fluid-cognition task involved a design that differed from the other tasks. However, this was also true for the other tasks, as the design also differed between the mechanical problem-solving task, the psychotechnical task, and the mentalizing task. Nevertheless, despite these distinctions, we found a consistent activation of the left area PF in these tasks with different designs including in the psychotechnical task, which seemed as difficult as the fluid-cognition task.

      “Region of interest (ROI) results. We conducted additional analyses to test the robustness of our findings. One of our results was that we did not report any specific activation of the left area PF in the fluid-cognition task contrary to the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. However, this negative result needed exploration at the ROI level. Therefore, we created a spherical ROI of the left area PF with a radius of 12 mm in the MNI standard space (–59; –31; 40). This ROI was literature-defined to ensure the independence of its selection (40). ROI results are shown in Fig. 4. The analyses confirmed the results obtained with the whole-brain analyses by indicating a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .35).” (p. 10-11)

      In sum, this is an interesting study that tests a neuro-cognitive model whereby the left PF forms a key node in a network of brain regions supporting technical reasoning for tool and non-tool-based tasks. Localizing area PF at the level of single participants and managing variance in accuracy is critically important before testing the proposed hypotheses.

      We thank Reviewer 2 for this positive evaluation and their suggestions. As detailed in our response, our revision took into consideration both the localization of the left area PF at the level of single participants and the variance in accuracy. 

      Reviewer #2 (Recommendations For The Authors):

      Did the fMRI data undergo high-pass temporal filtering prior to modeling the effects of interest? Participants engaged in a long (17-24 minutes) run of fMRI data collection. Highpass filtering of the data is critically important when managing temporal autocorrelation in the fMRI response (e.g., see Shinn et al., 2023, Functional brain networks reflect spatial and temporal autocorrelation. Nature Neuroscience).

      Yes. We added this information.

      “Regressors of non-interest resulting from 3D head motion estimation (x, y, z translation and three axes of rotation) and a set of cosine regressors for high-pass filtering were added to the design matrix.” (p. 25-26)

      Including scales in Figure 2 would help the reader interpret the magnitude of the BOLD effects.

      We added this information in Figure 3 (Figure 2 in the initial version of the MS).

      It was difficult to inspect the small thumbnail images of the task stimuli in Figure 1. Higher resolution versions of those stimuli would help facilitate understanding of the task design and trial structure.

      We changed both Figure 1 and Figure S1.

      Reviewer #3 (Public Review):

      Summary:

      This manuscript reports two neuroimaging experiments assessing commonalities and differences in activation loci across mechanical problem-solving, technical reasoning, fluid cognition, and "mentalizing" tasks. Each task includes a control task. Conjunction analyses are performed to identify regions in common across tasks. As Area PF (a part of the supramarginal gyrus of the inferior parietal lobe) is involved across 3 of the 4 tasks, the investigators claim that it is the hub of technical cognition.

      Strengths:

      The aim of finding commonalities and differences across related problem-solving tasks is a useful and interesting one.

      The experimental tasks themselves appear relatively well-thought-out, aside from the concern that they are differentially difficult.

      The imaging pipeline appears appropriate.

      We thank Reviewer 3 for their constructive comments, which helped us improve the MS.

      Weaknesses:

      (1) Methodological

      As indicated in the supplementary tables and figures, the experimental tasks employed differ markedly in 1) difficulty and 2) experimental trial time. Response latencies are not reported (but are of additional concern given the variance in difficulty). There is concern that at least some of the differences in activation patterns across tasks are the result of these fundamental differences in how hard various brain regions have to work to solve the tasks and/or how much of the trial epoch is actually consumed by "on-task" behavior. These difficulty issues should be controlled for by 1) separating correct and incorrect trials, and 2) for correct trials, entering response latency as a regressor in the Generalized Linear Models, 3) entering trial duration in the GLMs.

      We thank Reviewer 3 for this comment. It is true that the experimental conditions were more difficult than the control conditions, with some participants who performed at or below 50% in the experimental conditions. We added a section in the MS to stress this aspect. We could not conduct new analyses by separating correct and incorrect trials because, for each task, participants had to respond only on the last item of the block. Therefore, we did not record a response for each event. Nevertheless, we could examine whether this potential difficulty effect biased our interpretation, by conducting new ROI analyses in which we removed all the participants who performed at or below the chance level. These analyses revealed the same results as when no participant was excluded, suggesting that this did not bias our interpretation. 

      “As mentioned above, the experimental conditions of all the tasks were more difficult than their control conditions. As a result, the specific activation of the left area PF documented above could simply reflect that this area responds to a greater extent in a harder condition relative to an easy condition of a task. This interpretation is nevertheless ruled out by the results obtained with the fluid-cognition task. We did not report a specific activation of the left area PF in this task while its experimental condition was more difficult than its control condition. To test more directly this effect of difficulty, we conducted new ROI analyses by removing all the participants who performed at or below 50% (Fig. S2). These new analyses replicated the initial analyses by showing a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .48). In sum, the ROI analyses corroborated the wholebrain analyses and ruled out the potential effect of difficulty.” (p. 11)

      A related concern is that the control tasks also differ markedly in the degree to which they were easier and faster than their corresponding experimental task. Thus, some of the control tasks seem to control much better for difficulty and time on task than others. For example, the control task for the psychotechnical task simply requires the indication of which array contains a simple square shape (i.e., it is much easier than the psychotechnical task), whereas the control task for mechanical problem-solving requires mentally fitting a shape into a design, much like solving a jigsaw puzzle (i.e., it is only slightly easier than the experimental task).

      It is true that some control conditions could be easier than other ones. These differences reinforced the common activation found in the left area PF in the tasks hypothesized as involving technical reasoning, because this activation survived irrespective of the differences in terms of experimental design. For us, the rationale is the same as for a meta-analysis, in which we try to find what is common to a great variety of tasks. The only detrimental consequence we identified here is that this difference explained why we did not report a specific activation of the left area PF in the fluid-cognition task, as if the left area PF was more responsive when the task was difficult. This possibility assumes that the experimental condition of the fluid-cognition task is much more difficult than its control condition compared to what can be seen in the other tasks. As Reviewer 2 stressed in Point 1, this interpretation is unlikely, because the differences between the experimental and control conditions were similar to the fluid-cognition task in the mechanical problem-solving and psychotechnical tasks. In addition, again, the new ROI analyses in which we removed all the participants who performed at or below the chance level in expetimental conditions reproduced our initital results.

      (2) Theoretical 

      The investigators seem to overlook prior research that does not support their perspective and their writing seems to lack scientific objectivity in places. At times they over-reach in the claims that can be made based on the present data. Some claims need to be revised/softened.

      As this comment is also mentioned below, please find our response to it below.

      Reviewer #3 (Recommendations For The Authors):

      (1) Because of the high level of detail, Figures 1 and S2 (particularly the mentalizing task and mechanical problem-solving task, and their controls) are very hard to parse, even when examined relatively closely. It is suggested that these figures be broken down into separate panels for Experiment 1 and Experiment 2 to facilitate understanding.

      We changed both Figure 1 and Figure S1.

      (2) The behavioral data (including response latencies) should be reported in the main results section of the paper and not in a supplement.

      The behavioural data are now reported in the main results. We did not report response latencies because participants were not prompted to respond as quickly as possible.

      “Behavioural results. All the behavioural results are given in Fig. 2. As shown, scores were higher in the experimental conditions than for the control conditions for all the tasks (all p < .05). In other words, the experimental conditions were more difficult than the control conditions. This difference in terms of difficulty can also be illustrated by the fact that some participants performed at or below the chance level in the experimental conditions whereas none did so in the control conditions.” (p. 8)

      (3) The investigators seem to overlook prior research that does not support their perspective and their writing seems to lack scientific objectivity in places. At times they over-reach in the claims that can be made based on the present data. For example, claims that need to be revised/softened include:

      Abstract: "Area PF... can work along with social-cognitive skills to resolve day-to-day interactions that combine social and physical constraints". This statement is overly speculative.

      This statement is based on the fact that we reported a combined activation of the technical-reasoning network and the mentalizing network in the INT+PHYS condition of the mentalizing task. This suggests that both networks need to work together for solving a day-today problem in which both the physical constraints of the situation and the intention of the individual must be integrated. Our findings replicated previous ones with a similar task (e.g., Brunet et al. 2000; Völlm et al., 2006), in which the authors gave an interpretation similar to ours in considering that this task requires understanding physical and social causes. Perhaps that the reference to the results of the mentalizing task was not explicit enough. We added “dayto-day” before “problem” in the part of the discussion in which we discuss this possibility to make this aspect clearer.

      “In broad terms, the results of the mentalizing task indicate that causal reasoning has distinct forms and that it recruits distinct networks of the human brain (Social domain: Mentalizing; Physical domain: Technical reasoning), which can nevertheless interact together to solve day-to-day problems in which several domains are involved, such as in the INT+PHYS condition of the mentalizing task.” (p. 16)

      Introduction: "The manipulation-based approach... remains silent on the more general cognitive mechanisms...that must also encompass the use of unfamiliar or novel tools". This statement seems to be based on an overly selective literature review. There are a number of studies in which the relationship between a novel and familiar tool selection/use has been explored (e.g., Buchman & Randerath, 2017; Mizelle & Wheaton, 2010; Silveri & Ciccarelli, 2009; Stoll, Finkel et al., 2022; Foerster, 2023; Foerster, Borghi, & Goslin, 2020; Seidel, Rijntjes et al., 2023).

      We thank Reviewer 3 for this comment. Even if we accept the idea that we possess specific sensorimotor programs about tool manipulation, it remains that these programs cannot explain how an individual decides to bend a wire to make a hook or to pour water in a recipient to retrieve a target. As a matter of fact, such behaviour has been reported in nonhuman animals, such as crows (Weir et al., 2002, Nature) or orangutans (Mendes et al., 2007, Biology Letters). In these studies, the question is whether these nonhuman animals understand the physical causes or not, but the question of sensorimotor programs is never addressed (to our knowledge). This is also true in developmental studies on tool use (e.g., Beck et al., 2011, Cognition; Cutting et al., 2011, Journal of Experimental Child Psychology). This is what we meant here, that is, the manipulation-based approach is not equipped to explain how people solve physical problems by using or making tools – or any object – or by building constructions or producing technical innovations. However, we agree that some papers have been interested in exploring the link between common and novel tool use and have suggested that both could recruit common sensorimotor programs. It is noteworthy that these studies do not test the predictions from the manipulation-based approach versus the reasoning-based approach, so both interpretations are generally viable as stressed by Seidel et al. (2023), one of the papers recommended by Reviewer 3.

      “Apparently, the presentation of a graspable object that is recognizable as a tool is sufficient to provoke SMG activation, whether one tends to see the function of SMG to be either “technical reasoning” (Osiurak and Badets 2016; Reynaud et al. 2016; Lesourd et al. 2018; Reynaud et al. 2019) or “manipulation knowledge” (Sakreida et al. 2016; Buxbaum 2017; Garcea et al. 2019b).” (Seidel et al., 2023; p. 9)

      Regardless, as suggested by Reviewer 3, these papers deserve to be cited and this part needed to be rewritten to insist on the “making, construction, and innovation” dimension more than on the “unfamiliar and novel tool use” dimension to avoid any ambiguity.

      “This manipulation-based approach has provided interesting insights (12–16) and even elegant attempts to explain how these sensorimotor programs could support the use of both unfamiliar or novel tools (17–20), but remains silent on the more general cognitive mechanisms behind human technology that include the use of common and unfamiliar or novel tools but must also encompass tool making, construction behaviour, technical innovations, and transmission of technical content.” (p. 3)

      Introduction: "Here we focus on two important questions... to promote the technicalreasoning hypothesis as a comprehensive cognitive framework..."(italics added). This and other similar statements should be rewritten as testable scientific hypotheses rather than implying that the point of the research is to promote the investigators' preferred view.

      We agree that our phrasing could seem inappropriate here. What we meant here is that the technical-reasoning hypothesis could become an interesting framework for the study of the cognitive bases of human technology only if we are able to verify some of its key facets. As suggested, we rewrote this part. We also rewrote the abstract and the first paragraph of the discussion.

      “Here we focus on two key aspects of the technical-reasoning hypothesis that remain to be addressed: Generalizability and specificity. If technical reasoning is a specific form of reasoning oriented towards the physical world, then it should be implicated in all (the generalizability question) and only (the specificity question) the situations in which we need to think about the physical properties of our world.” (p. 5)

      Introduction: The Goldenberg and Hagmann paper cited actually shows that familiar tool use may be based either on retrieval from semantic memory or by inferring function from structure (mechanical problem solving); in other words, the investigators saw a role for both kinds of information, and the relationship between mechanical problem solving and familiar tool use was actually relatively weak. This requires correction.

      We disagree with Reviewer 3 on this point. The whole sentence is as follows:

      “This silence has been initially broken by a series of studies initiated by Goldenberg and Hagmann (9), which has documented a behavioural link in left brain-damaged patients between common tool use and the ability to solve mechanical problems by using and even sometimes making novel tools (e.g., extracting a target out from a box by bending a wire to create a hook) (9, 17).” (p. 3-4)

      We did not mention the interpretations given by Goldenberg and Hagmann about the link with the pantomime task, but only focused on the link they reported between common tool use and novel tool use. This is factual. In addition, we also disagree that the link between common tool use and novel tool use was weak.

      “The hypothesis put forward in the introduction predicts that knowledge about prototypical tool use assessed by pantomime of tool use and the ability to infer function from structure assessed by novel tool selection can both contribute to the use of familiar tools. Indeed results of both tests correlated signicantly with the use of familiar tools pantomime of tool use: r \= 0.77, novel tool selection: r \= 0.62; both P < 0.001), but there was also a signicant correlation between the two tests r \= 0.64, P < 0.001).” (Goldenberg & Hagmann, 1998; p. 585)

      As can be seen in this quote, they reported a significant correlation between novel tool selection and the use of familiar tools. It is also noteworthy that the novel tool selection test and the pantomime test correlated together. Georg Goldenberg told one of the authors (F. Osiurak; personal communication) that this result incited him to revise its idea that pantomime could assess “semantic knowledge”, which explains why he did not use it again as a measure of semantic knowledge. Instead, he preferred to use a classical semantic matching task in his 2009 Brain paper with Josef Spatt, in which they found a clearer dissociation between semantic knowledge and common/novel tool use not only at the behavioral level but also at the cerebral level.

      Introduction: Please expand and clarify this sentence "However, this involvement seems to be task-dependent, contrary to the systematic involvement of left are PF. The IFG and LOTC activations observed in prior studies are of interest as well. Were they indeed all taskdependent in these studies?

      We agree that this sentence is confusing. We meant that, in the studies reported just above in the paragraph, these regions were not systematically reported contrary to the left area PF. As we think that this information was not crucial for the logic of the paper, we preferred to remove it. 

      Introduction: If implicit mechanical knowledge is acquired through interactions with objects, how is that implicit knowledge conveyed to pass on the material culture to others?

      We thank Reviewer 3 for this comment. Although mechanical knowledge is implicit, it can be indirectly transmitted to other individuals, as shown in two papers we published in Nature Human Behaviour (Osiurak et al., 2021) and Science Advances (Osiurak et al., 2022). Actually, verbal teaching is not the only way to transmit information. There are many other ways of transmitting information such as gestural teaching (e.g., pointing the important aspects of a task to make them salient to the learner), observation without teaching (i.e., when we observe someone unbeknown to them) or reverse engineering (i.e., scrutinizing an artifact made by someone else). We have shown that even in reverse-engineering conditions, participants can benefit from what previous participants have done to increase their understanding of a physical system. In other words, all these forms of transmission allow the learners to understand new physical relationships without waiting that these relationships randomly occur in the environment. There is a wide literature on social learning, which describes very well how knowledge can be transmitted without using explicit communication. In fact, it is very likely that such forms of transmission were already present in our ancestors, allowing them to start accumulating knowledge without using symbolic language. We did not add this information in the MS because we think that this was a little bit beyond the scope of the MS. Nevetheless, we cited relevant literature on the topic to help the reader find it if interested in the topic.

      “Yet, recent accounts have proposed that non-social cognitive skills such as causal understanding or technical reasoning might have played a crucial role in cumulative technological culture (6, 29, 66). Support for these accounts comes from micro-society experiments, which have demonstrated that the improvement of technology over generations is accompanied by an increase in its understanding (67, 68), or that learners’ technical-reasoning skills are a good predictor of cumulative performance in such micro-societies (33, 69).” (p. 19)

      What distinguishes this implicit mechanical knowledge from stored knowledge about object manipulation? Are these two conceptualizations really demonstrably (testably) different?

      We agree that it is complex to distinguish between these two hypotheses as suggested by Seidel et al. (2023) cited above (see Reviewer 3 Point 8). We have conducted several studies to test the opposite predictions derived from each hypothesis. The main distinction concerns the understanding of physical materials and forces, which is central to the technical-reasoning hypothesis but not to the manipulation-based approach. Indeed, sensorimotor programs about tool manipulation are not assumed to contain information about physical materials and forces. In the present study, the understanding of physical materials and forces was needed in the four tasks hypothesized as requiring technical reasoning, i.e., the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task. We can illustrate this aspect with items of each of these tasks. Figure 1A is of the mechanical problem-solving task. 

      As explained in the MS, participants had memorized the five possible tools before the scanner session. Thus, for 4 seconds, they had to imagine which of these tools could be used to extract the target out from the box. We did so to incit them to reason about mechanical solutions based on the physical properties of the problem. Then, they had 3 seconds to select the tool with the appropriate shape, here the right one. In this case, the motor action remains the same (i.e., pulling). Another illustration can be given, with the psychotechnical task (Figure 1B).

      In this task, the participant had to reason as to whether the boat-tractor connection was better in the left picture or in the right picture. This needs to reason about physical forces, but there is no need to recruit sensorimotor programs about tool manipulation. Finally, a last example can be given with the PHYS-Only condition of the mentalizing task (but the logic is the same for the INT+PHYS condition except that the character’s intentions must also be taken into consideration) Figure 1D).

      Here the participant must reason about which picture shows what is physically possible. In this task, there is no need to recruit sensorimotor programs about tool manipulation. In sum, what is common between these three tasks is the requirement to reason about physical materials and forces. We do not ignore that motor actions could be simulated in the mechanical problemsolving task, but no motor action needed to be simulated in the other three tasks. Therefore, what was common between all these tasks was the potential involvement of technical reasoning but not of sensorimotor programs about tool manipulation. Of course, an alternative is to consider that motor actions are always needed in all the situations, including situations where no “manipulable tool” is presented, such as a tractor and a boat, a pulley, or a cannon. We cannot rule out this alternative, which is nevertheless, for us, prejudicial because it implies that it becomes difficult to test the manipulation-based approach as motor actions would be everywhere. We voluntarily decided not to introduce a debate between the reasoning-based approach and the manipulation-based approach and preferred a more positive writing by stressing the insights from the present study. Note that we stressed the merits of the manipulation-based approach in the introduction because we sincerely think that this approach has provided interesting insights. However, we voluntarily did not discuss the debate between the two approaches. Given Reviewer 3’s comment (see also Reviewer 1 Point 2), we understand and agree that some words must be nevertheless said to discuss how the manipulation-based approach could interpret our results, thus stressing the potential limitations of our interpretations. Therefore, we added a specific section in the discussion in which we discussed this aspect in more details.

      “The second limitation concerns the alternative interpretation that the left area PF is not central to technical reasoning but to the storage of sensorimotor programs about the prototypical manipulation of common tools. Here we show that the left area PF is recruited even in situations in which participants do not have to process common manipulable tools. For instance, some items of the psychotechnical task consisted of pictures of tractor, boat, pulley, or cannon. The fact that we found a common activation of the left area PF in such tasks as well as in the mechanical problem-solving task, in which participants could nevertheless simulate the motor actions of manipulating novel tools, indicates that this brain area is not central to tool manipulation but to physical understanding. That being said, some may suggest that viewing a boat or a cannon is enough to incite the simulation of motor actions, so our tasks were not equipped to distinguish between the manipulation-based approach and the reasoning-based approach. We have already shown that the left area PF is more involved in tasks that focus on the mechanical dimension of the tool-use action (e.g., the mechanical interaction between a tool and an object) than its motor dimension (i.e., the interaction between the tool and the effector [e.g., 24, 40]). Nevertheless, we recognize that future research is still needed to test the predictions derived from these two approaches.” (p. 18-19)

      Introduction and throughout: The framing of left Area PF as a special area for technical reasoning is overly reductionistic from a functional neuroanatomic perspective in that it ignores a large relevant literature showing that the region is involved with many other tasks that seem not to require anything like technical cognition. Indeed, entering the coordinates - 56, -29, 36 (reported as the peak coordinates in common across the studied tasks) in Neurosynth reveals that 59 imaging studies report activations within 3 mm of those coordinates; few are action-related (a brief review indicated studies of verbal creativity, texture processing, reading, somatosensory processing, stress reactions, attentional selection etc). Please acknowledge the difficulty of claiming that a large brain region should be labeled the brain's technical reasoning area when it seems to also participate in so much else. The left IPL (including area PF) is densely connected to the ventral premotor cortex, and this network is activated in language and calculation tasks as well as tool use tasks (e.g., Matsumoto, Nair, et al., 2012). What other constructs might be able to unite this disparate literature, and are any of these alternative constructs ruled out by the present data? Lacking this objective discussion, the manuscript does read as a promotion of the investigators' preferred viewpoint.

      We thank Reviewer 3 for this comment. As stressed in the initial version of the MS, we did not write that the left area PF is sufficient but central to the network that allows us to reason about the physical world. Regardless, we agree that an objective discussion was needed on this aspect to help the reader not misunderstand our purpose. We added a section in this aspect as suggested. 

      “Before concluding, we would like to point out two potential limitations of the present study. The first limitation concerns the fact that the literature has documented the recruitment of the left area PF in many neuroimaging experiments in which there was no need to reason about physical events (e.g., language tasks). This can be easily illustrated by entering the left area PF coordinates in the Neurosynth database.

      This finding could be enough to refute the idea that this brain area is specific to technical reasoning. Although this limitation deserves to be recognized, it is also true for many other findings. For instance, sensory or motor brain regions such as the precentral or the postcentral cortex have been found activated in many non-motor tasks, the visual word form area in non-language tasks, or the Heschl’s gyrus in nonmusical tasks. This remains a major challenge for scientists, the question being how to solve these inconsistencies that can result from statistical errors or stress that considerable effort is needed to understand the very functional nature of these brain areas. Thus, understanding that the left area PF is central to physical understanding can be viewed as a first essential step before discovering its fundamental function, as suggested by the functional polyhedral approach (56).” (p. 18)

      Discussion: The discussion of a small cluster in the IFG (pars opercularis) that nearly survived statistical correction is noteworthy in light of the above point. This further underscores the importance of discussing networks and not just single brain regions (such as area PF) when examining complex processes. The investigators note, "a plausible hypothesis is that the left IFG integrates the multiple constraints posed by the physical situation to set the ground for a correct reasoning process, such as it could be involved in syntactic language processing". In fact, the hypothesis that the IFG and SMG are together related to resolving competition has been previously proposed, as has the more specific hypothesis that the SMG buffers actions and that the context-appropriate action is then selected by the IFG (e.g., Buxbaum & Randerath, 2018). The parallels with the way the SMG is engaged with competing lexical or phonological alternatives (e.g., Peramunage, Blumstein et al., 2011) have also been previously noted.

      We added the Buxbaum and Randerath (2018)’s reference in this section.

      “The functional role of the left IFG in the context of tool use has been previously discussed (24) and a plausible hypothesis is that the left IFG integrates the multiple constraints posed by the physical situation to set the ground for a correct reasoning process, such as it could be involved in syntactic language processing (for a somewhat similar view, see [51]).” (p. 16-17)

      Introduction and Discussion: Please clarify how the technical reasoning network overlaps with or is distinct from the tool-use network reported by many previous investigators.

      We added a couple of sentences in the discussion to clarify this point.

      “It should be clear here that we do not advocate the localizationist position simply stating that activation in the left area PF is the necessary and sufficient condition for technical reasoning. We rather defend the view according to which it requires a network of interacting brain areas, one of them – and of major importance – being the left area PF. This allows the engagement of different configurations of cerebral areas in different technical-reasoning tasks, but with a central process acting as a stable component: The left area PF. Thus, when people intend to use physical tools, it can work in concert with brain regions specific to object manipulation and motor control, thereby forming another network, the tool-use network. It can also interact with brain regions specific to intentional gestures to form a “social-learning” network that allows people to enhance their understanding about the physical aspects of a technical task (e.g., the making of a tool) through communicative gestures such as pointing gestures (42). The major challenge for future research is to specify the nature of the cognitive process supported by the left area PF and that might be involved in the broad understanding of the physical world.” (p. 14)

      Discussion: All of the experimental tasks require a response from a difficult choice in an array, and all of the tasks except for the fluid cognition task are likely to require prediction or simulation of a motion trajectory-whether an embodied or disembodied trajectory is unclear. The Discussion does mention the related (but distinct) idea of an "intuitive physics engine", a "kind of simulator", Please clarify how this study can rule out these alternative interpretations of the data. If the study cannot rule out these alternatives, the claims of the study (and the paper title which labels PF as a technical cognition area) should be scaled back considerably. 

      We thank Reviewer 3 for this comment. The authors of the papers on intuitive physics engine or associative learning do not suggest that these processes are embodied. As discussed above, we clarified our perspective on the role of the left area PF and hope that these modifications help the reader better understand it. We warmly thank Reviewer 3 for their comments, which considerably helped us improve the MS.

    1. eLife Assessment

      This important study substantially advances our understanding of the circadian clock in Antarctic krill, a key species in the Southern Ocean ecosystem. Through logistically challenging shipboard experiments conducted across seasons, the authors provide compelling evidence for their conclusions. The study will be of broad interest to marine biologists and ecologists.

    2. Reviewer #1 (Public review):

      Hüppe and colleagues had already developed an apparatus and an analytical approach to capture swimming activity rhythms in krill. In a previous manuscript they explained the system, and here they employ it to show a circadian clock, supplemented by exogenous light, produces an activity pattern consistent with "twilight" diel vertical migration (DVM; a peak at sunset, a midnight sink, and a peak in the latter half of the night).

      They used light:dark (LD) followed by dark:dark (DD) photoperiods at two times of the year to confirm the circadian clock, coupled with DD experiments at four times a year to show rhythmicity occurs throughout the year along with DVM in the wild population. The individual activity data show variability in the rhythmic response, which is expected. However, their results showed rhythmicity was sustained in DD throughout the year, although the amplitude decayed quickly. The interpretation of a weak clock is reasonable, and they provide a convincing justification for the adaptive nature of such a clock in a species that has a wide distributional range and experiences various photic environments. These data also show that exogenous light increases the activity response and can explain the morning activity bouts, with the circadian clock explaining the evening and late-night bouts. This acknowledgement that vertical migration can be driven by multiple proximate mechanisms is important.

      The work is rigorously done, and the interpretations are sound. I see no major weaknesses in the manuscript. Because a considerable amount of processing is required to extract and interpret the rhythmic signals (see Methods and previous AMAZE paper), it is informative to have the individual activity plots of krill as a gut check on the group data.

      The manuscript will be useful to the field as it provides an elegant example of looking for biological rhythms in a marine planktonic organism and disentangling the exogenous response from the endogenous one. Furthermore, as high-latitude environments change, understanding how important organisms like krill have the potential to respond will become increasingly important. This work provides a solid behavioral dataset to complement the earlier molecular data suggestive of a circadian clock in this species.

    3. Reviewer #2 (Public review):

      Summary:

      This manuscript provides experimental evidence on circadian behavioural cycles in Antarctic krill. The krill were obtained directly from krill fishing vessels and the experiments were carried out on board using an advanced incubation device capable of recording activity levels over a number of days. A number of different experiments were carried out where krill were first exposed to simulated light:dark (L:D) regimes for some days followed by continuous darkness (DD). These were carried out on krill collected during late autumn and late summer. A further set of experiments was performed on krill across three different seasons (summer, autumn, winter), where incubations were all DD conditions. Activity was measured as the frequency by which an infrared beam close to the top of the incubation tube was broken over unit time. Results showed that patterns of increased and decreased activity that appeared synchronised to the LD cycle persisted during the DD period. This was interpreted as evidence of the operation of an internal (endogenous) clock. The amplitude of the behavioural cycles decreased with time in DD, which further suggests that this clock is relatively weak. The authors argued that the existence of a weak endogenous clock is an adaptation to life at high latitudes since allowing the clock to be modulated by external (exogenous) factors is an advantage when there is a high degree of seasonality. This hypothesis is further supported by seasonal DD experiments which showed that the periodicity of high and low activity levels differed between seasons.

      Strengths:

      Although there has been a lot of field observations of various circadian type behaviour in Antarctic krill, relatively few experimental studies have been published considering this behaviour in terms of circadian patterns of activity. Krill are not a model organism and obtaining them and incubating them in suitable conditions are both difficult undertakings. Furthermore, there is a need to consider what their natural circadian rhythms are without the overinfluence of laboratory-induced artefacts. For this reason alone, the setup of the present study is ideal to consider this aspect of krill biology. Furthermore, the equipment developed for measuring levels of activity is well-designed and likely to minimise artefacts.

    4. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1 (Public review):

      Hüppe and colleagues had already developed an apparatus and an analytical approach to capture swimming activity rhythms in krill. In a previous manuscript they explained the system, and here they employ it to show a circadian clock, supplemented by exogenous light, produces an activity pattern consistent with "twilight" diel vertical migration (DVM; a peak at sunset, a midnight sink, and a peak in the latter half of the night).

      They used light:dark (LD) followed by dark:dark (DD) photoperiods at two times of the year to confirm the circadian clock, coupled with DD experiments at four times of year to show rhythmicity occurs throughout the year along with DVM in the wild population. The individual activity data show variability in the rhythmic response, which is expected. However, their results showed rhythmicity was sustained in DD throughout the year, although the amplitude decayed quickly. The interpretation of a weak clock is reasonable, and they provide a convincing justification for the adaptive nature of such a clock in a species that has a wide distributional range and experiences various photic environments. These data also show that exogenous light increases the activity response and can explain the morning activity bouts, with the circadian clock explaining the evening and late-night bouts. This acknowledgement that vertical migration can be driven by multiple proximate mechanisms is important.

      The work is rigorously done, and the interpretations are sound. I see no major weaknesses in the manuscript. Because a considerable amount of processing is required to extract and interpret the rhythmic signals (see Methods and previous AMAZE paper), it is informative to have the individual activity plots of krill as a gut check on the group data.

      The manuscript will be useful to the field as it provides an elegant example of looking for biological rhythms in a marine planktonic organism and disentangling the exogenous response from the endogenous one. Furthermore, as high latitude environments change, understanding how important organisms like krill have the potential to respond will become increasingly important. This work provides a solid behavioral dataset to complement the earlier molecular data suggestive of a circadian clock in this species.

      We appreciate the positive evaluation of our work by Reviewer 1, acknowledging our approach to record locomotor activity in krill and the importance of the findings in assessing krill’s potential to respond to environmental change in their habitat.

      Reviewer #2 (Public review):

      Summary:

      This manuscript provides experimental evidence on circadian behavioural cycles in Antarctic krill. The krill were obtained directly from krill fishing vessels and the experiments were carried out on board using an advanced incubation device capable of recording activity levels over a number of days. A number of different experiments were carried out where krill were first exposed to simulated light:dark (L:D) regimes for some days followed by continuous darkness (DD). These were carried out on krill collected during late autumn and late summer. A further set of experiments was performed on krill across three different seasons (summer, autumn, winter), where incubations were all DD conditions. Activity was measured as the frequency by which an infrared beam close to the top of the incubation tube was broken over unit time. Results showed that patterns of increased and decreased activity that appeared synchronised to the LD cycle persisted during the DD period. This was interpreted as evidence of the operation of an internal (endogenous) clock. The amplitude of the behavioural cycles decreased with time in DD, which further suggests that this clock is relatively weak. The authors argued that the existence of a weak endogenous clock is an adaptation to life at high latitudes since allowing the clock to be modulated by external (exogenous) factors is an advantage when there is a high degree of seasonality. This hypothesis is further supported by seasonal DD experiments which showed that the periodicity of high and low activity levels differed between seasons.

      Strengths

      Although there has been a lot of field observations of various circadian type behaviour in Antarctic krill, relatively few experimental studies have been published considering this behaviour in terms of circadian patterns of activity. Krill are not a model organism and obtaining them and incubating them in suitable conditions are both difficult undertakings. Furthermore, there is a need to consider what their natural circadian rhythms are without the overinfluence of laboratory-induced artefacts. For this reason alone, the setup of the present study is ideal to consider this aspect of krill biology. Furthermore, the equipment developed for measuring levels of activity is well-designed and likely to minimise artefacts.

      We would like to thank Reviewer 2 for their positive assessment of our approach to study the influence of the circadian clock on krill behavior. We are delighted, that Reviewer 2 found our mechanistic approach in understanding daily behavioral patterns of Antarctic krill using the AMAZE set-up convincing, and that the challenging circumstances of working with a polar, non-model species are acknowledged.

      Weaknesses

      I have little criticism of the rationale for carrying out this work, nor of the experimental design. Nevertheless, the manuscript would benefit from a clearer explanation of the experimental design, particularly aimed at readers not familiar with research into circadian rhythms. Furthermore, I have a more fundamental question about the relationship between levels of activity and DVM on which I will expand below. Finally, it was unclear how the observational results made here related to the molecular aspects considered in the Discussion.

      (1) Explanation of experimental design - I acknowledge that the format of this particular journal insists that the Results are the first section that follows the Introduction. This nevertheless presents a problem for the reader since many of the concepts and terms that would generally be in the Methods are yet to be explained to the reader. Hence, right from the start of the Results section, the reader is thrown into the detail of what happened during the LD-DD experiments without being fully aware of why this type of experiment was carried out in the first place. Even after reading the Methods, further explanation would have been helpful. Circadian cycle type research of this sort often entrains organisms to certain light cycles and then takes the light away to see if the cycle continues in complete darkness, but this critical piece of knowledge does not come until much later (e.g. lines 369-372) leaving the reader guessing until this point why the authors took the approach they did. I would suggest the following (1) that more effort is made in the Introduction to explain the exact LD/DD protocols adopted (2) that a schematic figure is placed early on in the manuscript where the protocol is explained including some logical flow charts of e.g. if behavioural cycle continues in DD then internal clock exists versus if cycle does not continue in DD, the exogenous cues dominate - followed by - major decrease in cyclic amplitude = weak clock versus minor decrease = strong clock and so on

      We want to thank Reviewer 2 for pointing out that the experimental design and its rationale are not becoming clear early in the manuscript, especially for people outside the field of chronobiology. We added a new figure (now Fig. 1), illustrating the basic principle of chronobiological study design and how we adopted it. We also extended the description at the beginning of the Results section to clarify the rationale behind the experimental design.

      (2) Activity vs kinesis - in this study, we are shown data that (i) krill have a circadian cycle - incubation experiments; (ii) that krill swarms display DVM in this region - echosounder data (although see my later point). My question here is regarding the relationship between what is being measured by the incubation experiments and the in situ swarm behaviour observations. The incubation experiments are essentially measuring the propensity of krill to swim upwards since it logs the number of times an individual (or group) break a beam towards the top of the incubation tube. I argue that krill may be still highly active in the rest of the tube but just do not swim close to the surface, so this approach may not be a good measure of "activity". Otherwise, I suggest a more correct term of what is being measured is the level of "upward kinesis". As the authors themselves note, krill are negatively buoyant and must always be active to remain pelagic. What changes over the day-night cycle is whether they decide to expend that activity on swimming upwards, downwards or remaining at the same depth. Explaining the pattern as upward kinesis then also explains by swarms move upwards during the night. Just being more active at night may not necessarily result in them swimming upwards.

      We believe there is a slight misunderstanding in how what we call “activity” is measured. The experimental columns are equipped with five detector modules, evenly distributed over the height of the column. In our analysis we count all beam breaks caused by upward movement, i.e. every time a detector module is triggered after a detector module at a lower position has been triggered, and not only when the top detector module is triggered. In this way, we record upward swimming movements throughout the column, and not only when the krill swims all the way to the top of the column. This still means that what we are measuring is swimming activity, caused by upward swimming. We use this measure, to deliberately separate increased swimming activity, from baseline activity (i.e. swimming, which solely compensates for negative buoyancy) and inactivity (i.e. passive sinking).

      Higher activity is thus at first interpreted as an increase in swimming activity, which in the field may result in upwards-directed swimming but also could mean a horizontal increase in activity, for example, representing increased foraging and feeding activity. This would explain the daily activity pattern observed under LD cycles (now Fig. 3), which shows a general increase in activity during the dark phase. This nighttime increase could be used for both upward directed migration during sunset and horizontal directed swimming for feeding and foraging throughout the night.

      We added the following sentence to the description of the activity metric in the Methods section to clarify this point (lines 465-469):

      “To accomplish this, we organized the raw beam break data from all five detector modules in each experimental column in chronological order. We selected only those beam break detections that occurred after a detection in the detector module positioned lower on the column. Like this, we consider upward swimming movements throughout the full height of the column.”

      (3) Molecular relevance - Although I am interested in molecular clock aspects behind these circadian rhythms, it was not made clear how the results of the present study allow any further insight into this. In lines 282 to 284, the findings of the study by Biscontin et al (2017) are discussed with regard to how TIM protein is degraded by light via the clock photreceptor CRYTOCHROME 1. This element of the Discussion would be a lot more relevant if the results of the present study were considered in terms of whether they supported or refuted this or any other molecular clock model. As it stands, this paragraph is purely background knowledge and a candidate for deletion in the interest of shortening the Discussion.

      We agree that this part is not directly related to the data presented in the manuscript. We, therefore, omitted this part in the revised version of the manuscript to keep the discussion concise and focused on the results.

      Other aspects

      (i) 'Bimodal swimming' was used in the Abstract and later in the text without the term being fully explained. I could interpret it to mean a number of things so some explanation is required before the term is introduced.

      We thank the Reviewer for pointing this out. We provided an explanation for the term “bimodal” in the Results section, where the two clock driven activity bouts are described first, by extending the sentence in lines 161-164, which now reads:

      “This suggests that the circadian clock drives a distinct bimodal activity pattern with two activity peaks in one day, i.e. the evening and late-night activity bouts, while. In contrast, the morning activity bout is triggered by the onset of illumination in the experimental set-up.”.

      (ii) Midnight sinking - I was struck by Figure 2b with regards to the dip in activity after the initial ascent, as well as the rise in activity predawn. Cushing (1951) Biol Rev 26: 158-192 describes the different phases of a DVM common to a number of marine organisms observed in situ where there is a period of midnight sinking following the initial dusk ascent and a dawn rise prior to dawn descent. Tarling et al (2002) observe midnight sinking pattern in Calanus finmarchicus and consider whether it is a response to feeding satiation or predation avoidance (i.e. exogenous factors). Evidence from the present study indicates that midnight sinking (and potential dawn rise) behaviour could alternatively be under endogenous control to a greater or lesser degree. This is something that should certainly be mentioned in the Discussion, possibly in place of the molecular discussion element mentioned above - possibly adding to the paragraph Lines 303-319.

      We would like to thank the Reviewer for pointing this out and agree that adding the idea of an endogenous control of midnight sinking would be interesting to the discussion. We added the following section to the Discussion (lines 335-343):

      “Interestingly, the decrease in clock-controlled swimming activity during the early night, right after the evening activity bout, may further facilitate a phenomenon called “midnight sinking”, which describes the sinking of animals to intermediate depths after the evening ascent, followed by a second rise to the surface before the morning descend. This behavior has been observed in a number of zooplankton species, including calanoid copepods (see 69, 70 and references therein) and krill (71). While previous studies suggested several exogenous factors, such as satiation or predator presence, as drivers of the midnight sink (69, 70), our study suggests that this pattern may be partly under endogenous control.”

      (iii) Lines 200-207 - I struggled to follow this argument regarding Piccolin et al identifying a 12 h rhythm whereas the present study indicates a ~24 h rhythm. Is one contradicting the other - please make this clear.

      In our study, we found that the circadian clock drives a bimodal pattern of swimming activity in krill, meaning it controls two bouts of activity in a 24-hour cycle. Piccolin et al. (2020) identified a swimming activity pattern of ~12 h (i.e. two peaks in 24 h) at the group level, which aligns with our findings at the individual level. We revised the Section in the discussion for more clarity, which now reads:

      “Data from Piccolin et al. (20) showed a strong damping of the amplitude and indication of a remarkably short (~12 h) free running period (FRP) of vertical swimming behavior of a group of krill under constant darkness (20). The short period found in Piccolin et al. (20) complements is in line with our findings of a bimodal activity pattern the pattern of swimming activity under DD conditions on the individual level found in the present study, suggesting that the ~12 h rhythm in group swimming behavior in Piccolin et al. (20) could have resulted from a bimodal activity pattern at the individual level, as found in our study.” (lines 212-219).  

      (iv) Although I agree that the hydroacoustic data should be included and is generally supportive of the results, I think that two further aspects should be made clear for context (a) whether there was any groundtruthing that the acoustic marks were indeed krill and not potentially some other group know to perform DVM such as myctophids (b) how representative were these patterns - I have a sense that they were heavily selected to show only ones with prominent DVM as opposed to other parts of the dataset where such a pattern was less clear - I am aware of a lot of krill research where DVM is not such a clear pattern and it is disingenuous to provide these patterns as the definitive way in which krill behaves. I ask this be made clear to the reader (note also that there is a suggestion of midnight sinking in Fig 5b on 28/2).

      To clarify the mentioned points concerning the hydroacoustic data:

      a) As mentioned in the Methods section, only hydroacoustic data during active fishing was included in the analysis. E. superba occurs in large monospecific aggregations, and the fishery actively targets E. superba and monitors their catch and the proportion of non-target species continuously with cameras. Krill fishery bycatch rates are very low (0.1–0.3%, Krafft et al. 2022), and fishing operations would stop if non-target species were caught in significant proportions at any time. Therefore, and supported by our own observations when we conducted the experiments, we argue that it is a valid assumption that E. superba predominantly causes the backscattering signal shown in Figure 5 (now Fig. 6).

      b) We are aware of the fact that DVM patterns of Antarctic krill are highly variable and that normal DVM patterns do not need to be the rule (e.g. see our cited study on the plasticity of krill DVM by Bahlburg et al. 2023). The visualized data were not selected for their DVM pattern but represent the period directly preceding the sampling for behavioral experiments in four seasons (experiment 2), including the day of sampling. These periods were chosen to assess the DVM behavior of krill swarms in the field in the days before and during the sampling for behavioral experiments.

      To improve understanding, we modified the description in the Results, Discussion, and Methods sections, as well as the caption of Figure 5 (now Fig. 6), which now read:

      “To investigate whether krill swarms exhibited daily behavioral patterns in swimming behavior in the field before they were sampled for seasonal experiments, hydroacoustic data were recorded from the fishing vessel, continuously over a three-day period prior to sampling for the seasonal experiments described above…” (lines 191-194).

      “Furthermore, hydroacoustic recordings demonstrate that most krill swarms sampled exhibited synchronized DVM in the field in the days directly before sampling for behavioral experiments, indicating that in this region, krill remain behaviorally synchronized across a wide range of photoperiods.” (lines 397-400).

      “Hydroacoustic data were collected using a hull-mounted SIMRAD ES80 echosounder (Kongsberg Maritime AS) aboard the Antarctic Endurance, covering three days before the sampling for each of the seasonal behavioral experiments of experiment 2” (lines 512-515).

      “We only included data during active fishing periods and the vessel is specifically targeting E. superba, which occurs in large monospecific aggregations. Further, krill fishery bycatch rates are very low (0.1-0.3%, 84), which makes it highly probable that the recorded signal represents krill swarms.” (lines 523-526).

      “Hydroacoustic recordings showing the vertical distribution of krill swarms in the upper water column (<220 m) below the vessel, visualized by the mean volume backscattering signal (200 kHz), on the three days prior to krill sampling for experiments…” (lines 802-804).

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      As noted in the public review, this is a logical and well-written manuscript. I have very few comments to consider addressing.

      The Results lead with a paragraph outlining the experimental approach. This is good, but you use the term "experiments" to refer to both the two sets, and the two or four subsets of experiments. Perhaps consider the subset experiments as "treatments"? I understood what you meant, but it took a few read-throughs to be sure I got it.

      We thank the reviewer for pointing this out and changed the nomenclature of the experiments throughout the manuscript. We now refer to the two sets of experiments as experiment 1 and 2, to the subsets of experiment 1 as “short day treatment” and “long day treatment”, and to the subsets of experiment 2 as summer treatment, late summer treatment, autumn treatment, and winter treatment. We also believe that the new Figure 1 is now helping to follow the experimental design more efficiently.

      Ln 140: "...off and decrease at lights-on."

      We adjusted the sentence accordingly.

      Ln 244: Can you define "extreme photic conditions"? I get what you mean, but to be clear to the reader this would help.

      We adjusted the sentence, which now reads:

      “This could confer a significant adaptive advantage to species inhabiting environments characterized by extreme photic conditions (53, 54, 60), such as phases of polar night or midnight sun as well as rapid changes in daylength, or species that rely on precise photoperiodic time measurement for accurate seasonal adaptation.” (lines 258-261).

      Figures: Consider adding an LSP for groups in Fig 1. Also, it would be useful to have LSP period estimates for each individual tested. This could be a separate table, or it could be added to the individual activity plots. Should S3 and S4 be reversed?

      We thank the reviewer for their suggestion and added an LSP as figure 1d (now Fig. 2d) to statistically support the group activity shown in Figure 1c (now Fig. 2c) as suggested. We added the individual animals' LSP period estimates to supplementary figures S2, S7, S8, S9, and S10. We also reversed Figures S3 and S4 to match the appearance in the main text. 

      Fig 5: are the light regime bars for b and c correct? They look similar, but there are only 15 days apart, so perhaps they are correct as is.

      We double checked the light regime bars in Fig. 5b and c (now 6b and c) and they are correct as is.

    1. eLife Assessment

      This important study assessed the effects of food intake on sharp wave-ripples in the hippocampus of mice during subsequent sleep. Convincing evidence supports the conclusion that sharp wave-ripples are enhanced by food consumption. This work will likely interest researchers studying multiple functions including memory, metabolism, and brain-body physiology.

    2. Reviewer #1 (Public review):

      Summary:

      This manuscript by Kaya et al. studies the effect of food consumption on hippocampal sharp wave ripples (SWRs) in mice. The authors use multiple foods and forms of food delivery to show that the frequency and power of SWRs increase following food intake, and that this effect depends on the caloric content of food. The authors also studied the effects of administration of various food-intake-related hormones on SWRs during sleep, demonstrating that ghrelin negatively affects SWR rate and power, but not GLP-1, insulin, or leptin. Finally, the authors use fiber photometry to show that GABAergic neurons in the lateral hypothalamus, increase activity during a SWR event.

      Strengths:

      The experiments in this study seem to be well performed, and the data are well presented, visually. The data support the main conclusions of the manuscript that food intake enhances hippocampal SWRs. Taken together, this study is likely to be impactful to the study of the impact of feeding on sleep behavior, as well as the phenomena of hippocampal SWRs in metabolism.

      Weaknesses:

      None

    3. Reviewer #2 (Public review):

      Summary:

      Kaya et al uncover an intriguing relationship between hippocampal sharp wave-ripple production and peripheral hormone exposure, food intake, and lateral hypothalamic function. These findings significantly expand our understanding of hippocampal function beyond mnemonic processes and point a direction for promising future research.

      Strengths:

      Some of the relationships observed in this paper are highly significant. In particular, the inverse relationship between GLP1/Leptin and Insulin/Ghrelin are particularly compelling as this aligns well with opposing hormone functions on satiety.

    4. Reviewer #3 (Public review):

      Summary:

      The manuscript by Kaya et al. explores the effects of feeding on sharp wave-ripples (SWRs) in the hippocampus, which could reveal a better understanding of how metabolism is regulated by neural processes. Expanding on prior work that showed that SWRs trigger a decrease in peripheral glucose levels, the authors further tested the relationship between SWRs and meal consumption by recording LFPs from the dorsal CA1 region of the hippocampus before and after meal consumption. They found an increase in SWR magnitude during sleep after food intake, in both food-restricted and ad libitum fed conditions. Using fiber photometry to detect GABAergic neuron activity in the lateral hypothalamus, they found increased activity locked to the onset of SWRs. They conclude that the animal's satiety state modulates the amplitude and rate of SWRs, and that SWRs modulate downstream circuits involved in regulating feeding.

      The authors have addressed prior requests for revision and clarification, and provide a convincing case for SWRs being modulated by satiety state. These experiments provide an important step forward in understanding how metabolism is regulated in the brain. The study will likely be of great interest in the field of learning and memory while carrying broader implications for understanding brain-body physiology.

    5. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      This manuscript by Kaya et al. studies the effect of food consumption on hippocampal sharp wave ripples (SWRs) in mice. The authors use multiple foods and forms of food delivery to show that the frequency and power of SWRs increase following food intake, and that this effect depends on the caloric content of food. The authors also studied the effects of the administration of various food-intake-related hormones on SWRs during sleep, demonstrating that ghrelin negatively affects SWR rate and power, but not GLP1, insulin, or leptin. Finally, the authors use fiber photometry to show that GABAergic neurons in the lateral hypothalamus, increase activity during a SWR event.

      Strengths:

      The experiments in this study seem to be well performed, and the data are well presented, visually. The data support the main conclusions of the manuscript that food intake enhances hippocampal SWRs. Taken together, this study is likely to be impactful to the study of the impact of feeding on sleep behavior, as well as the phenomena of hippocampal SWRs in metabolism.

      Weaknesses:

      Details of experiments are missing in the text and figure legends. Additionally, the writing of the manuscript could be improved.

      We thank the reviewer for their favorable assessment of the work and its potential impact. We have added all requested details in the text and figure legends and revised the wording of the manuscript to improve its clarity.

      Reviewer #2 (Public review):

      Summary:

      Kaya et al uncover an intriguing relationship between hippocampal sharp wave-ripple production and peripheral hormone exposure, food intake, and lateral hypothalamic function. These findings significantly expand our understanding of hippocampal function beyond mnemonic processes and point a direction for promising future research.

      Strengths:

      Some of the relationships observed in this paper are highly significant. In particular, the inverse relationship between GLP1/Leptin and Insulin/Ghrelin are particularly compelling as this aligns well with opposing hormone functions on satiety.

      Weaknesses:

      I would be curious if there were any measurable behavioral differences that occur with different hormone manipulations.

      We thank the reviewer for their favorable assessment of the work and its contribution to our understanding of non-mnemonic hippocampal function. Whether there are behavioral differences that occur following administration of the different hormones is a great question, yet unfortunately our study design did not include fine behavioral monitoring to the degree that would allow answering it. While some previous studies have partially addressed the behavioral consequences of the delivery of these hormones (and we reference these studies in our Discussion), how these changes may interact with the hippocampal and hypothalamic effects we observe is a very interesting next step.

      Reviewer #3 (Public review):

      Summary:

      The manuscript by Kaya et al. explores the effects of feeding on sharp wave-ripples (SWRs) in the hippocampus, which could reveal a better understanding of how metabolism is regulated by neural processes. Expanding on prior work that showed that SWRs trigger a decrease in peripheral glucose levels, the authors further tested the relationship between SWRs and meal consumption by recording LFPs from the dorsal CA1 region of the hippocampus before and after meal consumption. They found an increase in SWR magnitude during sleep after food intake, in both food restricted and ad libitum fed conditions. Using fiber photometry to detect GABAergic neuron activity in the lateral hypothalamus, they found increased activity locked to the onset of SWRs. They conclude that the animal's satiety state modulates the amplitude and rate of SWRs, and that SWRs modulate downstream circuits involved in regulating feeding. These experiments provide an important step forward in understanding how metabolism is regulated in the brain. However, currently, the paper lacks sufficient analyses to control for factors related to sleep quality and duration; adding these analyses would further support the claim that food intake itself, as opposed to sleep quality, is primarily responsible for changes in SWR activity. Adding this, along with some minor clarifications and edits, would lead to a compelling case for SWRs being modulated by a satiety state. The study will likely be of great interest in the field of learning and memory while carrying broader implications for understanding brain-body physiology.

      Strengths:

      The paper makes an innovative foray into the emerging field of brain-body research, asking how sharp wave-ripples are affected by metabolism and hunger. The authors use a variety of advanced techniques including LFP recordings and fiber photometry to answer this question. Additionally, they perform comprehensive and logical follow-up experiments to the initial food-restricted paradigm to account for deeper sleep following meal times and the difference between consumption of calories versus the experience of eating. These experiments lay the groundwork for future studies in this field, as the authors pose several follow-up questions regarding the role of metabolic hormones and downstream brain regions.

      We thank the reviewer for their appreciation and constructive review of the work.

      Weaknesses:

      Major comments:

      (1) The authors conclude that food intake regulates SWR power during sleep beyond the effect of food intake on sleep quality. Specifically, they made an attempt to control for the confounding effect of delta power on SWRs through a mediation analysis. However, a similar analysis is not presented for SWR rate. Moreover, this does not seem to be a sufficient control. One alternative way to address this confound would be to subsample the sleep data from the ad lib and food restricted conditions (or high calorie and low calorie, etc), to match the delta power in each condition. When periods of similar mean delta power (i.e. similar sleep quality) are matched between datasets, the authors can then determine if a significant effect on SWR amplitude and rate remains in the subsampled data.

      This is an important point that we believe we addressed in a few complementary ways. First, the mediation analysis we implemented measures the magnitude and significance of the contribution of food on SWR power after accounting for the effects of delta power, showing a highly significant food-SWR contribution. While the objective of subsampling is similar, mediation is a more statistically robust approach as it models the relationship between food, SWR power, and delta power in a way that explicitly accounts for the interdependence of these variables. Further, subsampling introduces the risk of losing statistical power by reducing the sample size, due to exclusion of data that might contain relevant and valuable information. Mediation analysis, on the other hand, uses the full dataset and retains statistical power while modeling the relationships between variables more holistically. However, as we were not satisfied with a purely analytical approach to test this issue, we carried out a new set of experiments in ad-libitum fed mice, where there is no concern of food restriction impairing sleep quality in the presleep session. In these conditions food amount also significantly correlated with, and showed significant mediation of, the SWR power change. Finally, we acknowledge and discuss this point in the Discussion, highlighting that given the known relationship between cortical delta and SWRs, it is challenging to fully disentangle these signals. 

      (2) Relatedly, are the animals spending the same amount of time sleeping in the ad lib vs. food restricted conditions? The amount of time spent sleeping could affect the probability of entering certain stages of sleep and thus affect SWR properties. A recent paper (Giri et al., Nature, 2024) demonstrated that sleep deprivation can alter the magnitude and frequency of SWRs. Could the authors quantify sleep quantity and control for the amount of time spent sleeping by subsampling the data, similar to the suggestion above?

      Following the reviewer’s comment, we have quantified and compared the amount of time spent in NREM sleep in the Pre and Post session pairs in which the animals were food restricted, with 0-1.5 g of chow given between the sleep sessions. We found that there was no significant difference in the amount of time spent in NREM sleep in the Pre and Post sessions. We have added this result to the Results section of the manuscript and as a new Supplementary Fig. 1. 

      Additionally, we have added details to the Methods section that were missing in the original submission that are relevant to this point. Specifically, within the sleep sessions, the ongoing sleep states were scored using the AccuSleep toolbox (https://github.com/zekebarger/AccuSleep) using the EEG and EMG signals. NREM periods were detected based on high EEG delta power and low EMG power, REM periods were detected based on high EEG theta power and low EMG power, and Wake periods were detected based on high EMG power. Importantly, only NREM periods were included for subsequent SWR detection, quantification and analyses (in particular, reported SWR rates reflect the number of SWRs per second of NREM sleep). 

      (3) Plot 5I only reports significance but does not clearly show the underlying quantification of LH GABAergic activity. Upon reading the methods for how this analysis was conducted, it would be informative to see a plot of the pre-SWR and post-SWR integral values used for the paired t-test whose p-values are currently shown. For example, these values could be displayed as individual points overlaid on a pair of boxand-whisker plots of the pre- and post-distribution within the session (perhaps for one example session per mouse with the p-value reported, to supplement a plot of the distribution of p-values across sessions and mice). If these data are non-normal, the authors should also use a non-parametric statistical test.

      We have generated the summary plots the reviewer requested and have now included them in Supplementary Fig. 2. 

      Minor comments:

      (4) A brief explanation (perhaps in the discussion) of what each change in SWR property (magnitude, rate, duration) could indicate in the context of the hypothesis may be helpful in bridging the fields of metabolism and memory. For example, by describing the hypothesized mechanistic consequence of each change, could the authors speculate on why ripple rate may not increase in all the instances where ripple power increases after feeding? Why do the authors speculate that ripple duration does not increase, given that prior work (Fernandez-Ruiz et al. 2019) has shown that prolonged ripples support enhanced memory?

      This is an interesting point and we have added a section to the Discussion to discuss it (pg. 17, last paragraph)

      (5) The authors suggest that "SWRs could modulate peripheral metabolism" as a future implication of their work. However, the lack of clear effects from GLP-1, leptin and insulin complicates this interpretation. It might be informative for readers if the authors expanded their discussion of what specific role they speculate that SWRs could play in regulating metabolism, given these negative results.

      We have added a section to the Discussion proposing potential reasons for this point (pg. 16, last paragraph)

      Recommendations for the authors:  

      Reviewer #1 (Recommendations for the authors):

      Major Comments:

      (1) The experiments involve very precise windows of time for sleeping and eating that seem impossible to control. For example, the authors state that for the experiments in Figure 1, there was a 2-h sleep period, followed by a 1-h feeding period, followed by another 2-h sleep period. Without sleep deprivation procedures or other environmental manipulations, how can these periods be so well-defined? Even during the inactive period, mice typically don't sleep for 2-h bouts at once, and the addition of food would not likely lead to an exact 1-h period of wakefulness in the middle. The validity of these experimental times would be more believable if the authors provided much more data on these sessions. For example, the authors could provide a table or visual display of data for the actual timing of the pre-sleep, eating, and post-sleep phases with exact time measurements and/or visual display of sleep versus wakefulness.

      This is an important point, which we were not clear enough about in the original submission. While the durations of the Pre-sleep, Wake and Post-sleep sessions were indeed 2 h, 1 h and 2 h respectively, the animals did not actually sleep during the entirety of the sleep sessions. Importantly, we performed sleep state scoring on all sessions, and only analyzed identified NREM sleep for all SWR analyses. Following the reviewer’s comment (and that of Reviewer 1), we have quantified and compared the amount of time spent in NREM sleep in the Pre and Post session pairs in which the animals were food restricted and 0-1.5 g of chow were given between the sleep sessions. We found that there was no significant difference in the amount of time spent in NREM sleep in the Pre and Post sessions. We have added this result to the Results section of the manuscript and as a new Supplementary Fig. 1. 

      Additionally, we have added details to the Methods section that were missing in the original submission that are relevant to this point. Specifically, within the sleep sessions, the ongoing sleep states were scored using the AccuSleep toolbox (https://github.com/zekebarger/AccuSleep) using the EEG and EMG signals. NREM periods were detected based on high EEG delta power and low EMG power, REM periods were detected based on high EEG theta power and low EMG power, and Wake periods were detected based on high EMG power. Importantly, only NREM periods were included for subsequent SWR detection, quantification and analyses (in particular, reported SWR rates reflect the number of SWRs per second of NREM sleep). 

      (2) I may have missed this (although I tried searching in the text and figure legend), but the authors did not state the difference between green versus red bar colors in Figure 1 C-E. For Figures 1 F-J, do the individual dots represent both the test (fed) animals and control animals, or just the test animals?

      We thank the reviewer for the opportunity to clarify these points. Red bars in Fig. 1C-E represent the SWR changes observed following delivery of equal or more than 0.5 g of chow, while the green bars represent the changes observed following delivery of less than 0.5 g. Fig. 1F-J includes both the experimental and control animals- the control animals appearing as having received 0 food amount. This information has now been added to the figure legend.

      (3) For the jello experiments in Figure 3, was there only 1 trial per animal? Previous studies show that animals learn the caloric value of jello after subsequent trials, so whether or not multiple trials took place in each animal is important for interpretation of the results.

      In Figure 3, the datapoints within each panel represent different animals and this information has now been added to the figure legend. Nevertheless, the animals were previously habituated to all foods, including regular jello, sugar-free jello and chocolate. While we consider it unlikely that this prior experience was sufficient to underlie the differential effects on SWRs, we cannot fully rule out the possibility that it provided some ability to predict the caloric value and consequences of the different foods. We have added details to the acknowledgement of this point in the Discussion (pg. 17, second paragraph).

      (4) The experiments in Figure 5 are informative but don't relate to the experiments in the rest of the study. It is difficult to interpret their meaning given that these experiments take place over seconds while the other experiments take place over hours. Some attempt should be made to bridge these experiments over the timescales relevant for the behaviors studied in Figures 1-4.

      We have now further acknowledged and discussed the point that our investigation is limited to the timescale of seconds around SWRs, and thus identified a potential communication channel, but whether and how this communication changes across hours following feeding remains for future studies (pg. 18, second paragraph).

      (5) Figure 5B should depict the x-axis in seconds, not an arbitrary set of times from a recording.

      We have replaced these with a time scale bar.

      Minor Comments:

      (6) The writing of the manuscript can be improved in many places:

      Sometimes the writing could be more precise. For example, the Abstract states: "hippocampal sharp wave ripples (SWRs)... have been shown to influence peripheral glucose metabolism." Could this be written in a more informative way, rather than just staying "has been shown to influence?" A few more words would provide a lot more information. Similarly, at the end of the Introduction: "we set out to test the hypothesis that SWRs are modulated following meal times as part of the systems-level response to changing metabolic needs." This is not a strong hypothesis... could it be written to boldly state how the SWRs will be modulated (increase or decrease) and provide more assertive information?

      The writing can be grandiose at times. Phrases such as "life is a continuous journey" or "the hypothalamus is a master regulator of homeostasis" are a bit sophomoric and too colloquial.

      Finally, a representative recording should be referred to as just that-a "representative recording," as opposed to a "snippet," which is also colloquial. This word is used in the figure legends to Figures 1 and 5, and misspelled as "sinpper" in Figure 1

      We have reworded all these sentences and phrases to make them clearer, more concrete and more formal.

      (7) The methods state that the study used both male and female mice. Were they used in equal numbers across experiments?

      Only one female was used in the final dataset, and we have corrected the wording accordingly.

      Reviewer #2 (Recommendations for the authors):

      Great paper!

      Thanks!

      Reviewer #3 (Recommendations for the authors):

      Below are some minor requests for clarification, including in figures:

      (1) Fig. 5H y-axis should say "normalized dF/F."

      Done

      (2) Fig. 1B is missing a y-axis label. It may be clearer to display separate y-axis scale bars for each component (SWR envelope, ripple-filtered amplitude, etc).

      Done

      (3) Please include labels for brain areas and methodological components in Fig. 5A.

      Done

      (4) Should Fig. 5B have the same y-axis or scale bars as 1B?

      We have edited the figure labels and legends to be visually similar

      (5) In Fig. 5J, is the y-axis a count of sessions?

      Yes, we have added that to the y-axis label

      (6) Could the authors please clarify whether the sugar-free jello was sweetened with an artificial sweetener? If so, this is a robust control for the rewarding nature of the two jellos, so a quick clarification would highlight this strength of the experiment.

      We thank the reviewer for this great point. Indeed, the sugar free jello contained artificial sweeteners (Aspartame and Acesulfame Potassium). We have added this information to the Results and Methods.

      (7) It appears in Fig. 5 that there may be a reliable dip in activity **at** the time of SWR onset, followed by the increase afterward, as shown in the example FP trace and the individual ripple-triggered traces. Is this indeed the case, and does this dip fall significantly below baseline? This characterization would be interesting, but I acknowledge is not necessarily crucial to the study to include.

      This would indeed be an interesting finding, but upon examination and statistical testing, we found that this is not the case. We believe this may appear as such due to the normalization of the traces.

      (8) The authors mention a reduction in ripple rate following insulin under food restriction as the only significant effect for insulin, GLP-1, and leptin, yet there was also a significant increase (at p<0.05) in ripple duration for GLP-1 in the ab lib condition. Is this not considered noteworthy?

      This is a fair point and we have reworded the description of this result to simply state that there were no robust, consistent, dose-dependent effects of GLP-1, leptin and insulin on SWR attributes.

    1. eLife Assessment

      Using electrophysiological recordings in freely moving rats, this valuable study investigates the role of gamma oscillations in the development of spatial representations in the hippocampus. Specifically, solid evidence supports the claim that distinct gamma oscillatory inputs contribute to the emergence of 'theta sequences', which encode the animal's ongoing trajectory. This study will be of interest to neuroscientists working in the fields of spatial navigation and neuronal dynamics.

    2. Reviewer #2 (Public review):

      This manuscript addresses an important question which has not yet been solved in the field, what is the contribution of different gamma oscillatory inputs to the development of "theta sequences" in the hippocampal CA1 region. Theta sequences have received much attention due to their proposed roles in encoding short-term behavioral predictions, mediating synaptic plasticity, and guiding flexible decision-making. Gamma oscillations in CA1 offer a readout of different inputs to this region and have been proposed to synchronize neuronal assemblies and modulate spike timing and temporal coding. However, the interactions between these two important phenomena have not been sufficiently investigated. The authors conducted place cell and local field potential (LFP) recordings in the CA1 region of rats running on a circular track. They then analyzed the phase locking of place cell spikes to slow and fast gamma rhythms, the evolution of theta sequences during behavior and the interaction between these two phenomena. They found that place cell with the strongest modulation by fast gamma oscillations were the most important contributors to the early development of theta sequences and that they also displayed a faster form of phase precession within slow gamma cycles nested with theta.

      Comments on revisions:

      Several important shortcomings were noted in the original manuscript. These have been addressed in this revised version with the addition of multiple new analysis, controls and clarifications. The revised manuscript has been significantly improved and its conclusions are adequately supported by the results presented.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Reviewer #1 (Public review):

      This study presents evidence that a special group of place cells, those tuned to fast-gamma oscillations, play a key role in theta sequence development. How theta sequences are formed and developed during experience is an important question, because these sequences have been implicated in several cognitive functions of place cells, including memory-guided spatial navigation. The revised version of this paper has been significantly improved. Major concerns in the previous round of review on technical and conceptual aspects of the relationship between gamma oscillations and theta sequences are addressed. The main conclusion is supported by the data presented.

      Reviewer #2 (Public review):

      The authors have conducted new analysis to address the issues I and the other reviewers raised in our original revision. As a result, the revised manuscript has been substantially improved.

      We thank the two reviewers for their positive comments.

      Recommendations for the authors:

      Reviewer #2 (Recommendations for the authors):

      There are, however, still a few remaining issues that need further clarification.

      - Despite the authors explanation and comparison with Kitanishi et al., 2015, Neuron, I still find that the reduced number of significantly gamma phase-locked cells is at odds with most previous reports (e.g., Csicvari et al., 2003; Colgin et al., 2009; Belluscio et al., 2012; Schomburg et al., 2014; Cabral et al., 2014; Fernandez-Ruiz et al., 2017; Lopes dos Santos et al., 2018). There can be several issues to explain this difference, like the choice of LFP reference channel. The authors should at least acknowledge this difference in the text.

      We thank the reviewer for this suggestion.  We discussed the potential reasons causing the different proportion of gamma phase locked cells in the Discussion (lines367-380).

      - The new Figure R2 is very useful and should be included in the manuscript. It would be even better to expand the frequency range to higher frequencies to show where the maximum peak is. Still, the potential contribution of spike leakage should be acknowledged. While I agree that it will not account for all fast gamma spike modulation, it is certainly a contributing factor. A further evidence of this is that the coupling strength seems to keep increasing towards supra gamma frequency range in Fig R2. This is to be expected given that the authors have used the local LFP from the same tetrode where cells were recorded, which is never a good practice.

      We thank the reviewer for this suggestion. Now the Fig R2 has been moved to the manuscript as a part of Figure 2-figure supplement 2 (lines133-135). In terms of the contribution of spike leakage by using the local LFP, we also detected FG-cells by using LFP from a different tetrode, i.e. the central one of the bundle that located in the cell body layer, and found approximate proportion of FG-cells which phase locked to ~75Hz (Fig R3, now the Figure 2-figure supplement 2C-F). Thus, we think using the local LFP would not affect the main conclusion and we decide to keep the original results. We also acknowledged the potential contribution of spike leakage in the Discussion (lines 372-377).

      - From the authors answer I understand that recordings were almost exclusively conducted from the deep CA1 pyramidal layer. This would preclude any meaningful interpretation of the deep/ superficial differences in the distribution of FG and NFG cells. This is not a crucial point for the paper but needs to be acknowledged.

      We thank the reviewer for this suggestion.  We acknowledged the meaningful interpretation of the deep/ superficial differences in the distribution of FG- and NFG-cells in the Discussion (lines 380-386).

      - I am afraid that the authors interpreted my comment about authorship in the opposite way that I intended. I meant that the usual practice is that the last author of the manuscript is the person who has been the main intellectual driver of the work, not the most senior one necessarily. I guess that is Dr. Zheng not Dr. Ming. However, I leave this decision to the discretion of the authors.

      We thank the reviewer for this rigorous consideration.  Dr. Ming and Dr. Zheng were both the main intellectual drivers of this work.  Therefore, we decide to keep the current authors in the manuscript.

    1. eLife Assessment

      This important study combines genetic analysis, biochemistry, and structural modeling to reveal new insights into how changes in protein-protein structure activate signal transduction as part of the bacterial general stress response. The data, which was collected using validated and standard methods, and its interpretations are convincing; however, to fully meet the title's promise, additional experimental evidence is needed to strengthen the proposed model and its potential application to other systems. This manuscript will be of broad interest to microbiologists, structural biologists, and cell biologists.

    2. Reviewer #1 (Public review):

      Summary:

      This very interesting manuscript proposes a general mechanism for how activating signaling proteins respond to species specific signals arising from a variety of stresses. In brief, the authors propose that the activating signal alters the structure by a universal allosteric mechanism.

      Strengths:

      The unitary mechanism proposed is appealing and testable. The propose that the allosteric module consists of crossed alpha-helical linkers with similar architecture and that their attached regulatory domains connect to phosphatases or other molecules through coiled-coli domains, such that the signal is transduced via rigidifying the alpha helices, permitting downstream enzymatic activity. The authors present genetic and structural prediction data in favor of the model for the system they are studying, and stronger structural data in other systems.

      Weaknesses:

      I thank the authors for making significant revisions that addressed almost all of my concerns. I hope that the authors will consider addressing my last concern, which is that the title is inappropriate. However, I do not believe that this should hold up the publication of the ms.

      "A General Mechanism for Initiating the General Stress Response in Bacteria" is misleading because it suggests a broadly applicable, universal mechanism across all bacterial species, whereas the study primarily focuses on Bacillus subtilis and its RsbU phosphatase activation. While the authors propose that the mechanism may extend to other bacteria, the evidence is largely based on structural modeling rather than direct experimental validation across multiple phyla. Additionally, the phrase "General Stress Response" might imply that the paper broadly explains stress response regulation, but it specifically examines the activation of RsbU by RsbT, which is just one really small part of the broader GSR network. The redundancy in "A General Mechanism for the General Stress Response" could also create an impression of an oversimplified, universal model when stress responses are often species- and context-specific. Furthermore, the study builds upon existing knowledge of partner-switching mechanisms rather than introducing an entirely new concept, making the claim of a general mechanism overstated and misleading for the field.

      Title options could be "A Conserved Activation Mechanism for the General Stress Response Phosphatase in Bacteria", "Coiled-Coil Linker-Mediated Activation of a General Stress Response Phosphatase", all of which more accurately reflect the study's scope and findings.

    3. Reviewer #2 (Public review):

      Summary:

      While bacteria have the ability to induce genes in response to specific stresses, they also use the General Stress Response (GSR) to deal with growth conditions that presumably include a larger range of stresses (for instance, stationary phase growth). The activation of GSR-specific sigma factors is frequently at the heart of the induction of a GSR. Given the range of stresses that can lead to GSR induction, the regulatory inputs are frequently complex. In B. subtilis, the stressosome, a multi-protein complex, contains a set of proteins that, upon appropriate stresses, initiate partner switching cascades that free the sigma B sigma factor from an anti-sigma. The focus here is on the mode of activation of RsbU, a serine/threonine phosphatase of the PPM family, leading to sigB activation. RbsT, a component of the degradosome interacts with RsbU upon stress, activating the phosphatase activity. Once active, RsbU dephosphorylates its target (RsbV, an anti-antisigma), which in turn binds the anti-sigma. The conclusion is that flexible linker domains upstream of the phosphatase domain are the target for activation, resulting in a crossed-linker dimeric structure. The authors then use the information on RsbU to suggest that parallel approaches may be used to activate PPM phosphatases for the GSR response in other bacteria.

      Strengths and Weaknesses:

      (1) A strength of the work is the combination of modeling, genetics and biochemical approaches to support the idea that the flexibility of the linker of the RsbU phosphatase is critical to signalling and that this changes as a result of interactions of the signaling protein RsbT.

      (2) The impact of the work, beyond better understanding of this particular signalling system, lies in the suggested parallels with other GSR system regulators in a range of bacteria. The work here provides fairly clear indications of what mutational changes would be most likely to test the model.

      (3) Assuming that these predictions are shown to be correct in future work, that will leave as an intriguing question why this particular geometry has been conserved in GSR - whether they emerge from a common ancestor (found where?) and/or there is some characteristic (flexibility of modulating the response?) that is particularly important for GSR signal input. Coupled with this will be further understanding of how the linker and/or interacting proteins change in different systems.

    4. Reviewer #3 (Public review):

      Summary:

      The authors present a study building on their previous work on activation of the general stress response phosphatase, RsbU, from Bacillus subtilis. Using computed structural models of the RsbU dimer the authors map previously identified activating mutations onto the structure and suggest further protein variants to test the role of the predicted linker helix and the interaction with RsbT on the activation of the phosphatase activity.

      Using in vivo and in vitro activity assays, the authors demonstrate that linker variants can constitutively activate RsbU and increase the affinity of the protein for RsbT, thus showing a link between the structure of the linker region and RsbT binding.

      Small angle X-ray scattering experiments on RsbU variants alone, and in complex with RsbT show structural changes consistent with a decreased flexibility of the RsbU protein, which are hypothesised to indicate an disorder-order transition in the linker when RsbT binds. This interpretation of the data is consistent with the biochemical data presented by the authors.

      Further computed structure models are presented for other protein phosphates from different bacterial species and the authors propose a model for phosphatase activation by partner binding. They compare this to the activation mechanisms proposed for histidine kinase two-component systems and GGDEF proteins and suggest the individual domains could be swapped to give a toolkit of modular parts for bacterial signalling.

      Strengths:

      The key mutagenesis data is presented with two lines of evidence to demonstrate RsbU activation - in vivo sigma-b activation assays utilising a beta-galactosidase reporter and in vitro activity assays against the RsbV protein, which is the downstream target of RsbU. These data support the hypothesis for RsbT binding to the RsbU linker region as well as the dimerisation domain to activate the RsbU activity.

      Weaknesses:

      Small angle scattering curves are difficult to unambiguously interpret, but the authors present good interpretations that fit with the biochemical data presented. These interpretations should be considered as models for future testing with other methods - hydrogen/deuterium exchange mass spectrometry, would be a good additional method to use, as exchange rates in the linker region would be affected significantly by the disorder/order transition on RsbT binding.

      The interpretation of the computed structure models is provided with a few caveats related to the bias in the models returned by AlphaFold2. For the full-length models of RsbU and other phosphatase proteins, the relationship of the domains to each other is likely to be the least reliable part of the models - this is apparent from the PAE plots shown in supplementary figure 8.

      Comments on revisions:

      The authors have addressed the review comments satisfactorily for this manuscript to stand as a version of record.

    5. Author response:

      The following is the authors’ response to the original reviews

      Public reviews:

      Reviewer #1 (Public review):

      Summary:

      This very interesting manuscript proposes a general mechanism for how activating signaling proteins respond to species-specific signals arising from a variety of stresses. In brief, the authors propose that the activating signal alters the structure by a universal allosteric mechanism.

      Strengths:

      The unitary mechanism proposed is appealing and testable. They propose that the allosteric module consists of crossed alpha-helical linkers with similar architecture and that their attached regulatory domains connect to phosphatases or other molecules through coiled-coli domains, such that the signal is transduced via rigidifying the alpha helices, permitting downstream enzymatic activity. The authors present genetic and structural prediction data in favor of the model for the system they are studying, and stronger structural data in other systems.

      Weaknesses:

      The evidence is indirect - targeted mutations, structural predictions, and biochemical data. Therefore, these important generalizable conclusions are not buttressed by impeccable data, which would require doing actual structures in B. subtilis, confirming experiments in other organisms, and possibly co-evolutionary coupling. In the absence of such data, it is not possible to rule out variant models.

      We thank the reviewer for their feedback. A challenge of studying flexible proteins is that it is often not possible to directly obtain high resolution structural data. For the case of B. subtilis RsbU, the independent experimental approaches we applied (including two unbiased genetic screens, targeted mutagenesis, SAXS, enzymology, and structure prediction, which includes evolutionary coupling) converged upon a model for activation, which we feel is well supported. Frustratingly, our attempts at determining high resolution experimental structures have been unsuccessful, which we think is due to the flexibility of the proteins revealed by our SAXS experiments. For example, we collected X-ray diffraction data from crystals of a fragment of B. subtilis RsbU containing the N-terminal domain and linker in which the linker was almost entirely disordered in the maps. We agree that doing experiments in other organisms would be valuable next steps to test the hypothesis that this coiled-coil based transduction mechanism is conserved across species, and will modify the text to differentiate this more speculative section of the manuscript.

      We have modified the abstract to read:

      “This coiled-coil linker transduction mechanism additionally suggests a resolution to the mystery of how shared sensory domains control serine/threonine phosphatases, diguanylate cyclases and histidine kinases.”

      We have modified the results to read:

      "These predictions suggest a testable hypothesis that RsbP is controlled through an activation mechanism similar to that of RsbU (Fig. 5A)”

      “From this analysis, we speculate that linker-mediated phosphatase domain dimerization is an evolutionarily conserved, adaptable mechanism to control PPM phosphatase activity.”

      Based on this critique (and the critiques of the other reviewers), we plan to do energetic analysis of the predicted coiled coils from the enzymes we analyzed from other species and to incorporate this into the manuscript.

      We have modified the results to read:

      Consistent with a model in which the stability of the linker plays a conserved regulatory role, the AlphaFold2 models for many of the predicted structures have unfavorable polar residues buried in the coiled-coil interface (positions a and d, for which non-polar residues are most favorable) (Figure 5 – figure supplement 2).”

      Finally, in the manuscript, we have highlighted that this mechanism is not the only mechanism for activation of other proteins with effector domains connected to linkers, but rather one of many mechanisms (Fig 5G). The reviewer additionally made helpful suggestions about the text in detailed comments that we will incorporate as appropriate.

      Reviewer #2 (Public review):

      Summary:

      While bacteria have the ability to induce genes in response to specific stresses, they also use the General Stress Response (GSR) to deal with growth conditions that presumably include a larger range of stresses (for instance, stationary phase growth). The activation of GSR-specific sigma factors is frequently at the heart of the induction of a GSR. Given the range of stresses that can lead to GSR induction, the regulatory inputs are frequently complex. In B. subtilis, the stressosome, a multi-protein complex, contains a set of proteins that, upon appropriate stresses, initiate partner switching cascades that free the sigma B sigma factor from an anti-sigma. The focus here is on the mode of activation of RsbU, a serine/threonine phosphatase of the PPM family, leading to sigB activation. RbsT, a component of the degradosome interacts with RsbU upon stress, activating the phosphatase activity. Once active, RsbU dephosphorylates its target (RsbV, an anti-antisigma), which in turn binds the anti-sigma. The conclusion is that flexible linker domains upstream of the phosphatase domain are the target for activation, via binding of proteins to the N-terminal domain, resulting in a crossed-linker dimeric structure. The authors then use the information on RsbU to suggest that parallel approaches are used to activate PPM phosphatases for the GSR response in other bacteria. (Biology vs. Mechanism, evolution?)

      Strengths and Weaknesses:

      Many of these have to do with clarifying what was done and why. This includes the presentation and content of the figures.

      One issue relates to the background and context. A bit more information on the stresses that release RsbT would be useful here. The authors might also consider a figure showing the major conclusions and parallels for SpoIIE activation and possibly other partner switches that are discussed, introducing the switch change more clearly to set the stage for the work here (and the generalization). There are a lot of players to keep track of.

      We plan to carefully review the manuscript to improve the clarity of presentation and background. In particular, we thank the reviewer for pointing out the missing information about the release of RsbT from the stressosome. We will incorporate this information into the introduction and provide an additional figure.

      We have added the following text to the introduction:

      “RsbT is sequestered in a megadalton stress sensing complex called the stressosome, and is released to bind RsbU in response to specific stress signals including ethanol, heat, acid, salt, and blue light”

      We have added a new figure panel (2C) that shows the model for how Q94L, M166V, and RsbT binding induce conformational change of the PPM domain to recruit metal cofactor and activate RsbU (analogous, but slightly different from the mechanism for SpoIIE).

      The reviewer additionally provided detailed helpful comments that we will incorporate in the text and figures.

      Reviewer #3 (Public review):

      Summary:

      The authors present a study building on their previous work on activation of the general stress response phosphatase, RsbU, from Bacillus subtilis. Using computed structural models of the RsbU dimer the authors map previously identified activating mutations onto the structure and suggest further protein variants to test the role of the predicted linker helix and the interaction with RsbT on the activation of the phosphatase activity.

      Using in vivo and in vitro activity assays, the authors demonstrate that linker variants can constitutively activate RsbU and increase the affinity of the protein for RsbT, thus showing a link between the structure of the linker region and RsbT binding.

      Small angle X-ray scattering experiments on RsbU variants alone, and in complex with RsbT show structural changes consistent with a decreased flexibility of the RsbU protein, which is hypothesised to indicate a disorder-order transition in the linker when RsbT binds. This interpretation of the data is consistent with the biochemical data presented by the authors.

      Further computed structure models are presented for other protein phosphates from different bacterial species and the authors propose a model for phosphatase activation by partner binding. They compare this to the activation mechanisms proposed for histidine kinase two-component systems and GGDEF proteins and suggest the individual domains could be swapped to give a toolkit of modular parts for bacterial signalling.

      Strengths:

      The key mutagenesis data is presented with two lines of evidence to demonstrate RsbU activation - in vivo sigma-b activation assays utilising a beta-galactosidase reporter and in vitro activity assays against the RsbV protein, which is the downstream target of RsbU. These data support the hypothesis for RsbT binding to the RsbU linker region as well as the dimerisation domain to activate the RsbU activity.

      Weaknesses:

      Small angle scattering curves are difficult to unambiguously interpret, but the authors present reasonable interpretations that fit with the biochemical data presented. These interpretations should be considered as good models for future testing with other methods - hydrogen/deuterium exchange mass spectrometry, would be a good additional method to use, as exchange rates in the linker region would be affected significantly by the disorder/order transition on RsbT binding.

      We agree with the reviewer that the SAXS data has inherent ambiguity due to the nature of the measurement. However, SAXS is one of the best techniques to directly assess conformational flexibility. Our scattering data for RsbU have multiple signatures of flexibility supporting a high confidence conclusion. While the scattering data support a reduction in flexibility for the RsbT/RsbU complex, we agree that a high resolution structure would be valuable. However the combination of the scattering data with our biochemical and genetic data supports the validity of the AlphaFold predicted model. We thank the reviewer for the suggestion of future hydrogen/deuterium exchange experiments that would be complementary, but which we feel are beyond the scope of this work.

      The interpretation of the computed structure models should be toned down with the addition of a few caveats related to the bias in the models returned by AlphaFold2. For the full-length models of RsbU and other phosphatase proteins, the relationship of the domains to each other is likely to be the least reliable part of the models - this is apparent from the PAE plots shown in Supplementary Figure 8. Furthermore, the authors should show models coloured by pLDDT scores in an additional supplementary figure to help the reader interpret the confidence level of the predicted structures.

      We thank the reviewer for suggestions on how to clarify the discussion of AlphaFold models. We will decrease the emphasis on the computed models in the text and will add figures with the models colored by the pLDDT scores to aid in the interpretation.

      We have modified the text of the Abstract: “This coiled-coil linker transduction mechanism additionally suggests a resolution to the mystery of how shared sensory domains control serine/threonine phosphatases, diguanylate cyclases and histidine kinases.”

      We have modified the text of the Results: “These predictions suggest a testable hypothesis that RsbP is controlled through an activation mechanism similar to that of RsbU (Fig. 5A).”

      “From this analysis, we speculate that linker-mediated phosphatase domain dimerization is an evolutionarily conserved, adaptable mechanism to control PPM phosphatase activity”

      We have also added Figure 1 – figure supplement 2 with the AlphaFold2 models colored by the pLDDT scores.

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      Baral and colleagues investigate the regulatory mechanisms of the General Stress Response (GSR) in Bacillus subtilis, focusing on the phosphatase RsbU and its regulation by the protein RsbT. The GSR is a critical adaptive mechanism that allows bacteria to survive under various stress conditions by reshaping their physiology through a broad transcriptional response. RsbU, a key player in the GSR, facilitates the activation of the transcription factor SigB by dephosphorylating RsbV. This activation is mediated through a partner-switching mechanism involving RsbT. Baral and colleagues use a combination of genetic screening, structural predictions via AlphaFold2, and biophysical techniques such as SAXS and MALS to present a model for how RsbT regulates RsbU. Key findings include the identification of specific amino acid substitutions that enhance RsbU activity, the role of the α-helical linker in RsbU dimerization and activation, and the potential broader conservation of these mechanisms across bacterial species. However, as described below, additional work is required to solidify the results.

      Major Points

      (1) The manuscript is misnamed--it dissects a single step of the signal-transduction pathway regulating the general stress response. Instead, it is rather seeking a generalizable mechanism for kinase -phosphatase interactions across stresses.

      We have edited the title to “A General Mechanism for Initiating the General Stress Response in Bacteria” to reflect that that this study addresses the initiating event of the general stress response.

      (2) The genetic screen likely has limitations in detecting all possible variants that could affect RsbU activity. The readout is specific to σ^B activation, and the focus on specific amino acid substitutions may overlook other significant regions or mechanisms involved in the regulation of RsbU, particularly those involving RsbV and RsbT.

      Our screens were specifically designed to identify features of RsbU that contribute to regulation. Importantly, RsbU does not have any known targets other than RsbV and the downstream σ<sup>B</sup> response but agree that substitutions in either RsbV or RsbT could influence RsbU activation. In principle our suppressor screen with RsbU<sup>Y28I</sup> could have identified RsbT variants (rsbT was mutagenized in this screen), but we did not identify any such variants in the screen. We conducted a separate screen (published elsewhere) that specifically addressed how RsbU recognizes RsbV.

      (3) The authors largely focus on the biochemical and structural aspects of RsbU regulation. There is limited discussion on the broader functional implications of these findings in the context of bacterial physiology and stress response. Incorporating more in vivo studies to show how these mechanisms impact bacterial survival and adaptation would provide a more comprehensive understanding.

      We appreciate this comment, but did not conduct additional studies of survival and adaptation because the phenotypes of σ<sup>B</sup> deletion in B. subtilis under laboratory conditions are relatively mild and therefore difficult to assay. Future studies to address this in other systems could be highly informative.

      (4) The results primarily support the model of linker-mediated dimerization and rigidity. However, other potential regulatory mechanisms or interacting partners might also play significant roles in RsbU activation. A more thorough exploration of these possibilities would strengthen the study's conclusions.

      One of the major advantages of RsbU as a model for initiation of the general stress response is that the system is discreet with all evidence pointing to there being a single primary input (RsbT) and output (dephosphorylation of RsbV). While there are other possible variations on the system (for example RsbU may be directly activated by manganese stress), we focused on this system precisely because of its simplicity.

      (5) While the study presents evidence for the conservation of the described mechanism across different species, this assumption is based on structural predictions and limited experimental data. Broader experimental validation across diverse bacterial species would be necessary to substantiate this claim. Coevolution coupling along with conservation/evolutionary studies could be considered.

      We have altered the language in the paper to emphasize where we are making inferences from predictions that are therefore more speculative. We agree that a more detailed analysis of the evolutionary coupling would likely be fruitful. We note that these couplings are the major driving force of AlphaFold predictions, suggesting that these couplings contributed to the models that we analyzed.

      (6) The reliance on AlphaFold2 for structural predictions introduces potential biases and uncertainties inherent in computational models. Experimental validation of these models through additional techniques such as cryo-EM or X-ray crystallography would strengthen the conclusions.

      We agree with this point, which is why we performed extensive analysis and validation of the models for RsbU using SAXS, genetics, and biochemistry. The proposed techniques are made more challenging by flexibility and heterogeneity, which we detected in our experiments. Our attempts thus far at experimental structure determination are consistent with this being a major technical hurdle.

      (7) SAXS data provide low-resolution structural information, and the interpretation of flexibility versus rigidification might be overemphasized in its interpretation. This part of the study was difficult to interpret. Improving readability by breaking down the text into sections with clear headings for each figure panel and clarifying descriptions of the panels and methods would help. Complementary high-resolution techniques could provide a more definitive view of the linker's conformational changes.

      We have modified the presentation of the figures to clarify the SAXS analysis. The fact that the SAXS analysis suggests flexibility rather than a discrete inactive conformation means that high-resolution techniques may not be appropriate for this system.

      (8) The study primarily focuses on the model where RsbT binding rigidifies the RsbU linker. Alternative hypotheses, such as subtle conformational adjustments without complete rigidification, are not extensively explored or ruled out.

      Our analysis of the SAXS data strongly suggests that a subtle conformational change could not account for the scattering data that we obtained. We have modified the text to clarify this point.

      “Indicative of significant deviation between the RsbU structure in solution to the AlphaFold2 model, the scattering intensity profile (I(q) vs. q) was a poor fit (χ<sup>2</sup> 12.53) to a profile calculated from the AlphaFold2 model of an RsbU dimer using FoXS (Schneidman-Duhovny et al. 2016; Schneidman-Duhovny et al. 2013) (Fig. 4A). We therefore assessed the SAXS data for the RsbU dimer for features that report on flexibility (Kikhney & Svergun 2015). First, the scattering intensity data lacked distinct features caused by the multi-domain structure of RsbU from the AlphaFold2 model (Fig.4A).”

      (9) Future studies should aim to validate the AlphaFold2 predictions with high-resolution structural techniques. This would provide definitive evidence for the proposed conformational states of RsbU with and without RsbT.

      The fact that the SAXS analysis suggests flexibility rather than a discrete inactive conformation means that high-resolution techniques may not be appropriate for this system.

      (10) Investigating the RsbU-RsbT interaction in vivo using techniques like FRET, co-immunoprecipitation, or live-cell imaging would provide a more comprehensive understanding of their functional dynamics in a cellular context.

      We appreciate the reviewer’s suggestions for future experiments.

      (11) Exploring and testing alternative models of RsbU activation, such as partial rigidification or different modes of conformational change, would strengthen the conclusions.

      While our data strongly support that a flexible-to-rigid transition controls RsbU activation, we agree that it is possible that other mechanisms of linker modification could control other phosphatases and we discuss this at some length in the discussion.

      (12) The figure legends are quite dense and could benefit from some streamlining.

      We have edited the figure legends for clarity and length.

      Reviewer #2 (Recommendations for the authors):

      (1) Activation assays (Figures 1, 3, S2) are presented here as blue or white spots (reflecting a reporter activity). While off and on these are fairly clear, it is more difficult to compare the degree of activity (for instance that rsbU<sup>Q94L</sup> is more active than M166V). It would also be good to clearly present in the text the logic of asking if the mutant is RsbT independent or not (and the interpretation of that). Quantitative assays of these would be very useful.

      We chose not to perform quantitative-LacZ assays here because of several complications to interpreting these results that we encountered in our previously published study (Ho and Bradshaw, 2021). However, the level of blue pigmentation shown in Figure 1B for RsbU Q94L and RsbU M166V is qualitatively different, making the comparison possible. Most importantly, we observed cell density dependent changes in LacZ activity in the absence of rsbT for rsbU<sup>M166V</sup> expressing cells, meaning that comparisons between strains would be difficult. Additionally, we found that it was important to make a chromosomal replacement of rsbU to see the full effect of the M166V substitution. However, we were not able to construct a similar rsbU<sup>Q94L</sup> strain, likely because the high level σ<sup>B</sup> activity is lethal (we were able to construct this strain when σ<sup>B</sup> was deleted but only obtained strains with additional loss-of-function mutations in RsbU when σ<sup>B</sup> was present.

      We have modified the text to explain the logic of identifying RsbT independent variants: “We previously conducted a genetic screen (Ho & Bradshaw 2021) to identify features of RsbU that are important for phosphatase regulation by isolating gain-of-function variants that are active in the absence of RsbT.”

      (2) Explain Figure S8 graphs: as much as Alphafold is now in use, the authors should provide some further explanation of what is shown here. Blue (low error) is good, presumably. What are the A, B, C, and D sections showing? Different parts of a given letter region (and between them)? What is the x-axis? Is the top-ranked model used in every case in the text? How different are these models? The Methods section could be used for some of this (but doesn't in its current form). This also becomes important for the models generated later in the paper (Figure S7), which look rather different here.

      We have modified figure S8 to include additional labels and have added structures with the pLDDT scores shown. We have additionally modified the figure legends and methods to provide the requested information.

      (3) Figure 1C, D, Figure S2: amino acid ends of linker domains could be shown (text discusses 83-97 the linker as a two-turn coiled coil; Q94 is pretty close to the end of this coiled-coil? Figure S2 is even less clear - addresses of other amino acids would help, and or an added sequence showing the full linker and coiled-coil region). Some explanation for positions for readers to focus on for full coiled-coil would be useful in the legend of Figure S2. How strong a coiled-coil prediction is there for this region?

      We have added the sequence of the coiled-coil regions to the figures with numbering. For these analyses we used the Socket2 program, which analyzes a PDB file to identify coiled-coil regions and thus does not provide a confidence score. However, inspection of the sequence and the confidence scores of the AlphaFold2 models indicates that the coiled-coil regions are not ideal, consistent with this being a regulatory feature.

      Is it clear that the fully inactive proteins are still properly folded and soluble?

      In the case of RsbU, our biophysical analysis indicates that the inactive form of the protein is soluble. While phosphatase activity is substantially reduced, our unpublished comparison of single- and multiple-turnover reactions in the absence of RsbT indicates that nearly all of the enzyme is active.

      Finally, are there other positions that would also be expected, from this model, to stabilize the coiled-coil and thus bypass the requirement for RsbT? If so, it would be good to test these. Is it the burial of amino acid at position 94 that is important, or the ability to form crossed helices?

      Because of how short the predicted coiled-coil region is, we did not identify any obvious positions that would likely have the same effect as Q94 substitution. We considered making helix-breaking mutations, which would be predicted to block RsbU activation, but favored analysis of the wildtype protein because of limitations in interpreting the effects of loss-of-function mutations.

      (4) Figure 2A, RsbT binding to RsbU: It was not entirely clear to this reviewer why one would expect the RsbT binding, not needed for activation, to be increased by the mutation that stabilizes the crossed alpha helices. The change is impressive but doesn't the lack of a need for RsbT suggest that this mutation bypasses the normal mechanism? (Is dimerization enuf? Or other protein cross helices?).

      We have modified the text to clarify this point: “One prediction of our hypothesis that RsbT stabilizes the crossed alpha helices of the RsbU dimer, is that RsbT should bind more tightly to rsbU<sup>Q94L</sup> than to RsbU because the coiled-coil conformation that RsbT binds would be more energetically favorable.” Another way of putting this is that if the Q94L substitution activates RsbU through an on-pathway mechanism, RsbT must bind more tightly.

      (5) Figure 3A, Figure S3: Please label the yellow (interface) residues in RsbU and RsbT in Fig. S3 and the green (suppressor) spheres in Figure 3A.

      We have added labels to the figures as suggested.

      If RbsT interacts with the N-terminal dimerization domain and linker, why were residues 174 and 178 (from PPM domain) shown to be implicated in binding?

      The fact that residues in the switch region suppress a mutation that decreases RsbT binding suggests that this region is part of an allosteric network that links RsbT binding, the linker, and dimerization of the phosphatase domains. For example, any substitution that promotes a conformation of the phosphatase domain that is more favorable for dimerization would also promote RsbT binding. However, the precise details of how each mutation fits into this network is not clear and we have therefore chosen to not specify a particular model to avoid over interpreting our data.

      Are these marked in Figure S3?

      We have added labels to make this clear.

      Are these part of a dimerization interface in the C-terminal domain? Are any/all of these RsbU mutants suppressed by Q94L, as one might predict (apparently Y28I is since Q94L was again identified)?

      We chose to focus on Y28I because it was the best studied previously, but we would predict that Q94L would suppress other RsbT binding mutations.

      (6) Line 191-192: Is it surprising that no suppressors were isolated in RsbT?

      We didn’t have a preconception of whether or not it would be possible to identify similar suppressors in RsbT. Explanations for why we did not identify such suppressors could include that RsbT may be destabilized more easily by substitution, that RsbT is more constrained because it has other interaction partners, or that the particular substitutions that would suppress Y28I are less common by the PCR mutagenesis strategy we used.

      (7) Figure 3: Would the same mutants arise if the screen had been done in the absence of RsbT? Was RsbT-dependent tested for the rsbU alleles?

      Our prediction is that we would not have identified any of these mutations except for Q94L in the absence of rsbT. We tested a few of the alleles and found them all to be rsbT dependent, but did not systematically test all of the alleles and therefore did not include this analysis in the manuscript.

      Given the findings earlier in the paper for Q94L, suggesting that this stabilizes the coiled-coil and shows some activity in the absence of RsbT, it seems that the interpretation of other mutants in this region (and Q94L itself) as evidence that RsbT contacts the linker directly and that contact is necessary for activation may be an overinterpretation. If these are in fact RsbT independent, they support the importance of the linker (do they further stabilize coiled-coil formation?), rather than the role of RsbT here. Are G92 and T89 on the outside of the coiled-coil? If Q94 is buried, is it qualitatively different from these others?

      G92 and T89 are predicted to be exposed. The fact that these mutations are near Q94 is part of the reason that we focused on R91 and the predicted contact with D92 of RsbT as another approach to validate the predicted interface.

      (8) Figure 3C addresses the issue of direct interaction of RsbT with the RsbU linker to some extent, given that RsbU R91E doesn't appear to have a lot of activity without RsbT. It would be helped by telling the reader what the R91 contact is initially.

      We have modified the text to clarify this point: “To test the model that RsbT activates RsbU by directly interacting with the linker to dimerize the RsbU phosphatase domains, we introduced a charge swap at position R91 that would abolish a predicted salt-bridge with RsbT D92 (Fig. 3C).”

      (9) Figure 4 and the discussion of it in the text is not likely to be easily understandable for many readers. Aside from providing a bit more explanation of what these analyses are showing, it would be useful to start the whole section (or maybe even much earlier in the paper) with the information found on lines 261-264, that other studies show that the N-terminus dimerizes stably on its own (and is it known that the C-terminus does not?). Then the discussion of the alternative models early in this section would be clearer.

      We have updated the introduction to emphasize this point “RsbU has an N-terminal four-helix bundle domain that dimerizes RsbU and is also the binding site for RsbT, which activates RsbU as a phosphatase (Fig. 1C,D) (Delumeau et al. 2004).”

      We have also added clarification to the model presented at the beginning of this section: “A second possibility is that inactive RsbU is dimerized by the N-terminal domains but that the linkers of inactive RsbU are flexible and that the phosphatase domains only interact with each other when RsbT orders the linkers into a crossing conformation.”

      Is the dimerization of the N-terminal domains previously determined similar/the same as what is seen in the AlphaFold models used here (or the AlphaFold dimerization derived primarily from that data?).

      Yes, the dimerization in the AlphaFold models matches closely to the published structure.

      (10) Discussion and Figure 5: The final part of this work predicts AlphaFold models for a set of other phosphatases involved in initiating GSR across bacterial species, and suggests that linked-mediated phosphatase dimerization is the critical factor to activate the phosphatase. Clearly, this is the most speculative but interesting aspect of the paper. A number of possible questions are suggested by some of this:

      a. Do any of the activating mutants In RsbU and RsbP in the PPM domain (that apparently improve dimerization and thus activation) do a similar job in the other modeled proteins?

      This is an interesting question, but unfortunately most of these proteins have not been biochemically characterized. We highlight examples of RsbP and E. coli RssB for which similar activating mutations have been characterized.

      b. The legend (Figure 5G) suggests that all of the linker combinations will be coiled-coils, but that they will undergo different types of activating (and dimerizing?) transitions. Is that in fact what is being proposed here?

      Yes, this is our working hypothesis.

      c. If there is no dimerization (as noted, only weak dimerization has been reported for E. coli RssB), does that generalize the model to there are linkers and their structures are important? At the least, would the folding up of the E. coli RssB linker with antiadaptor binding be considered another mode of signal transduction or rather some sort of storage form?

      Interestingly, the P. aeruginosa RssB constitutively dimerizes, suggesting the E. coli is the outlier.

      d. Would the "toolkit" model, in which different changes occur in the linker regions, suggest that the interacting proteins are going to be critical for the type of linker changes that will be important? Or something about the nature of the linkers themselves?

      This is an interesting question that we cannot yet answer. We have chosen to focus on the possible flexibility of this mechanism and anticipate that a variety of mechanisms will be used.

      e. Given the extensive comparison to E. coli RssB, the authors might consider a figure to clarify the relative domain architecture, sequences that are akin to switch regions, and others important to the discussion here.

      We tried to highlight this in Figure 5C including coloring the regions similar to the switch regions.

      Reviewer #3 (Recommendations for the authors):

      Given the caveats noted above related to the reliability of computed structure models, I would recommend the authors make the following additions/modifications to their manuscript:

      (1) The authors should show alpha fold models coloured by pLDDT scores in an additional supplementary figure to help the reader interpret the confidence level of the predicted structures.

      We have added these models to figure 1 – figure supplement 2.

      (2) Because of the points mentioned above the authors should tone down the generalisation relating to the activation mechanism of this family of phosphatases presented in the discussion.

      We have modified the paper throughout to emphasize where we are speculating.

    1. eLife Assessment

      This is a saturation mutagenesis screening of CDKN2A gene, successfully assessing the functionality of the missense variants. The work is solid and well-prosecuted. The manuscript was improved during the revision process and this work will serve as a valuable resource for diagnostic labs as well as cancer geneticists.

    2. Reviewer #1 (Public review):

      Summary:

      Kimura et al performed a saturation mutagenesis study of CDKN2A to assess functionality of all possible missense variants and compare them to previously identified pathogenic variants. They also compared their assay result with those from in silico predictors.

      Strengths:

      CDKN2A is an important gene that modulate cell cycle and apoptosis, therefore it is critical to accurately assess functionality of missense variants. Overall, the paper reads well and touches upon major discoveries in a logical manner.

      Weaknesses:

      The paper lacks proper details for experiments and basic data, leaving the results less convincing. Analyses are superficial and does not provide variant-level resolution.

      Comments on revisions

      The manuscript was improved during the revision process.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Public Reviews:

      Reviewer #1:

      Summary:

      Kimura et al performed a saturation mutagenesis study of CDKN2A to assess functionality of all possible missense variants and compare them to previously identified pathogenic variants. They also compared their assay result with those from in silico predictors.

      Strengths:

      CDKN2A is an important gene that modulate cell cycle and apoptosis; therefore it is critical to accurately assess functionality of missense variants. Overall, the paper reads well and touches upon major discoveries in a logical manner.

      Weaknesses:

      The paper lacks proper details for experiments and basic data, leaving the results less convincing. Analyses are superficial and does not provide variant-level resolution. Many of which were addressed during the revision process.

      Comments on revisions:

      The manuscript was improved during the revision process.

      We thank the reviewer for their comments. We are grateful for the opportunity to provide additional information and data to clarify our approach and study results.

      Reviewer #2:

      Summary:

      This study describes a deep mutational scan across CDKN2A using suppression of cell proliferation in pancreatic adenocarcinoma cells as a readout for CDKN2A function. The results are also compared to in silico variant predictors currently utilized by the current diagnostic frameworks to gauge these predictors' performance. The authors also functionally classify CDKN2A somatic mutations in cancers across different tissues.

      Review:

      The goal of this paper was to perform functional classification of missense mutations in CDKN2A in order to generate a resource to aid in clinical interpretation of CDKN2A genetic variants identified in clinical sequencing. In our initial review, we concluded that this paper was difficult to review because there was a lack of primary data and experimental detail. The authors have significantly improved the clarity, methodological detail and data exposition in this revision, facilitating a fuller scientific review. Based on the data provided we do not think the functional characterization of CDKN2A variants is robust or complete enough to meet the stated goal of aiding clinical variant interpretation. We think the underlying assay could be used for this purpose but different experimental design choices and more replication would be required for these data to be useful. Alternatively, the authors could also focus on novel CDKN2A variants as there seems to be potential gain of function mutations that are simply lumped into "neutral" that may have important biological implications.

      Major concerns:

      Low experimental concordance. The p-value scatter plot (Figure 2 Figure Supplement 3A) across 560 variants shows low collinearity indicating poor replicability. These data should be shown in log2fold changes, but even after model fitting with the gamma GLM still show low concordance which casts strong doubt on the function scores.

      Concordance among non-significant p-values is generally low because most of the signal comes from random variability across repeats. If the observed log2 fold change between the repeats is entirely due to noise, one would expect two repeated p-values to behave like independent random uniforms. True concordance is typically more evident in significant p-values because they reflect consistent effects above random noise. Functionally deleterious variants are called when their associated p-value is significant. To confirm this statement, a scatter plot with the log2 normalized fold change was added in Figure 2 Supplement 3C. We see low concordance between repeats in the log2 normalized fold changes centered around 0, corresponding to log log2 normalized changes mainly due to noise. The concordance increases as the variants become significant. One can notice that the correlation coefficient between duplicate assay results was almost identical between the model-based p-values and log2normalized fold change (Figure 2-figure supplement 3A and 3C, Appendix 1-table 4, and Appendix 1-table 6). Also, importantly, no variant was functionally deleterious in one replicate and functionally neutral in another, implying a perfect concordance in calls if we exclude variants that were called indeterminate in one of the two repeats. Finally, of variants with discordant classifications, only 6/560 repeats (1.1%) were functionally deleterious (significant p-value) in one replicate and of indeterminate function in another. We have updated the text as follows:

      “Of variants with discordant classifications, 6 (1.1%) were functionally deleterious in one replicate and of indeterminate function in another. While 102 variants (18.2%) were functionally neutral in one replicate and of indeterminate function in another. Importantly, no variant that was functionally deleterious in one replicate and functionally neutral in another (Appendix 1 -table 4). Furthermore, the correlation coefficient between duplicate assay results was similar using the gamma GLM and log2 normalized fold change (Figure 2-figure supplement 3A and 3C).”

      The more detailed methods provided indicate that the growth suppression experiment is done in 156 pools with each pool consisting of the 20 variants corresponding to one of the 156 aa positions in CKDN2A. There are several serious problems with this design.

      Batch effects in each of the pools preventing comparison across different residues. We think this is a serious design flaw and not standard for how these deep mutational scans are done. The standard would be to combine all 156 pools in a single experiment. Given the sequencing strategy of dividing up CDKN2A into 3 segments, the 156 pools could easily have been collapsed into 3 (1 to 53, 54 to 110, 111 to 156). This would significantly minimize variation in handling between variants at each residue and would be more manageable for performance of further replicates of the screen for reproducibility purposes. The huge variation in confluency time 16-40 days for each pool suggest that this batch effect is a strong source of variation in the experiment.

      While there is variation in time to confluency between different amino acid residues, we do not anticipate this batch effect to significantly affect variant classifications in our study. For example, our results were generally consistent with previous classifications. All synonymous variants (one per residue) and benchmark benign variants assayed were classified as functionally neutral. Furthermore, of benchmark pathogenic variants assayed, none were classified as functionally neutral. 84% were classified as functionally deleterious and 16 percent were classified as indeterminate function.

      Lack of experimental/biological replication: The functional assay was only performed once on all 156 CDKN2A residues and was repeated for only 28 out of 156 residues, with only ~80% concordance in functional classification between the first and second screens. This is not sufficiently robust for variant interpretation. Why was the experiment not performed more than once for most aa sites?

      In our study we determined functional classifications for all CDKN2A missense variants while assessing variability with replicates across 28 residues. Of these variants, only 6 (1.1%) were functionally deleterious in one replicate and of indeterminate function in another. Furthermore, no variant was functionally deleterious in one replicate and functionally neutral in another (Appendix 1 -table 4).  As noted above, we provided additional context in the manuscript.

      For the screen, the methods section states that PANC-1 cells were infected at MOI=1 while the standard is an MOI of 0.3-0.5 to minimize multiple variants integrating into a single cell. At an MOI =1 under a Poisson process which captures viral integration, ~25% of cells would have more than 1 lentiviral integrant. So in 25% of the cells the effect of a variant would be confounded by one or more other variants adding noise to the assay.

      As noted previously, we are not able to differentiate effects due to multiple viral integrations per cells. However, we do not anticipate multiple viral integrations to significantly affect variant classifications in our study as our results are consistent with previous classifications, as described above.

      While the authors provide more explanation of the gamma GLM, we strongly advise that the heatmap and replicate correlations be shown with the log2 fold changes rather than the fit output of the p-values.

      Thank you for the suggestion. As noted, we provide additional explanation in the manuscript about why we classified variants using a gamma GLM. Using a gamma GLM, classification thresholds were determined using the change in representation of 20 non-functional barcodes in a pool of PANC-1 cells stably expressing CDKN2A after a period of in vitro proliferation. Our variant classifications were therefore not based on assay outputs for previously reported – benchmark – pathogenic or begin variants to determine thresholds. We strongly prefer using p-values and classifications using the gamma GLM in the manuscript. However, comparison of assay outputs using a gamma GLM and log2 fold change are included in the manuscript. Read counts, log2 fold change, and classifications based on log2 fold change are presented in the manuscript, for all variants. Readers who wish to use these data may do so and we refer them to the manuscript text, Appendix 1 -table 4, Appendix 1 -table 6, and Figure 2 -figure supplement 2.

      In this study, the authors only classify variants into the categories "neutral", "indeterminate", or "deleterious" but they do not address CDKN2A gain-of-function variants that may lead to decreased proliferation. For example, there is no discussion on variants at residue 104, whose proliferation values mostly consist of higher magnitude negative log2fold change values. These variants are defined as neutral but from the one replicate of the experiment performed, they appear to be potential gain-of-function variants.

      We have added a comment to the discussion to highlight that we did not identify potential gain-of-function variants. Specifically:

      “We classified CDKN2A missense variants using a gamma GLM, as either functionally deleterious, indeterminate functional or functionally neutral. However, we did not classify variants that may have gain-of-function effects, resulting in decreased representation in the cell pool. Future studies are necessary to determine the prevalence and significance of CDKN2A gain-of-function variants.”

      Minor concerns:

      The differentiation between variants of "neutral" and "indeterminate" function seems unnecessary and it seems like there are too many variants that fall into the "indeterminate" category. The authors seem to have set numerical thresholds for CDKN2A function using benchmark variants of known function. While the benchmark variants are important as a frame of reference for the "dynamic range" of the assay, their function scores should not necessarily be used to define hard cutoffs of whether a variant's function score can be interpreted.

      We did not utilize benchmark variants to define thresholds for functional classifications using a gamma GLM. This is one of the strengths of using a gamma GLM model for classification. As explained in our manuscript, classification thresholds were determined using the change in representation of 20 non-functional barcodes in a pool of PANC-1 cells stably expressing CDKN2A after a period of in vitro proliferation. Our variant classifications were therefore not based on assay outputs for previously reported – benchmark – pathogenic or begin variants. While not required when using a gamma GLM, we included indeterminate classifications, which are not uncommon.

      Figure 2 supplement 2 - on the x-axis, should "intermediate" be "indeterminate"?

      This, and a similar typographical error in Figure 2 -figure supplement 3, has been corrected.

    1. eLife Assessment

      This study is a valuable observation that deals with the toxic effects of an intermediary in lipid degradation [trans-2-hexadecenal (t-2-hex)] in yeast through modification of mitochondrial protein import via the TOM complex. We find that the claim that the TOM complex is a main target of t-2-hex are supported by solid evidence, however the molecular mechanism remains unclear allowing multiple interpretation. Despite the shortcomings, this study is inspiring for researchers from the organellar, protein trafficking and lipid field and serves as a starting point to further precise and mechanistic analyses of the phenomenon.

    2. Reviewer #3 (Public review):

      Summary:

      The authors investigate the effect of high concentrations of the lipid aldehyde trans-2-hexadecenal (t-2-hex) in a yeast deletion strain lacking the detoxification enzyme. Transcriptomic analyses as global read out reveals that a large range of cellular functions across all compartments are affected (transcriptomic changes affect 1/3 of all genes). The authors provide additional analyses, which indicate that mitochondrial protein import is affected.

      Strengths:

      Global analyses (transcriptomic and functional genomics approach) to obtain an overview of changes upon yeast treatment with high doses of t-2-hex.

      Weaknesses:

      The use of high concentrations of t-2-hex in combination with a deletion of the detoxifying enzyme Hfd1 limits the possibility to identify physiological relevant changes. For the follow-up analysis, the authors focus on mitochondrial proteins and describe an impairment of mitochondrial protein biogenesis, but the underlying molecular modification resulting in the observed impairment is not yet known.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Public Reviews:

      Reviewer #2 (Public Review):

      This study elucidates the toxic effects of the lipid aldehyde trans-2-hexadecenal (t-2-hex). The authors show convincingly that t-2-hex induces a strong transcriptional response, leads to proteotoxic stress and causes the accumulation of mitochondrial precursor proteins in the cytosol.

      The data shown are of high quality and well-controlled. The genetic screen for mutants that are hyper-and hypo-sensitive to t-2-hex is elegant and interesting, even if the mechanistic insights from the screen are rather limited. Moreover, the authors show evidence that t-2-hex affects subunits of the TOM complex. However, they do not formally demonstrate that the lipidation of a TOM subunit is responsible for the toxic effect of t-2-hex. A t-2-hex-resistant TOM mutant was not identified. Nevertheless, this is an interesting and inspiring study of high quality. The connection of proteostasis, mitochondrial biogenesis and sphingolipid metabolism is exciting and will certainly lead to many follow-up studies.

      Reviewer #3 (Public Review):

      Summary:

      The authors investigate the effect of high concentrations of the lipid aldehyde trans-2-hexadecenal (t-2-hex) in a yeast deletion strain lacking the detoxification enzyme. Transcriptomic analyses as global read out reveal that a large range of cellular functions across all compartments are affected (transcriptomic changes affect 1/3 of all genes). The authors provide additional analyses, from which they built a model that mitochondrial protein import caused by modification of Tom40 is blocked.

      Our initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant as an experimental approach is only a starting experiment and was aimed to identify as many determinants of t-2-hex toxicity as possible as stated in the manuscript. From this, we developed multiple independent approaches in wild-type (and mutant) cells at low t-2-hex concentrations, demonstrating that proteostasis and mitochondrial protein trafficking are physiologically important targets of the pro-apoptotic lipid. Specifically, proteostasis-specific PACE reporters are robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10 mM t-2-hex in detoxification competent wild type cells (new Figure 3F).

      We do not propose Tom40 as the lipid's primary target, while we show that several subunits of the TOM (and TIM) complex are directly targeted by low t-2-hex concentrations in vitro (Figure 8B), and Tom20 and Tom70 are important for lipid toxicity (Figure 8D) and mitochondrial protein trafficking in vivo (Suppl. Figure 2).

      Strengths:

      Global analyses (transcriptomic and functional genomics approach) to obtain an overview of changes upon yeast treatment with high doses of t-2-hex.

      Weaknesses:

      The use of high concentrations of t-2-hex in combination with a deletion of the detoxifying enzyme Hfd1 limits the possibility to identify physiological relevant changes. From the hundreds of identified targets the authors focus on mitochondrial proteins, which are not clearly comprehensible from the data.

      The initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant is a starting experiment and was aimed to identify as many determinants of t-2-hex toxicity as possible as stated in the manuscript. As stated (page 4), genes up-regulated (>2 log2FC) by t-2-hex were selected and subjected to GO category enrichment analysis (Supplemental Table 1). We found that “Mitochondrial organization” was the most numerous GO group activated by t-2-hex.  Among the strongly t-2-hex induced genes encoding mitochondrial proteins, CIS1 represented the most inducible gene with a known mitochondrial function. Cis1 is the central protein of the MitoCPR pathway, which is specifically induced upon and protects from mitochondrial protein import stress. We further show that proteostasis and mitochondrial protein trafficking are physiologically important targets at low t-2-hex doses in several independent experimental approaches: proteostasis-specific PACE reporters are robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10mM t-2-hex in detoxification competent wild type cells (new Figure 3F); mitochondrial pre-protein accumulation is induced by 10mM t-2-hex in wild type cells (Figure 5G); several subunits of the TOM and TIM complexes are lipidated by low (10mM) t-2-hex doses in wild type cell extracts (Figure 8B), mitochondrial import assays with mt-GFP in intact yeast wild type cells reveal that t-2-hex significantly inhibits import at low (5mM) t-2-hex concentrations (new Suppl. Figure 1). 5-10mM t-2-hex applied here is considerably lower than the published data in human cells with ³ 25mM on intact cells or cell extracts (Jarugumilli et al. 2018).

      The main claim of the manuscript that t-2-hex targets the TOM complex and inhibits mitochondrial protein import is not supported by experimental data as import was not experimentally investigated. The observed accumulation of precursor proteins could have many other reasons (e.g. dissipation of membrane potential, defects in mitochondrial presequence proteases, defects in cytosolic chaperones, modification of mitochondrial precursors by t-2-hex rendering them aggregation prone and thus non-import competent). However, none of these alternative explanations have been experimentally addressed or discussed in the manuscript.

      We have now performed additional experiments, alternative to the pre-protein quantifications, showing that t-2-hex specifically inhibits mitochondrial protein import. We investigated the effect of t-2-hex on mitochondrial protein import using flow cytometric GFP assays in live yeast cells. Specifically, we compared the expression and maturation of GFP targeted either to the cytosol or the mitochondrial matrix and show that low doses of t-2-hex (≥5 μM) significantly inhibited mt-GFP activity compared to cytosolic GFP in wild-type cells (new Supplemental Figure 1B). In contrast, this inhibition was not observed with the saturated derivative, t-2-hex-H2. Flow cytometric rhodamine123 assays revealed that t-2-hex did not alter ΔΨm within the concentration range that efficiently inhibits mt-GFP activity (new Supplemental Figure 1C). Alternative t-2-hex effects such as the direct modification of mitochondrial pre-proteins or cytosolic chaperones, potentially making the precursors prone to aggregation, are less likely, as the mitochondrial and cytosolic GFP used in these import studies differ only by the small, cysteine-free PreSu9 pre-peptide. This information is now included in the Results and Discussion sections.

      Furthermore, many of the results have been reported before (interaction of Tom22 and Tom70 with Hfd1) or observed before (TOM40 as target of t-2-hex in human cells).

      The interaction of Tom22 or Tom70 with Hfd1 has been only reported in high throughput pull-down studies in yeast (Opalinski et al., 2018 and Burri et al., 2006), and no functional connection between Hfd1 lipid detoxification and TOM has been investigated. Here we corroborate these high throughput results by targeted pull-down experiments, which strengthens the new finding that Hfd1 functionally interacts with the TOM complex. Tom40 has been found to be lipidated by high t-2-hex concentrations in human cell extracts in high throughput in vitro proteomic studies (Jarugumilli et al., 2018), but no functional connection between human TOM and t-2-hex has been investigated so far. Here we corroborate these high throughput results by targeted experiments, which strengthens the new findings that t-2-hex and TOM interact functionally.

      Recommendations for the authors:

      Reviewer #2 (Recommendations For The Authors):

      Congratulations on this exciting study. Even if some of the mechanistic details will have to be addressed in further studies (which of the modified sites are physiologically relevant; which sites are modified in vivo without external addition of t-2-hex) this study is inspiring and opens a new direction of mitochondrial research. I therefore fully support publication of this nice study in its current form.

      Reviewer #3 (Recommendations For The Authors):

      Two of the reviewers pointed out that the observation of precursors in whole cell extract is not sufficient to draw conclusions on mitochondrial protein import rates. The authors did not provide any new experiments but argued that a recent publication (Weidberg and Amon, 2018) had used the same readout for this conclusion. Why this manuscript was accepted with this statement is not known to this reviewer, but it does not change the fact, that the conclusion is not valid. Many alternative explanations are possible (see public review) and the claim that the import competence of the TOM complex is affected upon t-2-hex treatment is not appropriate.

      We have now performed new experiments addressing the inhibition of mitochondrial protein import by t-2-hex as an alternative to our precursor accumulation assays. We compared the induced expression of cytosolic and mitochondrial GFP by flow cytometry as a quantitative mitochondrial import assay (Sirk et al., Cytometry A. 2003 Nov; 56(1) 15-22). Low doses of t-2-hex (≥5 μM) significantly inhibited mt-GFP activity as compared to cytosolic GFP in wild-type cells (new Supplemental Figure 1B). This inhibition of mitochondrial GFP is independent of mitochondrial membrane potential perturbation (new Supplemental Figure 1C) and alternative t-2-hex effects, such as the direct modification of the mtGFP precursor or cytosolic chaperones are less likely, as the mitochondrial and cytosolic GFP used in these import studies differ only by the small, cysteine-free PreSu9 pre-peptide.

      The first sentence of the abstract states that t-2-hex „induces mitochondrial dysfunction in a conserved manner from yeast to human". I find two issues with this statement: 1) if the mechanism is known what is the question addressed in the present manuscript and 2) the second sentence of the results fully contradicts the above sentence „In human cells, t-2-hex causes mitochondrial dysfunction by directly stimulating Bax-oligomerisation at the outer mitochondrial membrane. In yeast, however, t-2-hex efficiently interferes with mitochondrial function and cell growth in a Bax independent manner."

      We agree that the first sentence was misleading, this has been fixed now in the revised version.

      The first reviewer requested a repetition of key experiments with lower concentrations and the authors provided additional in vitro data, however, for this, 10 uM is still very high. To gain valuable and physiological relevant data the initial transcriptomic analysis should be repeated with a low amount and in a wild-type yeast background.

      Published t-2-hex chemoproteomic experiments on human cell extracts were performed at higher concentrations (>25mM) and human Bax is hardly lipidated by 10mM t-2-hex (Jarugumilli et al., 2018), therefore the in vitro lipidation data provided in our study should be considered a low t-2-hex dose. The initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant is a starting experiment and was aimed at identifying as many determinants of t-2-hex toxicity as possible. Building on this, we further show that proteostasis and mitochondrial protein trafficking, the relevant cellular functions for our study, are physiologically important targets at low t-2-hex doses in several independent experimental approaches: proteostasis-specific gene expression is robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10mM t-2-hex in detoxification competent wild type cells (new Figure 3F); mitochondrial pre-protein accumulation is induced by 10mM t-2-hex in wild type cells (Figure 5G); several subunits of the TOM and TIM complexes are lipidated by low (10mM) t-2-hex doses in vitro in wild type extracts (Figure 8B), mitochondrial import assays with mt-GFP in intact yeast wild type cells reveal that t-2-hex significantly inhibits import at low (5mM) t-2-hex concentrations (new Suppl. Figure 1).

      As already stated above there are many alternative explanations for the observed accumulation of precursor proteins, e.g. the decreased proteasome activity could be cause and not consequence. Also, the modification of precursors directly upon translation in the cytosol could likely impact on their further transport and result in direct aggregation in the cytosol.

      As mentioned above, we have now corroborated the t-2-hex specific mitochondrial protein import defect by alternative in vivo experiments, which are not dependent on the accumulation of mitochondrial precursors. We have tested now the possibility that decreased proteasome activity could indirectly inhibit mitochondrial import. This is not the case because a rpn4 mutant with impaired proteasomal activity induces normal mtGFP levels (new Suppl. Figure 1D). We cannot exclude that the modification of precursors by t-2-hex upon translation might additionally impact on the transport of some mitochondrial pre-proteins. However, mitochondrial and cytosolic GFP used in the import studies only differ in the small cysteine-free PreSu9 pre-peptide making it very unlikely that precursor lipidation is secondarily responsible for the observed import defect.

      Many of the comments after first reviewing the manuscript were not addressed experimentally although many of the suggested experiments are easy to perform. I can only encourage the authors to provide more experimental support and controls, as the claims are currently not sufficiently supported.

      In the two revisions of our manuscript, we have included several control experiments to better link the pro-apoptotic lipid t-2-hex with mitochondrial import stress. These include: in vitro lipidation of TOM/TIM subunits by low t-2-hex concentrations, t-2-hex tolerance and recovery of mitochondrial protein import in specific tom mutants, inhibition of mitochondrial protein import (pre-protein and mtGFP assays) by low t-2-hex doses independently on mitochondrial membrane potential and proteasome activity, and induction of proteostasis specific gene expression by low t-2-hex doses.

    1. eLife Assessment

      In this manuscript, Lim and collaborators present a useful system for developing self-amplifying RNA that should not provoke a strong host inflammatory response. However, some of the claims are incomplete; additional experiments to investigate the effects on translation of the gene of interest and replication efficiency of the self-amplifying RNA could strengthen the manuscript.

    2. Reviewer #1 (Public review):

      Summary:

      The authors have developed self-amplifying RNAs (saRNAs) encoding additional genes to suppress dsRNA-related inflammatory responses and cytokine release. Their results demonstrate that saRNA constructs encoding anti-inflammatory genes effectively reduce cytotoxicity and cytokine production, enhancing the potential of saRNAs. This work is significant for advancing saRNA therapeutics by mitigating unintended immune activation.

      Strengths:

      This study successfully demonstrates the concept of enhancing saRNA applications by encoding immune-suppressive genes. A key challenge for saRNA-based therapeutics, particularly for non-vaccine applications, is the innate immune response triggered by dsRNA recognition. By leveraging viral protein properties to suppress immunity, the authors provide a novel strategy to overcome this limitation. The study presents a well-designed approach with potential implications for improving saRNA stability and minimizing inflammatory side effects.

      Weaknesses:

      (1) Impact on Cellular Translation:

      The authors demonstrate that modified saRNAs with additional components enhance transgene expression by inhibiting dsRNA-sensing pathways. However, it is unclear whether these modifications influence global cellular translation beyond the expression of GFP and mScarlet-3 (which are encoded by the saRNA itself). Conducting a polysome profiling analysis or a puromycin labeling assay would clarify whether the modified saRNAs alter overall translation efficiency. This additional data would strengthen the conclusions regarding the specificity of dsRNA-sensing inhibition.

      (2) Stability and Replication Efficiency of Long saRNA Constructs:

      The saRNA constructs used in this study exceed 16 kb, making them more fragile and challenging to handle. Assessing their mRNA integrity and quality would be crucial to ensure their robustness.<br /> Furthermore, the replicative capacity of the designed saRNAs should be confirmed. Since Figure 4 shows lower inflammatory cytokine production when encoding srIkBα and srIkBα-Smad7-SOCS1, it is important to determine whether this effect is due to reduced immune activation or impaired replication. Providing data on replication efficiency and expression levels of the encoded anti-inflammatory proteins would help rule out the possibility that reduced cytokine production is a consequence of lower replication.

      (3) Comparative Data with Native saRNA:

      Including native saRNA controls in Figures 5-7 would allow for a clearer assessment of the impact of additional genes on cytokine production. This comparison would help distinguish the effect of the encoded suppressor proteins from other potential factors.

      (4) In vivo Validation and Safety Considerations:

      Have the authors considered evaluating the in vivo potential of these saRNA constructs? Conducting animal studies would provide stronger evidence for their therapeutic applicability. If in vivo experiments have not been performed, discussing potential challenges - such as saRNA persistence, biodistribution, and possible secondary effects-would be valuable.

      (5) Immune Response to Viral Proteins:

      Since the inhibitors of dsRNA-sensing proteins (E3, NSs, and L*) are viral proteins, they would be expected to induce an immune response. Analyzing these effects in vivo would add insight into the applicability of this approach.

      (6) Streamlining the Discussion Section:

      The discussion is quite lengthy. To improve readability, some content - such as the rationale for gene selection-could be moved to the Results section. Additionally, the descriptions of Figure 3 should be consolidated into a single section under a broader heading for improved coherence.

    3. Reviewer #2 (Public review):

      Summary:

      Lim et al. have developed a self-amplifying RNA (saRNA) design that incorporates immunomodulatory viral proteins, and show that the novel design results in enhanced protein expression in vitro in mouse primary fibroblast-like synoviocytes. They test constructs including saRNA with the vaccinia virus E3 protein and another with E3, Toscana virus NS protein and Theiler's virus L protein (E3 + NS + L), and another with srIκBα-Smad7-SOCS1. They have also tested whether ML336, an antiviral, enables control of transgene expression.

      Strengths:

      The experiments are generally well-designed and offer mechanistic insight into the RNA-sensing pathways that confer enhanced saRNA expression. The experiments are carried out over a long timescale, which shows the enhance effect of the saRNA E3 design compared to the control. Furthermore, the inhibitors are shown to maintain the cell number, and reduce basal activation factor-⍺ levels.

      Weaknesses:

      One limitation of this manuscript is that the RNA is not well characterized; some of the constructs are quite long and the RNA integrity has not been analyzed. Furthermore, for constructs with multiple proteins, it's imperative to confirm the expression of each protein to confirm that any therapeutic effect is from the effector protein (e.g. E3, NS, L). The ML336 was only tested at one concentration; it is standard in the field to do a dose-response curve. These experiments were all done in vitro in mouse cells, thus limiting the conclusion we can make about mechanisms in a human system.

    4. Author response:

      Reviewer #1 (Public review):

      Summary:

      The authors have developed self-amplifying RNAs (saRNAs) encoding additional genes to suppress dsRNA-related inflammatory responses and cytokine release. Their results demonstrate that saRNA constructs encoding anti-inflammatory genes effectively reduce cytotoxicity and cytokine production, enhancing the potential of saRNAs. This work is significant for advancing saRNA therapeutics by mitigating unintended immune activation.

      Strengths:

      This study successfully demonstrates the concept of enhancing saRNA applications by encoding immune-suppressive genes. A key challenge for saRNA-based therapeutics, particularly for non-vaccine applications, is the innate immune response triggered by dsRNA recognition. By leveraging viral protein properties to suppress immunity, the authors provide a novel strategy to overcome this limitation. The study presents a well-designed approach with potential implications for improving saRNA stability and minimizing inflammatory side effects.

      We thank Reviewer #1 for their thorough review and for recognizing both the significance of our work and the potential of our strategy to expand saRNA applications beyond vaccines.

      Weaknesses:

      (1) Impact on Cellular Translation:

      The authors demonstrate that modified saRNAs with additional components enhance transgene expression by inhibiting dsRNA-sensing pathways. However, it is unclear whether these modifications influence global cellular translation beyond the expression of GFP and mScarlet-3 (which are encoded by the saRNA itself). Conducting a polysome profiling analysis or a puromycin labeling assay would clarify whether the modified saRNAs alter overall translation efficiency. This additional data would strengthen the conclusions regarding the specificity of dsRNA-sensing inhibition.

      We thank the reviewer for this helpful insight and suggestion. We aim to conduct a puromycin labelling assay to clarify the effect of the various saRNA constructs on translation efficiency.

      (2) Stability and Replication Efficiency of Long saRNA Constructs:

      The saRNA constructs used in this study exceed 16 kb, making them more fragile and challenging to handle. Assessing their mRNA integrity and quality would be crucial to ensure their robustness.

      Furthermore, the replicative capacity of the designed saRNAs should be confirmed. Since Figure 4 shows lower inflammatory cytokine production when encoding srIkBα and srIkBα-Smad7-SOCS1, it is important to determine whether this effect is due to reduced immune activation or impaired replication. Providing data on replication efficiency and expression levels of the encoded anti-inflammatory proteins would help rule out the possibility that reduced cytokine production is a consequence of lower replication.

      This is another very helpful comment. We will conduct an analysis of saRNA integrity and quality by denaturing gel electrophoresis. To examine replicative capacity of the saRNA constructs, we aim to conduct RT-qPCR experiments.

      (3) Comparative Data with Native saRNA:

      Including native saRNA controls in Figures 5-7 would allow for a clearer assessment of the impact of additional genes on cytokine production. This comparison would help distinguish the effect of the encoded suppressor proteins from other potential factors.

      Thank you for your suggestion. We will implement this change in the next version of the manuscript.

      (4) In vivo Validation and Safety Considerations:

      Have the authors considered evaluating the in vivo potential of these saRNA constructs? Conducting animal studies would provide stronger evidence for their therapeutic applicability. If in vivo experiments have not been performed, discussing potential challenges - such as saRNA persistence, biodistribution, and possible secondary effects-would be valuable.

      (5) Immune Response to Viral Proteins:

      Since the inhibitors of dsRNA-sensing proteins (E3, NSs, and L*) are viral proteins, they would be expected to induce an immune response. Analyzing these effects in vivo would add insight into the applicability of this approach.

      We recognize the importance of in vivo studies and immune cell responses and plan to incorporate in vivo imaging in future studies to investigate these interactions, as well as examining delivery of various cargoes via saRNA to determine potential therapeutic benefits in different animal models of inflammatory pain, but such studies are beyond the scope of this current investigation. As suggested by the reviewer, we will incorporate a section on potential challenges of in vivo saRNA work in the revised manuscript.

      (6) Streamlining the Discussion Section:

      The discussion is quite lengthy. To improve readability, some content - such as the rationale for gene selection-could be moved to the Results section. Additionally, the descriptions of Figure 3 should be consolidated into a single section under a broader heading for improved coherence.

      Thank you for your suggestions, we will make these changes in the next revision.

      Reviewer #2 (Public review):

      Summary:

      Lim et al. have developed a self-amplifying RNA (saRNA) design that incorporates immunomodulatory viral proteins, and show that the novel design results in enhanced protein expression in vitro in mouse primary fibroblast-like synoviocytes. They test constructs including saRNA with the vaccinia virus E3 protein and another with E3, Toscana virus NS protein and Theiler's virus L protein (E3 + NS + L), and another with srIκBα-Smad7-SOCS1. They have also tested whether ML336, an antiviral, enables control of transgene expression.

      Strengths:

      The experiments are generally well-designed and offer mechanistic insight into the RNA-sensing pathways that confer enhanced saRNA expression. The experiments are carried out over a long timescale, which shows the enhance effect of the saRNA E3 design compared to the control. Furthermore, the inhibitors are shown to maintain the cell number, and reduce basal activation factor-⍺ levels.

      We thank Reviewer #2 for their detailed assessment and recognition of the mechanistic insights provided by our study.

      Weaknesses:

      One limitation of this manuscript is that the RNA is not well characterized; some of the constructs are quite long and the RNA integrity has not been analyzed. Furthermore, for constructs with multiple proteins, it's imperative to confirm the expression of each protein to confirm that any therapeutic effect is from the effector protein (e.g. E3, NS, L). The ML336 was only tested at one concentration; it is standard in the field to do a dose-response curve. These experiments were all done in vitro in mouse cells, thus limiting the conclusion we can make about mechanisms in a human system.

      We agree that these are weaknesses of our work. We plan to address some of these weaknesses by performing a dose response curve for ML336, examining saRNA integrity through denaturing gel electrophoresis, and will also aim to provide additional evidence for effects of effector proteins through RT-qPCR. We are also looking into testing these constructs in patient-derived FLS.

    1. eLife Assessment

      This useful study describes a novel method for imaging NAD(P)H fluorescence lifetime and thus metabolic states in the Drosophila brain. These solid findings support recent work demonstrating the importance of energy homeostasis to sustain memory formation and maintenance. Further efforts to demonstrate the adequacy of the statistical methods and the significance of the observed differences in FLIM signals in the α/β KCs would greatly enhance the manuscript. The approach will be helpful for researchers working with systems where genetic manipulation is challenging.

    2. Reviewer #1 (Public review):

      Summary:

      The authors present a novel usage of fluorescence lifetime imaging microscopy (FLIM) to measure NAD(P)H autofluorescence in the Drosophila brain, as a proxy for cellular metabolic/redox states. This new method relies on the fact that both NADH and NADPH are autofluorescent, with a different excitation lifetime depending on whether they are free (indicating glycolysis) or protein-bound (indicating oxidative phosphorylation). The authors successfully use this method in Drosophila to measure changes in metabolic activity across different areas of the fly brain, with a particular focus on the main center for associative memory: the mushroom body.

      Strengths:

      The authors have made a commendable effort to explain the technical aspects of the method in accessible language. This clarity will benefit both non-experts seeking to understand the methodology and researchers interested in applying FLIM to Drosophila in other contexts.

      Weaknesses:

      (1) Despite being statistically significant, the learning-induced change in f-free in α/β Kenyon cells is minimal (a decrease from 0.76 to 0.73, with a high variability). The authors should provide justification for why they believe this small effect represents a meaningful shift in neuronal metabolic state.

      (2) The lack of experiments examining the effects of long-term memory (after spaced or massed conditioning) seems like a missed opportunity. Such experiments could likely reveal more drastic changes in the metabolic profiles of KCs, as a consequence of memory consolidation processes.

      (3) The discussion is mostly just a summary of the findings. It would be useful if the authors could discuss potential future applications of their method and new research questions that it could help address.

    3. Reviewer #2 (Public review):

      This manuscript presents a compelling application of NAD(P)H fluorescence lifetime imaging (FLIM) to study metabolic activity in the Drosophila brain. The authors reveal regional differences in oxidative and glycolytic metabolism, with a particular focus on the mushroom body, a key structure involved in associative learning and memory. In particular, they identify metabolic shifts in α/β Kenyon cells following classical conditioning, consistent with their established role in energy-demanding middle- and long-term memories.

      These results highlight the potential of label-free FLIM for in-vivo neural circuit studies, providing a powerful complement to genetically encoded sensors. This study is well-conducted and employs rigorous analysis, including careful curve fitting and well-designed controls, to ensure the robustness of its findings. It should serve as a valuable technical reference for researchers interested in using FLIM to study neural metabolism in vivo. Overall, this work represents an important step in the application of FLIM to study the interactions between metabolic processes, neural activity, and cognitive function.

    4. Reviewer #3 (Public review):

      This study investigates the characteristics of the autofluorescence signal excited by 740 nm 2-photon excitation, in the range of 420-500 nm, across the Drosophila brain. The fluorescence lifetime (FL) appears bi-exponential, with a short 0.4 ns time constant followed by a longer decay. The lifetime decay and the resulting parameter fits vary across the brain. The resulting maps reveal anatomical landmarks, which simultaneous imaging of genetically encoded fluorescent proteins helps to identify. Past work has shown that the autofluorescence decay time course reflects the balance of the redox enzyme NAD(P)H vs. its protein-bound form. The ratio of free-to-bound NADPH is thought to indicate relative glycolysis vs. oxidative phosphorylation, and thus shifts in the free-to-bound ratio may indicate shifts in metabolic pathways. The basics of this measure have been demonstrated in other organisms, and this study is the first to use the FLIM module of the STELLARIS 8 FALCON microscope from Leica to measure autofluorescence lifetime in the brain of the fly. Methods include registering the brains of different flies to a common template and masking out anatomical regions of interest using fluorescence proteins.

      The analysis relies on fitting an FL decay model with two free parameters, f_free and t_bound. F_free is the fraction of the normalized curve contributed by a decaying exponential with a time constant of 0.4 ns, thought to represent the FL of free NADPH or NADH, which apparently cannot be distinguished. T_bound is the time constant of the second exponential, with scalar amplitude = (1-f_free). The T_bound fit is thought to represent the decay time constant of protein-bound NADPH but can differ depending on the protein. The study shows that across the brain, T_bound can range from 0 to >5 ns, whereas f_free can range from 0.5 to 0.9 (Figure 1a). These methods appear to be solid, the full range of fits are reported, including maximum likelihood quality parameters, and can be benchmarks for future studies.

      The authors measure the properties of NADPH-related autofluorescence of Kenyon Cells (KCs) of the fly mushroom body. The results from the three main figures are:

      (1) Somata and calyx of mushroom bodies have a longer average tau_bound than other regions (Figure 1e);

      (2) The f_free fit is higher for the calyx (input synapses) region than for KC somata (Figure 2b);

      (3) The average across flies of average f_free fits in alpha/beta KC somata decreases from 0.734 to 0.718. Based on the first two findings, an accurate title would be "Autofluorecense lifetime imaging reveals regional differences in NADPH state in Drosophila mushroom bodies."

      The third finding is the basis for the title of the paper and the support for this claim is unconvincing. First, the difference in alpha/beta f_free (p-value of 4.98E-2) is small compared to the measured difference in f_free between somas and calyces. It's smaller even than the difference in average soma f_free across datasets (Figure 2b vs c). The metric is also quite derived; first, the model is fit to each (binned) voxel, then the distribution across voxels is averaged and then averaged across flies. If the voxel distributions of f_free are similar to those shown in Supplementary Figure 2, then the actual f_free fits could range between 0.6-0.8. A more convincing statistical test might be to compare the distributions across voxels between alpha/beta vs alpha'/beta' vs. gamma KCs, perhaps with bootstrapping and including appropriate controls for multiple comparisons.

      I recommend the authors address two concerns. First, what degree of fluctuation in autofluorescence decay can we expect over time, e.g. over circadian cycles? That would be helpful in evaluating the magnitude of changes following conditioning. And second, if the authors think that metabolism shifts to OXPHOS over glycolosis, are there further genetic manipulations they could make? They test LDH knockdown in gamma KCs, why not knock it down in alpha/beta neurons? The prediction might be that if it prevents the shift to OXPHOS, the shift in f_free distribution in alpha/beta KCs would be attenuated. The extensive library of genetic reagents is an advantage of working with flies, but it comes with a higher standard for corroborating claims.

      FLIM as a method is not yet widely prevalent in fly neuroscience, but recent demonstrations of its potential are likely to increase its use. Future efforts will benefit from the description of the properties of the autofluorescence signal to evaluate how autofluorescence may impact measures of FL of genetically engineered indicators.

    5. Author response:

      Thanks for the positive review of our manuscript and for appreciating our work.

      We align in many ways with the reviewers comments.. Our initial finding concerning the slight shift of f_free in a/b neurons after conditioning is interesting but we agree it would certainly deserve a follow-up to substantiate its link with memory formation. We also agree that an analysis in distribution rather than through an averaged signal might be more sensitive.

      We however have to cope with the fact that extending our investigation would require manpower resources that are no longer available. Therefore we appreciate the suggestion made by the 3 reviewers to restrain the claim and hence change the title to "In vivo NAD(P)H autofluorescence lifetime imaging reveals metabolic heterogeneity within the Drosophila mushroom body.". We find it matches better with the scope of this study which is mostly to showcase the potential of NAD(P)H FLIM to quantify variations in metabolism in Drosophila brain rather than firmly testing a specific hypothesis linked to memory formation. In this respect, we do provide quantitative results showing metabolic profile variations between brain tissues such as the somata and calyx regions but also between different Kenyon cells subtypes. We would then present the shifts of f_free induced by conditioning as a curio that might entice future work, as advised by Reviewer #2.

      Altogether, in the revised version we will change the title to restrain the claim, move two supplementary figures as main figures to better focus on and describe the registration process. We will also correct the figure panels pointed by the reviewers and add individual samples to our boxplots. We will also slightly compress the introduction and expand the discussion on potential applications. Finally, we will evaluate if statistical tests based on distributions may be more sensitive to observe a significant shift in FLIM signal in the a/b KCs after conditioning, to strengthen our last observation if confirmed.

    1. eLife Assessment

      This study presents an important finding on how lentiviral infection has driven the diversification of the HIV/SIV entry receptor CD4. Using a combination of molecular evolution approaches coupled with functional testing of extant and ancestral reconstructions of great ape CD4, the authors provide solid evidence to support the idea that endemic simian immunodeficiency virus infection in gorillas have selected for gorilla CD4 alleles that are more resistant to SIV infection. Expanding the study to interrogate the evolution and function of additional primate CD4 sequences could yield even stronger evidence.

    1. eLife Assessment

      This report explores the role of matrix metalloprotease MMP21 in left-right patterning in Xenopus. Based on a series of compelling experiments, the authors demonstrate that MMP21 can be secreted and acts upstream of dand5 without affecting cilia flow. The experiments are interesting and valuable; however, the claims by the authors lack consideration of other models that could also explain their findings.

    2. Joint Public Review:

      The manuscript describes the role of mmp21, a metallopeptidase, in left-right patterning. MMP21 has been implicated in genetic studies of patients with heterotaxy and the authors add an additional case. However, a molecular mechanism for Htx/LR patterning defects is not clear although one previous study implicated Notch signaling. The authors find that mmp21 does indeed cause LR patterning defects in Xenopus consistent with work in mice and zebrafish without affecting cilia motility. Importantly, the authors extend this work to place mmp21 in the LR pathway between dand5 (in the nodal cascade) and the cilia-driven sensation of flow. With RNA overexpression studies, the authors show MMP21 can induce Nodal signaling bilaterally suggesting it is an activator of the pathway, potentially through regulation of dand5 asymmetry. The authors also show that the role of MMP21 is upstream of another matrix metalloprotease CIROP which is tethered to the plasma membrane and possibly the cilium. They propose that mmp21, which is secreted, may represent a morphogen that is asymmetrically distributed along the LR axis due to cilia-driven flow and sensed by sensory cilia in the LRO.

      The authors attempt to address a highly controversial subject in the LR patterning field, that is, the debate between Nodal Vesicular Particles (NVP, ie morphogens) being driven by cilia to activate signaling on the left and the Two Cilia model which posits that mechanosensation of fluid flow and not morphogens drive asymmetric organogenesis.

      The model they propose is that mmp21 is secreted in the center of the LRO. LRO cilia generate leftward flow driving mmp21 to the left where sensory cilia at the LRO margin detect the mmp21 via cirop and suppress dand5, leading to activation of Nodal and Pitx2 expression.

      First and foremost, the authors need to consider alternative models in the discussion and acknowledge the strengths and weaknesses of their work. All three reviewers felt that their conclusion that mmp21 is a morphogen is premature and that other models could also fit their data which needs to be discussed. The authors need to soften the conclusion that other models have been excluded.

    1. eLife Assessment

      This fundamental work presents two clinically relevant BMP4 mutations that contribute to vertebrate development. The compelling evidence, both from wet lab and AI generated predictions, supports that the site-specific cleavage at the BMP4 pro-domain precisely regulates its function and provides mechanistic insight how homodimers and heterodimers behave differently. The work will be of board interest to researchers working on growth factor signaling mechanisms and vertebrate development.

    2. Reviewer #1 (Public review):

      Summary:

      The authors demonstrate that two human preproprotein human mutations in the BMP4 gene cause a defect in proprotein cleavage and BMP4 mature ligand formation, leading to hypomorphic phenotypes in mouse knock-in alleles and in Xenopus embryo assays.

      Strengths:

      They provide compelling biochemical and in vivo analyses supporting their conclusions, showing the reduced processing of the proprotein and concomitant reduced mature BMP4 ligand protein from impressively mouse embryonic lysates. They perform excellent analysis of the embryo and post-natal phenotypes demonstrating the hypomorphic nature of these alleles. Interesting phenotypic differences between the S91C and E93G mutants are shown with excellent hypotheses for the differences. Their results support that BMP4 heterodimers act predominantly throughout embryogenesis whereas BMP4 homodimers play essential roles at later developmental stages.

      Weaknesses:

      In the revision the authors have appropriately addressed the previous minor weaknesses.

    3. Reviewer #2 (Public review):

      Summary:

      The revised paper by Kim et al. reports two disease mutations in proBMP4, S91C and E93G, disrupt the FAM20C phosphorylation site at Ser91, blocking the activation of proBMP4 homodimers, while still allowing BMP4/7 heterodimers to function. Analysis of DMZ explants from Xenopus embryos expressing the proBMP4 S91C or E93G mutants showed reduced expression of pSmad1 and tbxt1. The expert amphibian tissue transplant studies were expanded to in vivo studies in Bmp4S91C/+ and Bmp4E93G/+ mice, highlighting the impact of these mutations on embryonic development, particularly in female mice, consistent with patient studies. Additionally, studies in mouse embryonic fibroblasts (MEFs) demonstrated that the mutations did not affect proBMP4 glycosylation or ER-to-Golgi transport but appeared to inhibit the furin-dependent cleavage of proBMP4 to BMP4. Based on these findings and AI modeling using AlphaFold of proBMP4, the authors speculate that pSer91 influences access of furin to its cleavage site at Arg289AlaLysArg292 in a new "Ideas and Speculation" section. Overall, the authors addressed the reviewers' comments, improving the presentation.

      Strengths:

      The strengths of this work continue to lie in the elegant Xenopus and mouse studies that elucidate the impact of the S91C and E93G disease mutations on BMP signaling and embryonic development. Including an "Ideas and Speculation" subsection for mechanistic ideas reduces some shortcomings regarding the analysis of the underlying mechanisms.

      Weaknesses:

      (Minor) In Figure S1 and lines 165-174 and 179-180, the authors should consider that, unlike the wild-type protein (Ser), which can be reversibly phosphorylated or dephosphorylated, phosphomimic mutations are locked into mimicking either the phosphorylated state (Asp) or the non-phosphorylated state (Ala). Consequently, if the S91D mutant exhibits lower activity than WT, it could imply that S91D interferes with other regulatory constraints, as the authors suggest. However, it may also be inhibiting activation. Therefore, caution is warranted when comparing S91D with S91C to conclude that Ser91 phosphorylation increases BMP4 activity. While additional experiments are not necessary, further consideration is essential.

      In Figure 4, panels A, E, and I, the proBMP bands in the mouse embryonic lysates and MEFs expressing the mutations show a clear size shift. Are these shifts a cause or a consequence of the lack of cleavage? Regardless, the size shifts should be explicitly noted.

      (Minor) In line 314, the authors should consider modifying the wording to: "is required for modulating proprotein convertase..."

      (Minor) In lines 394-399, the authors cleverly speculate that pS91 interacts with Arg289-the essential P4 arginine for furin processing. If so, this interaction could hinder the cleavage of proBMP4, as indicated by the results in Figure S1. The discussion would benefit from considering that, contrary to their favored model, dephosphorylation at Ser91 might actually facilitate cleavage.

    4. Reviewer #3 (Public review):

      Summary:

      The authors describe important new biochemical elements in the synthesis of a class of critical developmental signaling molecules, BMP4. They also present a highly detailed description of developmental anomalies in mice bearing known human mutations at these specific elements.

      Strengths:

      This paper presents exceptionally detailed descriptions of pathologies occurring in BMP4 mutant mice. Novel findings are shown regarding the interaction of propeptide phosphorylation and convertase cleavage, both of which will move the field forward. Lastly, a provocative hypothesis regarding furin access to cleavage sites is presented, supported by Alphafold predictions.

    5. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review): Summary:

      The authors demonstrate that two human preproprotein human mutations in the BMP4 gene cause a defect in proprotein cleavage and BMP4 mature ligand formation, leading to hypomorphic phenotypes in mouse knock-in alleles and in Xenopus embryo assays.

      Strengths:

      They provide compelling biochemical and in vivo analyses supporting their conclusions, showing the reduced processing of the proprotein and concomitant reduced mature BMP4 ligand protein from impressively mouse embryonic lysates. They perform excellent analysis of the embryo and post-natal phenotypes demonstrating the hypomorphic nature of these alleles. Interesting phenotypic differences between the S91C and E93G mutants are shown with excellent hypotheses for the differences. Their results support that BMP4 heterodimers act predominantly throughout embryogenesis whereas BMP4 homodimers play essential roles at later developmental stages.

      Weaknesses:

      (1) A control of BMP7 alone in the Xenopus assays seems important to excludeBMP7 homodimer activity in these assays.

      We and other have shown that BMP7 homodimers have weak or no activity while BMP4/7 heterodimers single at a much higher level than either BMP4 or BMP7 homodimers in Xenopus ectodermal and mesodermal cells. We have expanded the description of these published findings in the results section (lines 182-187). We have also added representative examples of experiments in which BMP4 and BMP7 alone controls are included (new Fig. S2). Since the level of activity of BMP7 + BMP4 variants is equivalent to that of BMP7 + WT BMP4, this cannot be accounted for by BMP7 homodimers.

      (2) The Discussion could be strengthened by more in-depth explanations of how BMP4 homodimer versus heterodimer signaling is supported by the results, so that readers do not have to think it all through themselves. Similarly, a discussion of why the S91C mutant has a stronger phenotype than E93G early in the Discussion would be helpful or least mention that it will be addressed later.

      We have revised the discussion as suggested by the reviewer. Please see responses to recommendations 2-4 below.

      Reviewer #1 (Recommendations for the authors):

      (1) A control of BMP7 injection alone seems missing when comparing the BMP4/7 variants. BMP4 in the embryo assays presented in Fig 1. Is it not possible that the activity observed is BMP7 homodimers, perhaps due to inhibited heterodimer formation by the BMP4 variant?

      Multiple published studies have shown that BMP7 homodimers have weak or no activity in Xenopus ectodermal and mesodermal cells, and that ½ dose of RNA encoding BMP4 and BMP7 together signals at a higher level than does a full dose of RNA encoding either BMP4 or BMP7 alone. We have expanded our description of these published findings (lines 182-187), have included additional details about RNA doses that were injected (line 156, 175, 182) and have added representative examples of experiments in which BMP4 and BMP7 controls were included in a new Figure (Fig. S2).

      (2) In reading the Discussion, I was continually thinking of the stronger phenotype of the S91C mutant compared to the E93G one, although both are discussed together throughout most of the Discussion. Only at the end of the Discussion is the stronger phenotype of S91C discussed with a compelling explanation for the stronger phenotype, not related to the phosphorylation site function. I wonder if it would be better placed earlier in Discussion or at least mentioned the difference in phenotypes that will be discussed later.

      We have moved the possible explanation of differences between Bmp4<sup>S91C</sup> and Bmp4<sup>E93G</sup> mutants to immediately follow the introductory paragraph of the results section.

      (3) Along these same lines, why is it that the E93G exhibits rather normal cleavage at E10.5? Might the mechanisms of cleavage vary in different contexts with phosphorylation-dependent cleavage not functioning at early stages of development? I believe the hypothesis is that it is cleaved due to heterodimerization with BMP7. More discussion of this excellent hypothesis should be provided with clear statements, rather than inferences, if I'm understanding this correctly. For example, I had to read 3 times the first sentence of the last paragraph on p.14 before I understood it. Better to break that sentence down and the one that follows it, so it is easier to understand.

      We have rewritten and expanded the paragraphs describing phenotypic and biochemical evidence for defective homodimer but not heterodimer signaling as suggested (lines 343-375). We have also more explicitly stated the possibility that normal cleavage of BMP4<sup>E93G</sup> in embryonic lystates may be due to a predominance of BMP4/7 heterodimers in early embryonic stages or spatiotemporal differences in phosphorylation-dependent cleavage of BMP4 homodimers (lines 369-372)

      (4) Similarly the last paragraph of the Discussion mentions that the authors provide evidence of BMP4 homodimer signaling. I agree with the authors, but I had to think through the evidence myself. Better if the authors clearly explain the evidence that points to this, as this is a very good point of

      See response to point 3, above. Thank you for these useful suggestions.

      (5) Last sentence, first paragraph on p.11 should be qualified for the E93G mutant to E13.5, since it was normal at E10.5 regarding Figure 4 results.

      Thank you for pointing this out. It has been corrected.

      (6) Skip the PC acronym, since it is only repeated once in the text and hard to remember almost 10 pages later when it is used again.

      We have corrected this.

      (7) In the Discussion, a typo in "a single intramolecular disulfide bond that stabilizes the dimer", should be 'intermolecular'.

      Thank you for catching our switch in the use of inter- and intramolecular. We have corrected this (lines 334-335).

      (8) At times the E93G mutant is referred to having early lethality, often in conjunction with S91C, while other times it is referred to as late lethality. Considering that the homozygotes die postnatally after weaning, most would consider it late lethality. In contrast S91C is indeed an early lethal.

      We have changed the wording in the introduction to state that “mice carrying Bmp4<sup>S91C</sup> or Bmp4<sup>E93G</sup> knock in mutations show embryonic or enhanced postnatal lethality, respectively,… (lines 141-143)” and have removed the word “early” from the title.

      Reviewer #2 (Public review): Summary:

      Kim et al. report that two disease mutations in proBMP4, Ser91Cys and Glu93Gly, which disrupt the Ser91 FAM20C phosphorylation site, block the activation of proBMP4 homodimers. Consequently, analysis of DMZ explants from Xenopus embryos expressing the proBMP4 S91C or E93G mutants showed reduced pSmad1 and tbxt1 expression. The block in BMP4 activity caused by the mutations could be overcome by co-expression of BMP7, suggesting that the missense mutations selectively affect the activity of BMP4 homodimers but not BMP4/7 heterodimers. The expert amphibian tissue transplant studies were extended to in vivo studies in Bmp4S91C/+ and Bmp4E93G/+ mice, demonstrating the impact of these mutations on embryonic development, particularly in female mice, in line with patient studies. Finally, studies in MEFs revealed that the mutations did not affect proBMP4 glycosylation or ER-to-Golgi transport but appeared to inhibit the furin-dependent cleavage of proBMP4 to BMP4. Based on these findings and AI (AlphaFold) modeling of proBMP4, the authors speculate that pSer91 influences access of furin to its cleavage site at Arg289AlaLysArg292.

      Strengths:

      The Xenopus and mouse studies are valuable and elegantly describe the impact of the S91C and E93G disease mutations on BMP signaling and embryonic development.

      Weaknesses:

      The interpretation of how the mutations may disturb the furin-mediated cleavage of proBMP4 is underdeveloped and does not consider all of their data. Understanding how pS91 influences the furin-dependent cleavage at Arg292 seems to be the crux of this work and thus warrants more consideration. Specifically:

      (1) Figure S1 may be significantly more informative than implied. The authors report that BMP4S91D activates pSmad1 only incrementally better than S91C and much less than WT BMP4. However, Fig. S1B does not support the conclusion on page 7 (numbering beginning with title page); "these findings suggest that phosphorylation of S91 is required to generate fully active BMP4 homodimers". The authors rightly note that the S91C change likely has manifold effects beyond inhibiting furin cleavage. The E93G change may also affect proBMP4 beyond disturbing FAM20C phosphorylation. Additional mutation analyses would strengthen the work.

      The major goal of generating and comparing the activity of the S91D mutant with S91C was to control for phosphorylation independent defects cause by the deleterious introduction of a cysteine residue, which might cause aberrant disulfide bonding. We opted to introduce S91D since “phosphomimics” can sometimes approximate the phosphorylated state. S91D has significantly higher activity than S91C (p<0.01) and has a less significant loss of activity (p<0.05) than does S91C (<p<0.0001) relative to wild type BMP4 (Fig. S1), consistent with deleterious effects of the cysteine residue and supporting a possible explanation for the more severe phenotype of S91C vs E93G mice. We have rewritten this section to clarify our interpretation (lines 165-174)and have changed our statement that our activity data “suggest the importance of phosphorylation” to a statement that they are consistent with this possibility (lines 179-180). We do not believe that further mutational analysis using activity assays in Xenopus would shed light on how or whether phosphorylation affects proteolytic activation of BMP4.

      (2) These findings in Figure S1 are potentially significant because they may inform how proBMP4 is protected from cleavage during transit through the TGN and entry into peripheral cellular compartments. Intriguing modeling studies in Figure 6 suggest that pSer91 is proximal to the furin cleavage site. Based on their presentation, pSer91 may contact Arg289, the critical P4 residue at the furin site. If so, might that suggest how pS91 may prevent furin cleavage, thus explaining why the S91D mutation inhibits processing as presented, and possibly how proBMP4 processing is delayed until transit to distal compartments (perhaps activated by a change in the endosomal microenvironment or a Ser91 phosphatase)? Have the authors considered or ruled out these possibilities? In addition to additional mutation analyses of the FAM20C site, moving the discussion of this model to an "Ideas and Speculation" subsection may be warranted.

      The model shown in Fig. 6B proposes the possibility that phosphorylation unmasks (rather than preventing) the furin cleavage motif due to the proximity of Ser91 to the cleavage site (lines 399-402). If S91D truly mimicked phosphorylation, we would predict it would facilitate processing rather than inhibiting it. We do not have data comparing cleavage of S91D relative to wild type BMP4 and have not generated knock in S91D mice to test this idea. While the reviewers questions are intriguing, they cannot be answered by mutational analysis of the FAM20C site and are beyond the scope of the current studies that sought to understand the impact of human pS91C and pE93G mutations and cell biological implications. We have moved the models to an “Ideas and Speculation” subsection as suggested (lines 377-414) since these models are meant to provoke further thought rather than provide definitive answers based on our data.

      (3) The lack of an in vitro protease assay to test the effect of the S91 mutations on furin cleavage is problematic.

      Although we routinely perform in vitro cleavage assays with recombinant furin, we don’t believe they would be informative on how S91 phosphorylation or mutation of this residue impacts cleavage since in vitro synthesized substrate used in these assays is neither dimerized not post-translationally modified, and cleavage would be tested in isolation from the endogenous trafficking environment that we propose influences cleavage.

      Reviewer #2 (Recommendations for the authors):

      (1) The impact of BMPS91A should be determined and paired with the S91D phosphomimic data to reveal if it causes proBMP4 to be cleaved prematurely and disturbs pSmad1 expression. Data for S93G should also be included.

      Our major goal in comparing the activity of S91D with S91C was to control for phosphorylation independent defects cause by the deleterious introduction of a cysteine residue in S91C, which might cause aberrant disulfide bonding. We opted to introduce S91D since “phosphomimics” can sometimes approximate the phosphorylated state. We note that S91D has significantly higher activity than S91C, consistent with deleterious effects of the cysteine residue and supporting a possible explanation for the more severe phenotype of S91C vs E93G mice. We have revised the wording of this section to clarify this. Our models predict that S91D would be cleaved more efficiently than S91C or S91A, if it really mimics the endogenous phosphorylated state, rather than being cleaved prematurely. Our biochemical analysis compares cleavage of endogenous BMP4 in wild type and mutant MEFs. Generation of S91D, S91A or S93G mutant mice to compare cleavage is beyond the scope of the current work.

      (2) Is the distance between pS91 and Arg289 close enough to form a hydrogen bond? If so, might this interaction influence furin access?

      AI modeling does not provide high probability prediction of structures surrounding the furin motif (see Fig. S7) and thus we cannot comment on whether or not these residues are close enough to form a hydrogen bond. We have revised the wording of the discussion to state “This simple model building indicates the possibility of direct contact between pSer91 and Arg289, and that phosphorylation is required for furin to access the cleavage site, although we note that predictions surrounding the furin motif represent low probability conformations (Fig. S7) (lines 399-402).”

      (3) The genotypes in Figure 2 are labeled awkwardly. Consider labeling the headers for the three subsections of panels (A-F, G-L, and M-O) differently.

      We have revised Fig. 2 to clarify that the three subsections of panels are distinct, and to emphasize that the middle subsection represents views of the right and left side of the same embryo.

      (4) The tables should be reformatted. As is, the labeling is frequently cut off, and the numbers of expected and observed progeny should both be stated to aid the reader.

      We thank the reviewer for noting the formatting errors in the tables, which we have corrected. We have also changed the tables so that normal or abnormal mendelian distributions are reported as numbers of observed/expected progeny rather than numbers/percent observed progeny.

      Reviewer #3 (Public review):

      Summary:

      The authors describe important new biochemical elements in the synthesis of a class of critical developmental signaling molecules, BMP4. They also present a highly detailed description of developmental anomalies in mice bearing known human mutations at these specific elements.

      Strengths:

      Exceptionally detailed descriptions of pathologies occurring in mutant mice. Novel findings regarding the interaction of propeptide phosphorylation and convertase cleavage, both of which will move the field forward. Provocative hypothesis regarding furin access to cleavage sites, supported by Alphafold predictions.

      Weaknesses:

      Figure 6A presents two testable models for pre-release access of furin to cleavage sites since physical separation of enzyme from substrate only occurs in one model; could immunocytochemistry resolve?

      Available reagents are not sensitive enough to detect endogenous furin and BMP4 with high resolution. Because PC/substrate interactions are transient, whereas the bulk of furin and BMP4 is distributed throughout the secretory pathway, it is not possible to co-immunolocalize furin and BMP4 in vivo at present. Studies using more advanced cell biological techniques such along with tagged proteins may enable us to test these hypotheses in the future.

      Reviewer #3 (Recommendations for the authors):

      This interesting paper presents new data on an important family of developmental signaling molecules, BMPs. Mutations at FAM20C consensus sites within BMP prodomains are known to cause birth defects. The authors have here explored differential effects of human mutations on hetero- and homodimer activity and maturation, issues that may well arise during human development. In addition to demonstrating the profound effect of these mutations on development in Xenopus and mice, the authors also show differential processing of BMP4 precursors bearing these mutations in MEF cells prepared from mutant embryos. Finally, they show that FAM20C plays a role in BMP4 prodomain processing with quite differing outcomes in homo- vs heterodimers, which they suggest is due to structural differences impacting furin access. While this latter idea remains speculative due to the lack of crystal structures (models are based on Alphafold) it is a highly promising line of work.

      The data are beautifully presented and will be of clear interest to all developmental biologists. Certain cell biology results may also extrapolate to other phosphorylated precursor molecules undergoing the interesting (and as yet unexplained) phenomenon of convertase cleavage immediately before secretion, for example, FGF23. I have only a few minor comments regarding the presentation, which is remarkably clear.

      (1) The introduction of BMP7 in the Abstract is abrupt. It should be described as a preferred dimerization partner for BMP4.

      Thank you for noting this. We have revised the first sentence of the abstract to better introduce BMP7(lines 49-50).

      (2) In Figure 1A, what is the small light green box?

      This is a small fragment released from the prodomain by the second cleavage. We have clarified this in the introduction (lines 112-114) and in the legend to Figure 1 (lines 758-759).

      (3) In the Discussion it might be relevant to mention that FAM20C propeptide is not cleaved by convertases but by S1P (Chen 2021).

      We have added this information to clarify (lines 394-396).

      (4) Figure 3, define VSD; Figure 5, Endo H removes sugars only from immature (nonsialylated) sugars, not from all chains as implied. More importantly, EndoH and PNGase remove N-linked sugars, yet Results refer only to O-linked glycosylation.

      Thank you for noting these oversights. We have defined VSD in Figure 3. We have also revised the headers for Fig. 5 and for the relevant subsection of the results to include N-linked glycosylation and note in the results that EndoH removes only immature N-linked carbohydrates (lines 301-304).

      (5) Figure 5- for clarity, I suggest it be broken up into two larger panels labeled "Embryos" and "MEFs"

      Thank you for this suggestion, we have subdivided the Figure into two panels.

      (6) Figure 6A presents two testable models for pre-release access of furin to cleavage sites since the physical separation of the enzyme from substrate only occurs in one model; could confocal immunocytochemistry resolve?

      Available reagents are not sensitive enough to detect endogenous furin and BMP4 with high resolution and PC/substrate interactions are transient whereas the bulk of both furin and BMP4 is in transit through the secretory pathway. For these reasons it is not possible to co-immunolocalize furin and BMP4 in vivo. Future studies using advanced cell biological techniques may enable us to test these hypotheses in the future.

    1. eLife Assessment

      This fundamental article significantly advances our understanding of FGF signalling, and in particular highlights the complex modifications affecting this pathway. The evidence for the authors' claims is convincing, combining state of the art conditional gene deletion in the mouse lens with histological and molecular approaches. This work should be of great interest to molecular and developmental biologists beyond the lens community.

    2. Reviewer #2 (Public review):

      Summary:

      In this paper, the authors first examined lens phenotypes in mice with Le-Cre-mediated knockdown (KD) of all the four FGFR (FGFR1-4), and found that pERK signals, Jag1 and foxe3 expression are absent or drastically reduced, indicating that FGF signaling is essential for lens induction. Next, the authors examined lens phenotypes of FGFR1/2-KD mice and found that lens fiber differentiation is compromised, and that proliferative activity and cell survival are also compromised in lens epithelium. Interestingly, Kras activation rescues defects in lens growth and lens fiber differentiation in FGFR1/2-KD mice, indicating that Ras activation is a key step for lens development, downstream of FGF signaling. Next, the authors examined the role of Frs2, Shp2 and Grb2 in FGF signaling for lens development. They confirmed that lens fiber differentiation is compromised in FGFR1/3-KD mice combined with Frs2-dysfunctional FGFR2 mutants, which is similar to lens phenotypes of Grb2-KD mice. However, lens defects are milder in mice with Shp2YF/YF and Shp2CS mutant alleles, indicating that involvement of Shp2 is limited for the Grb2 recruitment for lens fiber differentiation. Lastly, the authors showed new evidence on the possibility that another adapter protein, Shc1, promotes Grb2 recruitment independent of Frs2/Shp2-mediated Grb2 recruitment.

      Strengths:

      Overall, the manuscript provides valuable data on how FGFR activation leads to Ras activation through the adapter platform of Frs2/Shp2/Grb2, which advances our understanding on complex modification of FGF signaling pathway. The authors applied a genetic approach using mice, whose methods and results are valid to support the conclusion. The discussion also well summarizes the significance of their findings.

      Weaknesses:

      The authors found that the new adaptor protein Shc1 is involved in Grb2 recruitment in response to FGF receptor activation. However, the main data on Shc1 are only histological sections and statistical evaluation of lens size. Cellular-level evidence on Shc1 makes the authors' conclusion more convincing.

      Comments on latest version:

      In the 2nd revised version of the manuscript, the authors responded to my recommendation to show the number of biological replicates for Prox1 and αA-crystallin (Fig. 1F) and conductedstatistical analysis for pmTOR, and pS6 (Supplementary figure 1B).

      The authors also explained why the animals are no longer available for the additional experiments that I requested. I may understand the situation, but hope that the authors will investigate the cellular-level evidence on Shc1 in more detail and report it maybe as another paper in future.

    3. Author response:

      The following is the authors’ response to the previous reviews

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      This manuscript uses the eye lens as a model to investigate basic mechanisms in the Fgf signaling pathway. Understanding Fgf signaling is of broad importance to biologists as it is involved in the regulation of various developmental processes in different tissues/organs and is often misregulated in disease states. The Fgf pathway has been studied in embryonic lens development, namely with regards to its involvement in controlling events such as tissue invagination, vesicle formation, epithelium proliferation and cellular differentiation, thus making the lens a good system to uncover the mechanistic basis of how the modulation of this pathway drives specific outcomes. Previous work has suggested that proteins, other than the ones currently known (e.g., the adaptor protein Frs2), are likely involved in Fgfr signaling. The present study focuses on the role of Shp2 and Shc1 proteins in the recruitment of Grb2 in the events downstream of Fgfr activation.

      Strengths:

      The findings reveal that the juxtamembrane region of the Fgf receptor is necessary for proper control of downstream events such as facilitating key changes in transcription and cytoskeleton during tissue morphogenesis. The authors conditionally deleted all four Fgfrs in the mouse lens that resulted in molecular and morphological lens defects, most importantly, preventing the upregulation of the lens induction markers Sox2 and Foxe3 and the apical localization of F-actin, thus demonstrating the importance of Fgfrs in early lens development, i.e. during lens induction. They also examined the impact of deleting Fgfr1 and 2, on the following stage, i.e. lens vesicle development, which could be rescued by expressing constitutively active KrasG12D. By using specific mutations (e.g. Fgfr1ΔFrs lacking the Frs2 binding domain and Fgfr2LR harboring mutations that prevent binding of Frs2), it is demonstrated that the Frs2 binding site on Fgfr is necessary for specific events such as morphogenesis of lens vesicle. Further, by studying Shp2 mutations and deletions, the authors present a case for Shp2 protein to function in a context-specific manner in the role of an adaptor protein and a phosphatase enzyme. Finally, the key surprising finding from this study is that downstream of Fgfr signaling, Shc1 is an important alternative pathway - in addition to Shp2 - involved in the recruitment of Grb2 and in the subsequent activation of Ras. The methodologies, namely, mouse genetics and state-of-the-art cell/molecular/biochemical assays are appropriately used to collect the data, which are soundly interpreted to reach these important conclusions. Overall, these findings reveal the flexibility of the Fgf signaling pathway and it downstream mediators in regulating cellular events. This work is expected to be of broad interest to molecular and developmental biologists.

      Weaknesses:

      A weakness that needs to be discussed is that Le-Cre depends on Pax6 activation, and hence its use in specific gene deletion will not allow evaluation of the requirement of Fgfrs in the expression of Pax6 itself. But since this is the earliest Cre available for deletion in the lens, mentioning this in the discussion would make the readers aware of this issue.

      Reviewer #2 (Public review):

      Summary

      I have reviewed the revised manuscript submitted by Wang et al., which is entitled "Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development". In this paper, the authors first examined lens phenotypes in mice with Le-Cre-mediated knockdown (KD) of all four FGFR (FGFR1-4), and found that pERK signals, Jag1 and foxe3 expression are absent or drastically reduced, indicating that FGF signaling is essential for lens induction. Next, the authors examined lens phenotypes of FGFR1/2-KD mice and found that lens fiber differentiation is compromised and that proliferative activity and cell survival are also compromised in lens epithelium. Interestingly, Kras activation rescues defects in lens growth and lens fiber differentiation in FGFR1/2-KD mice, indicating that Ras activation is a key step for lens development, downstream of FGF signaling. Next, the authors examined the role of Frs2, Shp2 and Grb2 in FGF signaling for lens development. They confirmed that lens fiber differentiation is compromised in FGFR1/3-KD mice combined with Frs2-dysfunctional FGFR2 mutants, which is similar to lens phenotypes of Grb2-KD mice. However, lens defects are milder in mice with Shp2YF/YF and Shp2CS mutant alleles, indicating that involvement of Shp2 is limited for the Grb2 recruitment for lens fiber differentiation. Lastly, the authors showed new evidence on the possibility that another adapter protein, Shc1, promotes Grb2 recruitment independent of Frs2/Shp2-mediated Grb2 recruitment.

      Strength

      Overall, the manuscript provides valuable data on how FGFR activation leads to Ras activation through the adapter platform of Frs2/Shp2/Grb2, which advances our understanding on complex modification of FGF signaling pathway. The authors applied a genetic approach using mice, whose methods and results are valid to support the conclusion. The discussion also well summarizes the significance of their findings.

      Weakness

      The authors found that the new adaptor protein Shc1 is involved in Grb2 recruitments in response to FGF receptor activation. However, the main data on Shc1 are only histological sections and statistical evaluation of lens size. In the revised manuscript, the authors did not answer my major concern that cellular-level data are missing, which is not fully enough to support their main conclusion on the involvement of Shc1 in Grb2 recruitment of FGF signaling for lens development. Since the title of this manuscript is that Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development, it is important to provide the cellular-level evidence on Shc1.

      Reviewer #3 (Public review):

      Summary:

      The manuscript entitled "Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development" by Wang et al., investigates the molecular mechanism used by FGFR signaling to support lens development. The lens has long been known to depend on FGFR-signaling for proper development. Previous investigations have demonstrated the FGFR signaling is required for embryonic lens cell survival and for lens fiber cell differentiation. The requirement of FGFR signaling for lens induction has remained more controversial as deletion of both Fgfr1 and Fgfr2 during lens placode formation does not prevent the induction of definitive lens markers such as FOXE3 or αA-crystallin. Here the authors have used the Le-Cre driver to delete all four FGFR genes from the developing lens placode demonstrating a definitive failure of lens induction in the absence of FGFR-signaling. The authors focused on FGFR1 and FGFR2, the two primary FGFRs present during early lens development and demonstrated that lens development could be significantly rescued in lenses lacking both FGFR1 and FGFR2 by expressing a constitutively active allele of KRAS. They also showed that the removal of pro-apoptotic genes Bax and Bak could also lead to a substantial rescue of lens development in lenses lacking both FGFR1 and FGFR2. In both cases, the lens rescue included both increased lens size and the expression of genes characteristic of lens cells.

      Significantly the authors concentrated on the juxtamembrane domain, a portion of the FGFRs associated with FRS2. Previous investigations have demonstrated the importance of FRS2 activation for mediating a sustained level of ERK activation. FRS2 is known to associate both with GRB2 and SHP2 to activate RAS. The authors utilized a mutant allele of Fgfr1, lacking the entire juxtamembrane domain (Fgfr1ΔFrs) and an allele of Fgfr2 containing two-point mutations essential for Frs2 binding (Fgfr2LR). When combining three floxed alleles and leaving only one functional allele (Fgfr1ΔFrs or Fgfr2LR) the authors got strikingly different phenotypes. When only the Fgfr1ΔFrs allele was retained, the lens phenotype matched that of deleting both Fgfr1 and Fgfr2. However, when only the Fgfr2LR allele was retained the phenotype was significantly milder, primarily affecting lens fiber cell differentiation, suggesting that something other than FRS2 might be interacting with the juxtamembrane domain to support FGFR signaling in the lens. The authors also deleted Grb2 in the lens and showed that the phenotype was similar to that of the lenses only retaining the Fgfr2LR allele, resulting a failure of lens fiber cell differentiation and decreased lens cell survival. However, mutating the major tyrosine phosphorylation site of GRB2 did not affect lens development. The authors additionally investigated the role of SHP2 in lens development by either deleting SHP2 or by making mutations in the SHP2 catalytic domain. The deletion of the SHP2 phosphatase activity did not affect lens development as severely as total loss of SHP2 protein, suggesting a function for SHP2 outside of its catalytic activity. Although the loss of Shc1 alone has only a slight effect on lens size and pERK activation in the lens, the authors showed that the loss of Shc1 exacerbated the lens phenotype in lenses lacking both Frs2 and Shp2. The authors suggest that SHC1 binds to the FGFR juxtamembrane domain allowing for the recruitment of GRB2 in independently of FRS2.

      Strengths:

      (1) The authors used a variety of genetic tools to carefully dissect the essential signals downstream of FGFR signaling during lens development.

      (2) The authors made a convincing case that something other than FRS2 binding mediates FGFR signaling in the juxtamembrane domain.

      (3) The authors demonstrated that despite the requirement of both the adaptor function and phosphatase activity of SHP2 are required for embryonic survival, neither of these activities is absolutely required for lens development.

      (4) The authors provide more information as to why FGFR loss has a phenotype much more severe than the loss of FRS2 alone during lens development.

      (5) The authors followed up their work analyzing various signaling molecules in the context of lens development with biochemical analyses of FGF-induced phosphorylation in murine embryonic fibroblasts (MEFs).

      (6) In general, this manuscript represents a Herculean effort to dissect FGFR signaling in vivo with biochemical backing with cell culture experiments in vitro.

      Weaknesses:

      (1) The authors demonstrate that the loss of FGFR1 and FGFR2 can be compensated by a constitutive active KRAS allele in the lens and suggest that FGFRs largely support lens development only by driving ERK activation. However, the authors also saw that lens development was substantially rescued by preventing apoptosis through the deletion of BAK and BAX. To my knowledge, the deletion of BAK and BAX should not independently activate ERK. The authors do not show whether ERK activation is restored in the BAK/BAX deficient lenses. Do the authors suggest the FGFR3 and/or FGFR4 provide sufficient RAS and ERK activation for lens development when apoptosis is suppressed? Alternatively, is it the survival function of FGFR-signaling as much as a direct effect on lens differentiation?

      (2) Do the authors suggest that GRB2 is required for RAS activation and ultimately ERK activation? If so, do the authors suggest that ERK activation is not required for FGFR-signaling to mediate lens induction? This would follow considering that the GRB2 deficient lenses lack a problem with lens induction.

      (3) The increase in p-Shc is only slightly higher in the Cre FGFR1f/f FGFR2r/LR than in the FGFR1f/Δfrs FGFR2f/f. Can the authors provide quantification?

      (4) The authors have not shown directly that Shc1 binds to the juxtamembrane region of either Fgfr1 or Fgfr2.

      Recommendations for the authors:

      Reviewer #2 (Recommendations for the authors):

      In the revised manuscript, the authors have responded to my recommendations to revise the original manuscript, except for three suggestions below.

      (1) The original recommendation: Results (page 6, line 8): The authors mentioned "we observed .... expression of Foxe3 in ...mutant lens cells (Figure 1E, arrows). However, Foxe3-expressing lens cells are a very small population in Figure 1E. It is important to state the decreased number of Foxe3-expressing lens cells in FGFR1/2 mutants. In addition, I would like to request the authors to show histograms indicating sample size and statistical analysis for marker expression: Foxe3 (Figure 1E), Prox1 and aA-crystallin (Fig. 1F), cyclin D1 and TUNEL (Fig. 1G) and pmTOR and pS6 (Supplementary figure 1B).

      Author's response: We added a statement indicating that the number of Foxe3-expressing cells is reduced in FGFR1/2 mutants, which is now quantified in Fig. 1H. Quantifications for Cyclin D1 and TUNEL are now shown in Fig. 1I and J, respectively. However, we chose not to quantify Prox1, αA-crystallin, pmTOR, and pS6, as the FGFR1/2 mutants showed no staining for these markers.<br /> My recommendation: Although the authors have responded to revise the quantification of Foxe3-expressing cells, Cyclin D1 and TUNEL, they did not conduct statistical analysis of Prox1, αA-crystallin, pmTOR, and pS6, because of absence of these marker signals. I understand that no signal makes statistical analysis no meaningful. However, it is still important to indicate how many the authors repeated experiments to confirm the same result. Please indicate the number of biological replicates or independent experiments in the figure legends, for example "Biological replicates, n=3" or "Three independent experiments show similar results". As for pS6 labeling, there seems to be a weak signal in Supplementary Figure 1B, so please show statistical analysis to indicate its histogram.

      We have added the number of biological replicates for Prox1 and αA staining in the legend of Fig.1. The review is correct that there is weak staining of pS6, and also pmTOR. The quantification of pS6 and pmTOR staining are now shown in Supplementary Fig. 1C and D.

      (2) The original recommendation: Results (page 6, line 19- page 7, line 6): The authors showed that inducible expression of constitutive active Kras, KrasG12D, using Le-Cre, recovered lens size to the half level of wild-type control. However, in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D, pERK was detected in the most posterior edge of the lens fiber core, whereas pERK was detected in the broader area of the lens in control. Furthermore, pMEK was detected in the whole lens of mice with Le-Cre; FGFR1/2f/f; and LSL-KrasG12D, whereas pMEK was detected only in the lens epithelial cells at the equator. So, the spatial profile of pERK and pMEK expression was different from those of wild-type, although the authors observed that Prox1 and Crystallin expression are normally induced in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D. I wonder whether the lens normally develops in mice with Le-Cre; LSL-KrasG12D? Is the lens growth enhanced in mice with Le-Cre; LSL-KrasG12D? Please add the panels of mice with Le-Cre; LSL-KrasG12D in Figure 2B and 2C. In addition, I wonder whether apoptosis is suppressed in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D?

      Authors' response: Response: As we previously reported (Developmental Biology 355, 2011, 12-20), Le-Cre; LSL-KrasG12D did not lead to enhanced lens growth. While we agree that including images of Le-Cre; LSL-KrasG12D as controls in Fig. 2B and C and evaluating apoptosis in Le-Cre; FGFR1/2f/f; LSL-KrasG12D mutants would be appropriate, we regretfully no longer have these animals available to conduct these experiments.

      My recommendation: I would like to suggest the authors conduct these experiments again, because the recovery of lens formation by Bax/Bak KD in Fgfr1/2 KD mice (Fig. 2F) suggests that KrasG12D activates the AKT-mediated cell survival pathway as well as that MEK/MAPK pathway downstream of FGF signaling pathway. Regarding the availability of mouse strains, in general, it is necessary to keep animal strains available for sincere response to reviewers' suggestions. Please clarify why these strains are now not available and justify the reason in the response to reviewers' recommendations.

      We acknowledge the reviewer's suggested experiments. However, our research utilized multiple mouse strains that are costly to maintain, a challenge that was exacerbated during and after the COVID-19 pandemic. Unfortunately, we no longer have access to the specific mouse strains required to conduct these additional studies.

      (3) The original recommendation: Figures 7E, and 7F: The authors showed that lens morphology and lens size evaluation in genetic combinations: control, Frs2/Shc1 KD, Frs2/Shp2 KD, and Frs2/Shp2/Shc1 KD. However, I would like to request the authors to show more detailed data in these genetic combinations, for example, pERK, foxe3, Maf, Prox1, Jag1, p57, cyclin D3, g-crystallin, and TUNEL.

      Authors' response: Unfortunately, we no longer have these mutant mice to perform these detailed staining.

      My recommendation: As I mentioned in the statement on weakness above, it is important to provide the cellular-level evidence to support the main conclusion on the involvement of Shc1 in Grb2 recruitment of FGF signaling for lens development, because this is the main novel finding in this manuscript. Regarding the availability of mouse strains, it is generally necessary to keep animal strains available for sincere response to reviewers' suggestions. Please clarify why these strains are now not available and justify the reason in the response to the reviewers' suggestions.

      We regret that we did not anticipate these experiments suggested by the reviewer. Unfortunately, we are unable to perform these studies as we no longer maintain the required mouse strains in our colony.

      Reviewer #3 (Recommendations for the authors):

      The changes made by the authors improved the manuscript. I have no further suggestions.

    1. eLife Assessment

      This is an important study that aims to investigate the behavioral relevance of multisensory responses recorded in the auditory cortex. The experiments are elegant and well-designed and are supported by appropriate analyses of the data. Although solid evidence is presented that is consistent with learning-dependent encoding of visual information in auditory cortex, further work is needed to establish the origin and nature of these non-auditory signals and to definitively rule out any effects of movement-related activity.

    2. Reviewer #1 (Public review):

      Summary:

      Chang and colleagues use tetrode recordings in behaving rats to study how learning an audiovisual discrimination task shapes multisensory interactions in auditory cortex. They find that a significant fraction of neurons in auditory cortex responded to visual (crossmodal) and audiovisual stimuli. Both auditory-responsive and visually-responsive neurons preferentially responded to the cue signaling the contralateral choice in the two-alternative forced choice task. Importantly, multisensory interactions were similarly specific for the congruent audiovisual pairing for the contralateral side.

      Strengths:

      The experiments are conducted in a rigorous manner. Particularly thorough are the comparisons across cohorts of rats trained in a control task, in a unisensory auditory discrimination task and the multisensory task, while also varying the recording hemisphere and behavioral state (engaged vs. anesthesia). The resulting contrasts strengthen the authors' findings and rule out important alternative explanations regarding the effect of experience. Through the comparisons, they show that the enhancements of activity in multisensory trials in auditory cortex are specific to the paired audiovisual stimulus and specific to contralateral choices in correct trials and thus dependent on learned associations in a task engaged state.

      Weaknesses:

      The main result that multisensory interactions are specific for contralateral paired audiovisual stimuli is consistent across experiments and interpretable as a learned task-dependent effect. However, the alternative interpretation of behavioral signals is crucial to rule out, which would also be specific to contralateral, correct trials in trained animals. Although the authors focus on the first 150 ms after cue onset, some of the temporal profiles of activity suggest that choice-related activity could confound some of the results.

      The main concern (noted by all reviewers) is the interpretation of the evoked activity in visual trials. In the revised manuscript, the authors have not provided much data to disentangle movement related activity from sensory related activity. The only new data is on the visual response dynamics in supplementary figure 2, which is unconvincing both in terms of visual response latency and response dynamics. Therefore, the response of the authors has been insufficient to prove the visual nature of the evoked responses.

      In this supplemental figure 2 the same example neuron as in the original manuscript is shown again as well as the average z-scored visual response. First, the visual response latency is inconsistent with literature. The first evoked activity in mouse V1 (!) is routinely reported around 50 ms (for example, 45 ms in Niell Stryker 2008, 52 ms, Schnabel et al. 2018, 54 ms in Oude Lohuis et al. 2024). According to the authors the potential route of crossmodal modulation of AC can occur through either corticocortical connections (which will impose further polysynaptic delays - monosynaptic projection from dLGN or V1 incredibly sparse), or through pulvinar (but pulvinar visual responses are much later (they find 170 vs 80 ms in dLGN, Roth et al. 2019) as expected from a higher order thalamic nucleus). One can also critique the estimation of the response latency which depends on the signal strength (visual response is smaller) and thus choice of threshold. With a different arbitrary threshold one would come to different conclusions.

      Second, the temporal response dynamics to visual input are the same as the auditory response. It can be observed that if the data were normalized by the max response the dynamics are very similar, with the response back to near baseline levels at 100 ms post stimulus. I am not aware of publications that have observed response dynamics that are similar between A and V stimuli, nor such short-lasting visual response. In the visual system, mean activity typically drops again around 150-200ms.<br /> With the nature of the observed activity unclear, careful interpretation is warranted about audiovisual interactions in auditory cortex.

    3. Reviewer #2 (Public review):

      In this revision the authors have made a solid effort to address each of the points raised by all three reviewers. Due to the fact that animals in this study were freely moving, and there has not been any high-speed video recordings to measure whisker movements or other possible stimulus-induced motor effects it is still not possible to rule out motor effects completely. However, the fact that the multisensory enhancements are stimulus specific, much stronger in the multisensory case than the visual only condition, and short in latency it does seem the most parsimonious explanation is likely that these responses are visual in nature.

      The delayed auditory stimulus offers some explanation for the very small latency difference between audio and visual stimulus elements. Studies using LED flashes in rat V2 report latencies around ~50 ms (e.g. 2017 paper from Brian Allman's group). The response latencies for visual stimuli in this manuscript are of this order of magnitude, albeit still shorter than that (which presumably means they don't originate from V2).

      There are still parts of the manuscript that are inappropriately causal - e.g. line 283 "this suggests that strong multisensory integration is critical for behavior" - it could just as well be the case that high attention / motivation / arousal leads to both strong integration and good behavior.

    4. Reviewer #3 (Public review):

      Summary:

      The manuscript by Chang et al. aims to investigate how the behavioral relevance of auditory and visual stimuli influences the way in which the primary auditory cortex encodes auditory, visual and audiovisual information. The main results is that behavioral training induces an increase in the encoding of auditory and visual information and in multisensory enhancement that is mainly related to the choice located contralaterally with respect to the recorded hemisphere.

      Strengths:

      The manuscript reports the results of an elegant and well planned experiment meant to investigate if auditory cortex encodes visual information and how learning shapes visual responsiveness in auditory cortex. Analyses are typically well done and properly address the questions raised

      Weaknesses:

      The authors have addressed most of my comments satisfactorily. However, I am still not convinced by the authors' claim that the use of LED should lead to visually-evoked responses with faster dynamics compared to the use of normal screens. In fact, previous studies using screen-emitted flashed did not report such faster dynamics. Visually-evoked responses in V1 (which are expected to occur earlier than A1) typically do not show onset latencies faster than 40 ms, and have a peak latency of about 100-120 ms. The dynamics shown in the new supplementary Fig. 2 are still faster than this, and thus should be explained. The authors' claim is in fact not supported by cited literature. The authors should at least provide evidence that a similar effect has been observed previously, or otherwise collect evidence themselves. In the absence of such evidence, I remain dubious about the visual nature of the observed activity, especially since, in contrast with what the authors say elsewhere in the rebuttal, involuntary motor reaction to (at least auditory) stimuli can be extremely fast (<40 ms; Clayton et al. 2024) and might thus potentially, at least partially, explain the observed "visual" response.

    5. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1 (Public review):

      Summary:

      Chang and colleagues used tetrode recordings in behaving rats to study how learning an audiovisual discrimination task shapes multisensory interactions in the auditory cortex. They found that a significant fraction of neurons in the auditory cortex responded to visual (crossmodal) and audiovisual stimuli. Both auditory-responsive and visually-responsive neurons preferentially responded to the cue signaling the contralateral choice in the two-alternative forced choice task. Importantly, multisensory interactions were similarly specific for the congruent audiovisual pairing for the contralateral side.

      Strengths:

      The experiments were conducted in a rigorous manner. Particularly thorough are the comparisons across cohorts of rats trained in a control task, in a unisensory auditory discrimination task, and the multisensory task, while also varying the recording hemisphere and behavioral state (engaged vs. anesthesia). The resulting contrasts strengthen the authors' findings and rule out important alternative explanations. Through the comparisons, they show that the enhancements of multisensory responses in the auditory cortex are specific to the paired audiovisual stimulus and specific to contralateral choices in correct trials and thus dependent on learned associations in a task-engaged state.

      We thank Reviewer #1 for the thorough review and valuable feedback.

      Weaknesses:

      The main result is that multisensory interactions are specific for contralateral paired audiovisual stimuli, which is consistent across experiments and interpretable as a learned task-dependent effect. However, the alternative interpretation of behavioral signals is crucial to rule out, which would also be specific to contralateral, correct trials in trained animals. Although the authors focus on the first 150 ms after cue onset, some of the temporal profiles of activity suggest that choice-related activity could confound some of the results.

      We thank the reviewer for raising this important point regarding the potential influence of choice-related activity on our results. In our experimental setup, it is challenging to completely disentangle the effects of behavioral choice from multisensory interaction. However, we conducted relevant analyses to examine the influence of choice-related components on multisensory interaction.

      First, we analyzed neural responses during incorrect trials and found a significant reduction in multisensory enhancement for the A<sup>10k</sup>-V<sup>vt</sup> pairing (Fig. 4). In contrast, for the A<sup>3k</sup>-V<sup>hz</sup> pairing, there was no strong multisensory interaction during either correct (right direction) or incorrect (left direction) choices. This finding suggests that the observed multisensory interactions are strongly associated with specific cue combinations during correct task performance.

      Second, we conducted experiments with unisensory training, in which animals were trained separately on auditory and visual discriminations without explicit multisensory associations. The results demonstrated that unisensory training did not lead to the development of selective multisensory enhancement or congruent auditory-visual preferences, as observed in the multisensory training group. This indicates that the observed multisensory interactions in the auditory cortex are specific to multisensory training and cannot be attributed solely to behavioral signals or choice-related effects.

      Finally, we specifically focused on the early 0-150 ms time window after cue onset in our main analyses to minimize contributions from motor-related or decision-related activity, which typically emerge later. This time window allowed us to capture early sensory processing while reducing potential confounds.

      Together, these findings strongly suggest that the observed choice-dependent multisensory enhancement is a learned, task-dependent phenomenon that is specific to multisensory training.

      The auditory stimuli appear to be encoded by short transient activity (in line with much of what we know about the auditory system), likely with onset latencies (not reported) of 15-30 ms. Stimulus identity can be decoded (Figure 2j) apparently with an onset latency around 50-75 ms (only the difference between A and AV groups is reported) and can be decoded near perfectly for an extended time window, without a dip in decoding performance that is observed in the mean activity Figure 2e. The dynamics of the response of the example neurons presented in Figures 2c and d and the average in 2e therefore do not entirely match the population decoding profile in 2j. Population decoding uses the population activity distribution, rather than the mean, so this is not inherently problematic. It suggests however that the stimulus identity can be decoded from later (choice-related?) activity. The dynamics of the population decoding accuracy are in line with the dynamics one could expect based on choice-related activity. Also the results in Figures S2e,f suggest differences between the two learned stimuli can be in the late phase of the response, not in the early phase.

      We appreciate the reviewer’s detailed observations and questions regarding the dynamics of auditory responses and decoding profiles in our study. In our experiment, primary auditory cortex (A1) neurons exhibited short response latencies that meet the established criteria for auditory responses in A1, consistent with findings from many other studies conducted in both anesthetized and task-engaged animals. While the major responses typically occurred during the early period (0-150ms) after cue onset (see population response in Fig. 2e), individual neuronal responses in the whole population were generally dynamic, as illustrated in Figures 2c, 2d, and 3a–c. As the reviewer correctly noted, population decoding leverages the distribution of activity across neurons rather than the mean activity, which explains why the dynamics of population decoding accuracy align well with choice-related activity. This also accounts for the extended decoding window observed in Figure 2j, which does not entirely match the early population response profiles in Figure 2e.

      To address the reviewer’s suggestion that differences between the two learned stimuli might arise in the late phase of the response, we conducted a cue selectivity analysis during the 151–300 ms period after cue onset. The results, shown below, indicate that neurons maintained cue selectivity in this late phase for each modality (Supplementary Fig. 5), though the selectivity was lower than in the early phase. However, interpreting this late-phase activity remains challenging. Since A<sup>3k</sup>, V<sup>hz</sup>, and A<sup>3k</sup>-V<sup>hz</sup> were associated with the right choice, and A<sup>10k</sup>, V<sup>vt</sup>, and A<sup>10k</sup>-V<sup>vt</sup> with the left choice, it is difficult to disentangle whether the responses reflect choice, sensory features, or a combination of both.

      To further investigate, we examined multisensory interactions during the late phase, controlling for choice effects by calculating unisensory and multisensory responses within the same choice context. Our analysis revealed no evident multisensory enhancement for any auditory-visual pairing, nor significant differences between pairings—unlike the robust effects observed in the early phase (Supplementary Fig. 5). We hypothesize that early responses are predominantly sensory-driven and exhibit strong multisensory integration, whereas late responses likely reflect task-related, choice-related, or combined sensory-choice activity, where sensory-driven multisensory enhancement is less prominent. As the focus of this manuscript is on multisensory integration and cue selectivity, we prioritized a detailed analysis of the early phase, where these effects are most prominent. However, the complexity of interpreting late-phase activity remains a challenge and warrants further investigation. We cited Supplementary Fig. 5 in revised manuscript as the following:

      “This resulted in a significantly higher mean MSI for the A<sup>10k</sup>-V<sup>vt</sup> pairing compared to the A<sup>3k</sup>-V<sup>hz</sup> pairing (0.047 ± 0.124 vs. 0.003 ± 0.096; paired t-test, p < 0.001). Among audiovisual neurons, this biasing is even more pronounced (enhanced vs. inhibited: 62 vs. 2 in A<sup>10k</sup>-V<sup>vt</sup> pairing, 6 vs. 13 in A<sup>3k</sup>-V<sup>hz</sup> pairing; mean MSI: 0.119±0.105 in A<sup>10k</sup>-V<sup>vt</sup> pairing vs. 0.020±0.083 A<sup>3k</sup>-V<sup>hz</sup> pairing, paired t-test, p<0.00001) (Fig. 3f). Unlike the early period (0-150ms after cue onset), no significant differences in multisensory integration were observed during the late period (151-300ms after cue onset) (Supplementary Fig. 5).”

      First, it would help to have the same time axis across panels 2,c,d,e,j,k. Second, a careful temporal dissociation of when the central result of multisensory enhancements occurs in time would discriminate better early sensory processing-related effects versus later decision-related modulations.

      Thank you for this valuable feedback. Regarding the first point, we used a shorter time axis in Fig. 2j-k to highlight how the presence of visual cues accelerates the decoding process. This visualization choice was intended to emphasize the early differences in processing speed. For the second point, we have carefully analyzed multisensory integration across different temporal windows. The results presented in the Supplementary Fig. 5 (also see above) already address the late phase, where our data show no evidence of multisensory enhancement for any auditory-visual pairings. This distinction helps clarify that the observed multisensory effects are primarily related to early sensory processing rather than later decision-related modulations. We hope this addresses the concerns raised and appreciate the opportunity to clarify these points.

      In the abstract, the authors mention "a unique integration model", "selective multisensory enhancement for specific auditory-visual pairings", and "using this distinct integrative mechanisms". I would strongly recommend that the authors try to phrase their results more concretely, which I believe would benefit many readers, i.e. selective how (which neurons) and specific for which pairings?

      We appreciate the reviewer’s suggestion to clarify our phrasing for better accessibility. To address this, we have revised the relevant sentence in the abstract as follows:

      "This model employed selective multisensory enhancement for the auditory-visual pairing guiding the contralateral choice, which correlated with improved multisensory discrimination."

      Reviewer #2 (Public review):

      Summary

      In this study, rats were trained to discriminate auditory frequency and visual form/orientation for both unisensory and coherently presented AV stimuli. Recordings were made in the auditory cortex during behaviour and compared to those obtained in various control animals/conditions. The central finding is that AC neurons preferentially represent the contralateral-conditioned stimulus - for the main animal cohort this was a 10k tone and a vertically oriented bar. Over 1/3rd of neurons in AC were either AV/V/A+V and while a variety of multisensory neurons were recorded, the dominant response was excitation by the correctly oriented visual stimulus (interestingly this preference was absent in the visual-only neurons). Animals performing a simple version of the task in which responses were contingent on the presence of a stimulus rather than its identity showed a smaller proportion of AV stimuli and did not exhibit a preference for contralateral conditioned stimuli. The contralateral conditioned dominance was substantially less under anesthesia in the trained animals and was present in a cohort of animals trained with the reverse left/right contingency. Population decoding showed that visual cues did not increase the performance of the decoder but accelerated the rate at which it saturated. Rats trained on auditory and then visual stimuli (rather than simultaneously with A/V/AV) showed many fewer integrative neurons.

      Strengths

      There is a lot that I like about this paper - the study is well-powered with multiple groups (free choice, reversed contingency, unisensory trained, anesthesia) which provides a lot of strength to their conclusions and there are many interesting details within the paper itself. Surprisingly few studies have attempted to address whether multisensory responses in the unisensory cortex contribute to behaviour - and the main one that attempted to address this question (Lemus et al., 2010, uncited by this study) showed that while present in AC, somatosensory responses did not appear to contribute to perception. The present manuscript suggests otherwise and critically does so in the context of a task in which animals exhibit a multisensory advantage (this was lacking in Lemus et al.,). The behaviour is robust, with AV stimuli eliciting superior performance to either auditory or visual unisensory stimuli (visual were slightly worse than auditory but both were well above chance).

      We thank the reviewer for their positive evaluation of our study.

      Weaknesses

      I have a number of points that in my opinion require clarification and I have suggestions for ways in which the paper could be strengthened. In addition to these points, I admit to being slightly baffled by the response latencies; while I am not an expert in the rat, usually in the early sensory cortex auditory responses are significantly faster than visual ones (mirroring the relative first spike latencies of A1 and V1 and the different transduction mechanisms in the cochlea and retina). Yet here, the latencies look identical - if I draw a line down the pdf on the population level responses the peak of the visual and auditory is indistinguishable. This makes me wonder whether these are not sensory responses - yet, they look sensory (very tightly stimulus-locked). Are these latencies a consequence of this being AuD and not A1, or ... ? Have the authors performed movement-triggered analysis to illustrate that these responses are not related to movement out of the central port, or is it possible that both sounds and visual stimuli elicit characteristic whisking movements? Lastly, has the latency of the signals been measured (i.e. you generate and play them out synchronously, but is it possible that there is a delay on the audio channel introduced by the amp, which in turn makes it appear as if the neural signals are synchronous? If the latter were the case I wouldn't see it as a problem as many studies use a temporal offset in order to give the best chance of aligning signals in the brain, but this is such an obvious difference from what we would expect in other species that it requires some sort of explanation.

      Thank you for your insightful comments. I appreciate the opportunity to clarify these points and strengthen our manuscript. Below, I address your concerns in detail:

      We agree that auditory responses are typically faster than visual responses due to the distinct transduction mechanisms. However, in our experiment, we intentionally designed the stimulus setup to elicit auditory and visual responses within a similar time window to maximize the potential for multisensory integration. Specifically, we used pure tone sounds with a 15 ms ramp and visual stimuli generated by an LED array, which produce faster responses compared to mostly used light bars shown on a screen (see Supplementary Fig. 2a). The long ramp of the auditory stimulus slightly delayed auditory response onset, while the LED-generated bar (compared to the bar shown on the screen) elicited visual responses more quickly. This alignment likely facilitated the observed overlap in response latencies.

      Neurons’ strong spontaneous activity in freely moving animals complicates the measurement of first spike latencies. Despite that, we still can infer the latency from robust cue-evoked responses. Supplementary Fig. 2b illustrates responses from an exemplar neuron (the same neuron as shown in Fig. 2c), where the auditory response begins 9 ms earlier than the visual response. Given the 28 ms auditory response latency observed here using 15 ms-ramp auditory stimulus, this value is consistent with prior studies in the primary auditory cortex usually using 5 ms ramp pure tones, where latencies typically range from 7 to 28 ms. Across the population (n=559), auditory responses consistently reached 0.5 of the mean Z-scored response 15 ms earlier than visual responses (Supplementary Fig. 2c). The use of Gaussian smoothing in PSTHs supports the reliability of using the 0.5 threshold as an onset latency marker. We cited Supplementary Fig. 2 in the revised manuscript within the Results section (also see the following):

      “This suggests multisensory discrimination training enhances visual representation in the auditory cortex. To optimize the alignment of auditory and visual responses and reveal the greatest potential for multisensory integration, we used long-ramp pure tone auditory stimuli and quick LED-array-elicited visual stimuli (Supplementary Fig. 2). While auditory responses were still slightly earlier than visual responses, the temporal alignment was sufficient to support robust integration.”

      We measured the time at which rats left the central port and confirmed that these times occur significantly later than the neuronal responses analyzed (see Fig. 1c-d). While we acknowledge the potential influence of movements such as whiskering, facial movements, head direction changes, or body movements on neuronal responses, precise monitoring of these behaviors in freely moving animals remains a technical challenge. However, given the tightly stimulus-locked nature of the neuronal responses observed, we believe they primarily reflect sensory processing rather than movement-related activity.

      To ensure accurate synchronization of auditory and visual stimuli, we verified the latencies of our signals. The auditory and visual stimuli were generated and played out synchronously with no intentional delay introduced. The auditory amplifier used in our setup introduces minimal latency, and any such delay would have been accounted for during calibration. Importantly, even if a small delay existed, it would not undermine our findings, as many studies intentionally use temporal offsets to facilitate alignment of neural signals. Nonetheless, the temporal overlap observed here is primarily a result of our experimental design aimed at promoting multisensory integration.

      We hope these clarifications address your concerns and highlight the robustness of our findings.

      Reaction times were faster in the AV condition - it would be of interest to know whether this acceleration is sufficient to violate a race model, given the arbitrary pairing of these stimuli. This would give some insight into whether the animals are really integrating the sensory information. It would also be good to clarify whether the reaction time is the time taken to leave the center port or respond at the peripheral one.

      We appreciate your request for clarification. In our analysis, reaction time (RT) is defined as the time taken for the animal to leave the center port after cue onset. This measure was chosen because it reflects the initial decision-making process and the integration of sensory information leading to action initiation. The time taken to respond at the peripheral port, commonly referred to as movement time, was not included in our RT measure. However, movement time data is available in our dataset, and we are open to further analysis if deemed necessary.

      To determine whether the observed acceleration in RTs in the audiovisual (AV) condition reflects true multisensory integration rather than statistical facilitation, we tested for violations of the race model inequality (Miller, 1982). This approach establishes a bound for the probability of a response occurring within a given time interval under the assumption that the auditory (A) and visual (V) modalities operate independently. Specifically, we calculated cumulative distribution functions (CDFs) for the RTs in the A, V, and AV conditions (please see Author response image 1). In some rats, the AV_RTs exceeded the race model prediction at multiple time points, suggesting that the observed acceleration is not merely due to statistical facilitation but reflects true multisensory integration. Examples of these violations are shown in Panels a-b of the following figure. However, in other rats, the AV_RTs did not exceed the race model prediction, as illustrated in Author response image 1c-d.

      This variability may be attributed to task-specific factors in our experimental design. For instance, the rats were not under time pressure to respond immediately after cue onset, as the task emphasized accuracy over speed. This lack of urgency may have influenced their behavioral responses and movement patterns. The race model is typically applied to assess multisensory integration in tasks where rapid responses are critical, often under conditions that incentivize speed (e.g., time-restricted tasks). In our study, the absence of strict temporal constraints may have reduced the likelihood of observing consistent violations of the race model. Furthermore, In our multisensory discrimination task, animals should discriminate multiple cues and make a behavioral choice have introduced additional variability in the degree of integration observed across individual animals. Additionally, factors such as a decline in thirst levels and physical performance as the task progressed may have significantly contributed to the variability in our results. These considerations are important for contextualizing the race model findings and interpreting the data within the framework of our experimental design.

      Author response image 1.

      Reaction time cumulative distribution functions (CDFs) and race model evaluation. (a) CDFs of reaction times (RTs) for auditory (blue), visual (green), and audiovisual stimuli (red) during the multisensory discrimination task. The summed CDF of the auditory and visual conditions (dashed purple, CDF_Miller) represents the race model prediction under independent sensory processing. The dashed yellow line represents the CDF of reaction times predicted by the race model. According to the race model inequality, the CDF for audiovisual stimuli (CDF_AV) should always lie below or to the right of the sum of CDF_A and CDF_V. In this example, the inequality is violated at nearly t = 200 ms, where CDF_AV is above CDF_Miller. (b) Data from another animal, showing similar results. (c, d) CDFs of reaction times for two other animals. In these cases, the CDFs follow the race model inequality, with CDF_AV consistently lying below or to the right of CDF_A + CDF_V.

      The manuscript is very vague about the origin or responses - are these in AuD, A1, AuV... ? Some attempts to separate out responses if possible by laminar depth and certainly by field are necessary. It is known from other species that multisensory responses are more numerous, and show greater behavioural modulation in non-primary areas (e.g. Atilgan et al., 2018).

      Thank you for highlighting the importance of specifying the origin of the recorded responses. In the manuscript, we have detailed the implantation process in both the Methods and Results sections, indicating that the tetrode array was targeted to the primary auditory cortex. Using a micromanipulator (RWD, Shenzhen, China), the tetrode array was precisely positioned at stereotaxic coordinates 3.5–5.5 mm posterior to bregma and 6.4 mm lateral to the midline, and advanced to a depth of approximately 2–2.8 mm from the brain surface, corresponding to the primary auditory cortex. Although our recordings were aimed at A1, it is likely that some neurons from AuD and/or AuV were also included due to the anatomical proximity.

      In fact, in our unpublished data collected from AuD, we observed that over 50% of neurons responded to or were modulated by visual cues, consistent with findings from many other studies. This suggests that visual representations are more pronounced in AuD compared to A1. However, as noted in the manuscript, our primary focus was on A1, where we observed relatively fewer visual or audiovisual modulations in untrained rats.

      Regarding laminar depth, we regret that we were unable to determine the specific laminar layers of the recorded neurons in this study, a limitation primarily due to the constraints of our recording setup.

      Reviewer #3 (Public review):

      Summary:

      The manuscript by Chang et al. aims to investigate how the behavioral relevance of auditory and visual stimuli influences the way in which the primary auditory cortex encodes auditory, visual, and audiovisual information. The main result is that behavioral training induces an increase in the encoding of auditory and visual information and in multisensory enhancement that is mainly related to the choice located contralaterally with respect to the recorded hemisphere.

      Strengths:

      The manuscript reports the results of an elegant and well-planned experiment meant to investigate if the auditory cortex encodes visual information and how learning shapes visual responsiveness in the auditory cortex. Analyses are typically well done and properly address the questions raised.

      We sincerely thank the reviewer for their thoughtful and positive evaluation of our study.

      Weaknesses:

      Major

      (1) The authors apparently primarily focus their analyses of sensory-evoked responses in approximately the first 100 ms following stimulus onset. Even if I could not find an indication of which precise temporal range the authors used for analysis in the manuscript, this is the range where sensory-evoked responses are shown to occur in the manuscript figures. While this is a reasonable range for auditory evoked responses, the same cannot be said for visual responses, which commonly peak around 100-120 ms, in V1. In fact, the latency and overall shape of visual responses are quite different from typical visual responses, that are commonly shown to display a delay of up to 100 ms with respect to auditory responses. All traces that the authors show, instead, display visual responses strikingly overlapping with auditory ones, which is not in line with what one would expect based on our physiological understanding of cortical visually-evoked responses. Similarly, the fact that the onset of decoding accuracy (Figure 2j) anticipates during multisensory compared to auditory-only trials is hard to reconcile with the fact that visual responses have a later onset latency compared to auditory ones. The authors thus need to provide unequivocal evidence that the results they observe are truly visual in origin. This is especially important in view of the ever-growing literature showing that sensory cortices encode signals representing spontaneous motor actions, but also other forms of non-sensory information that can be taken prima facie to be of sensory origin. This is a problem that only now we realize has affected a lot of early literature, especially - but not only - in the field of multisensory processing. It is thus imperative that the authors provide evidence supporting the true visual nature of the activity reported during auditory and multisensory conditions, in both trained, free-choice, and anesthetized conditions. This could for example be achieved causally (e.g. via optogenetics) to provide the strongest evidence about the visual nature of the reported results, but it's up to the authors to identify a viable solution. This also applies to the enhancement of matched stimuli, that could potentially be explained in terms of spontaneous motor activity and/or pre-motor influences. In the absence of this evidence, I would discourage the author from drawing any conclusion about the visual nature of the observed activity in the auditory cortex.

      We thank the reviewers for highlighting the critical issue of validating the sensory origin of the reported responses, particularly regarding the timing of visual responses and the potential confound of motor-related activity.

      We analyzed neural responses within the first 150 ms following cue onset, as stated in the manuscript. This temporal window encompasses the peak of visual responses. The responses to visual stimuli occur predominantly within the first 100 ms after cue onset, preceding the initiation of body movements in behavioral tasks. This temporal dissociation aligns with previous studies, which demonstrate that motor-related activity in sensory cortices generally emerges later and is often associated with auditory rather than visual stimuli

      We acknowledge that auditory responses are typically faster than visual responses due to distinct transduction mechanisms. However, in our experiment, we intentionally designed the stimulus setup to elicit auditory and visual responses within a similar time window to maximize the potential for multisensory integration. Specifically, we used pure tone sounds with a 15 ms ramp and visual stimuli generated by an LED array, which produce faster responses compared to commonly used light bars shown on a screen. The long ramp of the auditory stimulus slightly delayed auditory response onset, while the LED-generated bar elicited visual responses more quickly (Supplementary Fig. 2). This alignment facilitated the observed overlap in response latencies. As we measured in neurons with robust visual response, first spike latencies is approximately 40 ms, as exemplified by a neuron with a low spontaneous firing rate and a strong, stimulus-evoked response (Supplementary Fig. 4). Across the population (n = 559 neurons), auditory responses reached 0.5 of the mean Z-scored response 15 ms earlier than visual responses on average (Supplementary Fig. 2). We cited Supplementary Fig. 4 in the Results section as follows:

      “Regarding the visual modality, 41% (80/196) of visually-responsive neurons showed a significant visual preference (Fig. 2f). The visual responses observed within the 0–150 ms window after cue onset were consistent and unlikely to result from visually evoked movement-related activity. This conclusion is supported by the early timing of the response (Fig. 2e) and exemplified by a neuron with a low spontaneous firing rate and a robust, stimulus-evoked response (Supplementary Fig. 4).”

      We acknowledge the growing body of literature suggesting that sensory cortices can encode signals related to motor actions or non-sensory factors. To address this concern, we emphasize that visual responses were present not only during behavioral tasks but also in anesthetized conditions, where motor-related signals are absent. Additionally, movement-evoked responses tend to be stereotyped and non-discriminative. In contrast, the visual responses observed in our study were highly consistent and selective to visual cue properties, further supporting their sensory origin.

      In summary, the combination of anesthetized and behavioral recordings, the temporal profile of responses, and their discriminative nature strongly support the sensory (visual) origin of the observed activity within the early response period. While the current study provides strong temporal and experimental evidence for the sensory origin of the visual responses, we agree that causal approaches, such as optogenetic silencing of visual input, could provide even stronger validation. Future work will explore these methods to further dissect the visual contributions to auditory cortical activity.

      (2) The finding that AC neurons in trained mice preferentially respond - and enhance - auditory and visual responses pertaining to the contralateral choice is interesting, but the study does not show evidence for the functional relevance of this phenomenon. As has become more and more evident over the past few years (see e.g. the literature on mouse PPC), correlated neural activity is not an indication of functional role. Therefore, in the absence of causal evidence, the functional role of the reported AC correlates should not be overstated by the authors. My opinion is that, starting from the title, the authors need to much more carefully discuss the implications of their findings.

      We fully agree that correlational data alone cannot establish causality. In light of your suggestion, we will revise the manuscript to more carefully discuss the implications of our findings, acknowledging that the preferred responses observed in AC neurons, particularly in relation to the contralateral choice, are correlational. We have updated several sentences in the manuscript to avoid overstating the functional relevance of these observations. Below are the revisions we have made:

      Abstract section

      "Importantly, many audiovisual neurons in the AC exhibited experience-dependent associations between their visual and auditory preferences, displaying a unique integration model. This model employed selective multisensory enhancement for the auditory-visual pairing guiding the contralateral choice, which correlated with improved multisensory discrimination."

      (Page 8, fourth paragraph in Results Section)

      "This aligns with findings that neurons in the AC and medial prefrontal cortex selectively preferred the tone associated with the behavioral choice contralateral to the recorded cortices during sound discrimination tasks, potentially reflecting the formation of sound-to-action associations. However, this preference represents a neural correlate, and further work is required to establish its causal link to behavioral choices."

      (rewrite 3rd paragraph in Discussion Section)

      "Consistent with prior research(10,31), most AC neurons exhibited a selective preference for cues associated with contralateral choices, regardless of the sensory modality. This suggests that AC neurons may contribute to linking sensory inputs with decision-making, although their causal role remains to be examined. "

      "These results indicate that multisensory training could drive the formation of specialized neural circuits within the auditory cortex, facilitating integrated processing of related auditory and visual information. However, further causal studies are required to confirm this hypothesis and to determine whether the auditory cortex is the primary site of these circuit modifications."

      MINOR:

      (1) The manuscript is lacking what pertains to the revised interpretation of most studies about audiovisual interactions in primary sensory cortices following the recent studies revealing that most of what was considered to be crossmodal actually reflects motor aspects. In particular, recent evidence suggests that sensory-induced spontaneous motor responses may have a surprisingly fast latency (within 40 ms; Clayton et al. 2024). Such responses might also underlie the contralaterally-tuned responses observed by the authors if one assumes that mice learn a stereotypical response that is primed by the upcoming goal-directed, learned response. Given that a full exploration of this issue would require high-speed tracking of orofacial and body motions, the authors should at least revise the discussion and the possible interpretation of their results not just on the basis of the literature, but after carefully revising the literature in view of the most recent findings, that challenge earlier interpretations of experimental results.

      Thank you for pointing out this important consideration. We have revised the discussion (paragraph 8-9) as follows:

      “There is ongoing debate about whether cross-sensory responses in sensory cortices predominantly reflect sensory inputs or are influenced by behavioral factors, such as cue-induced body movements. A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(48). Several studies have demonstrated sensory neurons can encode signals associated with whisking(49), running(50), pupil dilation (510 and other movements(52). In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset. This early timing suggests that the observed responses likely reflect direct sensory inputs, rather than being modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(53).

      A recent study by Clayton et al. (2024) demonstrated that sensory stimuli can evoke rapid motor responses, such as facial twitches, within 50 ms, mediated by subcortical pathways and modulated by descending corticofugal input(56). These motor responses provide a sensitive behavioral index of auditory processing. Although Clayton et al. did not observe visually evoked facial movements, it is plausible that visually driven motor activity occurs more frequently in freely moving animals compared to head-fixed conditions. In goal-directed tasks, such rapid motor responses might contribute to the contralaterally tuned responses observed in our study, potentially reflecting preparatory motor behaviors associated with learned responses. Consequently, some of the audiovisual integration observed in the auditory cortex may represent a combination of multisensory processing and preparatory motor activity. Comprehensive investigation of these motor influences would require high-speed tracking of orofacial and body movements. Therefore, our findings should be interpreted with this consideration in mind. Future studies should aim to systematically monitor and control eye, orofacial, and body movements to disentangle sensory-driven responses from motor-related contributions, enhancing our understanding of motor planning’s role in multisensory integration.”

      (2) The methods section is a bit lacking in details. For instance, information about the temporal window of analysis for sensory-evoked responses is lacking. Another example: for the spike sorting procedure, limited details are given about inclusion/exclusion criteria. This makes it hard to navigate the manuscript and fully understand the experimental paradigm. I would recommend critically revising and expanding the methods section.

      Thank you for raising this point. We clarified the temporal window by including additional details in the methods section, even though this information was already mentioned in the results section. Specifically, we now state:

      (Neural recordings and Analysis in methods section)

      “...These neural signals, along with trace signals representing the stimuli and session performance information, were transmitted to a PC for online observation and data storage. Neural responses were analyzed within a 0-150ms temporal window after cue onset, as this period was identified as containing the main cue-evoked responses for most neurons. This time window was selected based on the consistent and robust neural activity observed during this period.”

      We appreciate your concern regarding spike sorting procedure. To address this, we have expanded the methods section to provide more detailed information about the quality of our single-unit recordings. we have added detailed information in the text, as shown below (Analysis of electrophysiological data in methods section):

      “Initially, the recorded raw neural signals were band-pass filtered in the range of 300-6000 Hz to eliminate field potentials. A threshold criterion, set at no less than three times the standard deviation (SD) above the background noise, was applied to automatically identify spike peaks. The detected spike waveforms were then subjected to clustering using template-matching and built-in principal component analysis tool in a three-dimensional feature space. Manual curation was conducted to refine the sorting process. Each putative single unit was evaluated based on its waveform and firing patterns over time. Waveforms with inter-spike intervals of less than 2.0 ms were excluded from further analysis. Spike trains corresponding to an individual unit were aligned to the onset of the stimulus and grouped based on different cue and choice conditions. Units were included in further analysis only if their presence was stable throughout the session, and their mean firing rate exceeded 2 Hz. The reliability of auditory and visual responses for each unit was assessed, with well-isolated units typically showing the highest response reliability.”

      Reviewer #1 (Recommendations for the authors):

      (1) Some of the ordering of content in the introduction could be improved. E.g. line 49 reflects statements about the importance of sensory experience, which is the topic of the subsequent paragraph. In the discussion, line 436, there is a discussion of the same findings as line 442. These two paragraphs in general appear to discuss similar content. Similarly, the paragraph starting at line 424 and at line 451 both discuss the plasticity of multisensory responses through audiovisual experience, as well as the paragraph starting at line 475 (but now audiovisual pairing is dubbed semantic). In the discussion of how congruency/experience shapes multisensory interactions, the authors should relate their findings to those of Meijer et al. 2017 and Garner and Keller 2022 (visual cortex) about enhanced and suppressed responses and their potential role (as well as other literature such as Banks et al. 2011 in AC).

      We thank the reviewer for their detailed observations and valuable recommendations to improve the manuscript's organization. Below, we address each point:

      We deleted the sentence, "Sensory experience has been shown to shape cross-modal presentations in sensory cortices" (Line 49), as the subsequent paragraph discusses sensory experience in detail.

      To avoid repetition, we removed the sentence, "This suggests that multisensory training enhances AC's ability to process visual information" (Lines 442–443).

      Regarding the paragraph starting at Line 475, we believe its current form is appropriate, as it focuses on the influence of semantic congruence on multisensory integration, which differs from the topics discussed in the other paragraphs.

      We have cited the three papers suggested by the reviewer in the appropriate sections of the manuscript.

      (Paragraph 6 in discussion section)

      “…A study conducted on the gustatory cortex of alert rats has shown that cross-modal associative learning was linked to a dramatic increase in the prevalence of neurons responding to nongustatory stimuli (24). Moreover, in the primary visual cortex, experience-dependent interactions can arise from learned sequential associations between auditory and visual stimuli, mediated by corticocortical connections rather than simultaneous audiovisual presentations (26).”

      (Paragraph 2 in discussion section)

      “...Meijer et al. reported that congruent audiovisual stimuli evoke balanced enhancement and suppression in V1, while incongruent stimuli predominantly lead to suppression(6), mirroring our findings in AC, where multisensory integration was dependent on stimulus feature…”

      (Paragraph 2 in introduction section)

      “...Anatomical investigations reveal reciprocal nerve projections between auditory and visual cortices(4,11-15), highlighting the interconnected nature of these sensory systems. Moreover, two-photon calcium imaging in awake mice has shown that audiovisual encoding in the primary visual cortex depends on the temporal congruency of stimuli, with temporally congruent audiovisual stimuli eliciting balanced enhancement and suppression, whereas incongruent stimuli predominantly result in suppression(6).”

      (2) The finding of purely visually responsive neurons in the auditory cortex that moreover discriminate the stimuli is surprising given previous results (Iurilli et al. 2012, Morrill and Hasenstaub 2018 (only L6), Oude Lohuis et al. 2024, Atilgan et al. 2018, Chou et al. 2020). Reporting the latency of this response is interesting information about the potential pathways by which this information could reach the auditory system. Furthermore, spike isolation quality and histological verification are described in little detail. It is crucial for statements about the auditory, visual, or audiovisual response of individual neurons to substantiate the confidence level about the quality of single-unit recordings and where they were recorded. Do the authors have data to support that visual and audiovisual responses were not restricted to posteromedial tetrodes or clusters with poor quality? A discussion of finding V-responsive units in AC with respect to literature is warranted. Furthermore, the finding that also in visual trials behaviorally relevant information about the visual cue (with a bias for the contralateral choice cue) is sent to the AC is pivotal in the interpretation of the results, which as far as I note not really considered that much.

      We appreciate the reviewer’s thoughtful comments and have addressed them as follows:

      Discussion of finding choice-related V-responsive units in AC with respect to literature and potential pathways

      3rd paragraph in the Discussion section

      “Consistent with prior research(10,31), most AC neurons exhibited a selective preference for cues associated with contralateral choices, regardless of the sensory modality. This suggests that AC neurons may contribute to linking sensory inputs with decision-making, although their causal role remains to be examined. Associative learning may drive the formation of new connections between sensory and motor areas of the brain, such as cortico-cortical pathways(35). Notably, this cue-preference biasing was absent in the free-choice group. A similar bias was also reported in a previous study, where auditory discrimination learning selectively potentiated corticostriatal synapses from neurons representing either high or low frequencies associated with contralateral choices(32)…”

      6th paragraph in the Discussion section

      “Our results extend prior finding(4,47), showing that visual input not only reaches the AC but can also drive discriminative responses, particularly during task engagement. This task-specific plasticity enhances cross-modal integration, as demonstrated in other sensory systems. For example, calcium imaging studies in mice showed that a subset of multimodal neurons in visual cortex develops enhanced auditory responses to the paired auditory stimulus following coincident auditory–visual experience(25)…”

      8th paragraph in the Discussion section

      “…In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset, suggesting that visual information reaches the AC through rapid pathways. Potential candidates include direct or fast cross-modal inputs, such as pulvinar-mediated pathways(8) or corticocortical connections(5,54), rather than slower associative mechanisms. This early timing indicates that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55).”

      Response Latency

      Regarding the latency of visually driven responses, we have included this information in our response to the second reviewer’s first weakness (please see the above). Briefly, we analyzed neural responses within a 0-150ms temporal window after cue onset, as this period captures the most consistent and robust cue-evoked responses across neurons.

      Purely Visually Responsive Neurons in A1

      We agree that the finding of visually responsive neurons in the auditory cortex may initially seem surprising. However, these neurons might not have been sensitive to target auditory cues in our task but could still respond to other sound types. Cortical neurons are known to exhibit significant plasticity during the cue discrimination tasks, as well as during passive sensory exposure. Thus, the presence of visually responsive neurons is not inconsistent with prior findings but highlights task-specific sensory tuning. We confirm that responses were not restricted to posteromedial tetrodes or low-quality clusters (see an example of a robust visually responsive neuron in supplementary Fig. 4). Histological analysis verified electrode placements across the auditory cortex.

      For spike sorting, we have added detailed information in the text, as shown below:

      “Initially, the recorded raw neural signals were band-pass filtered in the range of 300-6000 Hz to eliminate field potentials. A threshold criterion, set at no less than three times the standard deviation (SD) above the background noise, was applied to automatically identify spike peaks. The detected spike waveforms were then subjected to clustering using template-matching and built-in principal component analysis tool in a three-dimensional feature space. Manual curation was conducted to refine the sorting process. Each putative single unit was evaluated based on its waveform and firing patterns over time. Waveforms with inter-spike intervals of less than 2.0 ms were excluded from further analysis. Spike trains corresponding to an individual unit were aligned to the onset of the stimulus and grouped based on different cue and choice conditions. Units were included in further analysis only if their presence was stable throughout the session, and their mean firing rate exceeded 2 Hz. The reliability of auditory and visual responses for each unit was assessed, with well-isolated units typically showing the highest response reliability.”

      (3) In the abstract it seems that in "Additionally, AC neurons..." the connective word 'additionally' is misleading as it is mainly a rephrasing of the previous statement.

      Replaced "Additionally" with "Furthermore" to better signal elaboration and continuity.

      (4) The experiments included multisensory conflict trials - incongruent audiovisual stimuli. What was the behavior for these trials given multiple interesting studies on the neural correlates of sensory dominance (Song et al. 2017, Coen et al. 2023, Oude Lohuis et al. 2024).

      We appreciate your feedback and have addressed it by including a new figure (supplemental Fig. 8) that illustrates choice selection during incongruent audiovisual stimuli. Panel (a) shows that rats displayed confusion when exposed to mismatched stimuli, resulting in choice patterns that differed from those observed in panel (b), where consistent audiovisual stimuli were presented. To provide clarity and integrate this new figure effectively into the manuscript, we updated the results section as follows:

      “...Rats received water rewards with a 50% chance in either port when an unmatched multisensory cue was triggered. Behavioral analysis revealed that Rats displayed notable confusion in response to unmatched multisensory cues, as evidenced by their inconsistent choice patterns (supplementary Fig. 8).”

      (5) Line 47: The AC does not 'perceive' sound frequency, individual brain regions are not thought to perceive.

      e appreciate the reviewer’s observation and have revised the sentence to ensure scientific accuracy. The updated sentence in the second paragraph of the Introduction now reads:

      “Even irrelevant visual cues can affect sound discrimination in AC<sup>10</sup>.”

      (6) Line 59-63: The three questions are not completely clear to me. Both what they mean exactly and how they are different. E.g. Line 60: without specification, it is hard to understand which 'strategies' are meant by the "same or different strategies"? And Line 61: What is meant by the quotation marks for match and mismatch? I assume this is referring to learned congruency and incongruency, which appears almost the same question as number 3 (how learning affects the cortical representation).

      We have revised the three questions for improved clarity and distinction as follows:<br /> “This limits our understanding of multisensory integration in sensory cortices, particularly regarding: (1) Do neurons in sensory cortices adopt consistent integration strategies across different audiovisual pairings, or do these strategies vary depending on the pairing? (2) How does multisensory perceptual learning reshape cortical representations of audiovisual objects? (3) How does the congruence between auditory and visual features—whether they "match" or "mismatch" based on learned associations—impact neural integration?”

      (7) Is the data in Figures 1c and d only hits?

      Only correct trials are included. We add this information in the figure legend. Please see Fig. 1 legend. Also, please see below

      “c Cumulative frequency distribution of reaction time (time from cue onset to leaving the central port) for one representative rat in auditory, visual and multisensory trials (correct only). d Comparison of average reaction times across rats in auditory, visual, and multisensory trials (correct only).”

      (8) Figure S1b: Preferred frequency is binned in non-equidistant bins, neither linear nor logarithmic. It is unclear what the reason is.

      The edges of the bins for the preferred frequency were determined based on a 0.5-octave increment, starting from the smallest boundary of 8 kHz. Specifically, the bin edges were calculated as follows:

      8×2<sup>0.5</sup>=11.3 kHz;

      8×2<sup>1</sup>=16 kHz;

      8×2<sup>1.5</sup>=22.6 kHz;

      8×2<sup>2</sup>=32 kHz;

      This approach reflects the common practice of using changes in octaves to define differences between pure tone frequencies, as it aligns with the logarithmic perception of sound frequency in auditory neuroscience.

      (9) Figure S1d: why are the responses all most neurons very strongly correlated given the frequency tuning of A1 neurons? Further, the mean normalized response presented in Figure S2e does seem to indicate a stronger response for 10kHz tones than 3kHz, in conflict with the data from anesthetized rats presented in Figure S2e.

      There is no discrepancy in the data. In Figure S1d, we compared neuronal responses to 10 kHz and 3 kHz tones, demonstrating that most neurons responded well to both frequencies. This panel does not aim to illustrate frequency selectivity but rather the overall responsiveness of neurons to these tones. For detailed information on sound selectivity, readers can refer to Figures S3a-b, which show that while more neurons preferred 10 kHz tones, the proportion is lower than in neurons recorded during the multisensory discrimination task. This distinction explains the observed differences and aligns with the results presented.

      (10) Line 79: For clarity, it can be added that the multisensory trials presented are congruent trials (jointly indicated rewarded port), and perhaps that incongruent trials are discussed later in the paper.

      We believe additional clarification is unnecessary, as the designations "A<sup>3k</sup>V<sup>hz</sup>" and "A<sup>10k</sup>V<sup>vt</sup>" clearly indicate the specific combinations of auditory and visual cues presented during congruent trials. Additionally, the discussion of incongruent trials is provided later in the manuscript, as noted by the reviewer.

      (11) Line 111: the description leaves unclear that the 35% reflects the combination of units responsive to visual only and responsive to auditory or visual.

      The information is clearly presented in Figure 2b, which shows the proportions of neurons responding to auditory-only (A), visual-only (V), both auditory and visual (A, V), and audiovisual-only (VA) stimuli in a pie chart. Readers can refer to this figure for a detailed breakdown of the neuronal response categories.

      (12) Figure 2h: consider a colormap with diverging palette and equal positive and negative maximum (e.g. -0.6 to 0.6) and perhaps reiterate in the color bar legend which stimulus is preferred for which selectivity index.

      We appreciate the suggestion; however, we believe that the current colormap effectively conveys the data and the intended interpretation. The existing color bar legend already provides clear information about the selectivity index, and the stimulus preference is adequately explained in the figure caption. As such, further adjustments are not necessary.

      (13) Line 160: "a ratio of 60:20 for V<sup>vt</sup> 160 preferred vs. V<sup>hz</sup> preferred neurons." Is this supposed to add up to 100, or is this a ratio of 3:1?

      We rewrite the sentence. Please see below:

      “Similar to the auditory selectivity observed, a greater proportion of neurons favored the visual stimulus (V<sup>vt</sup>) associated with the contralateral choice, with a 3:1 ratio of V<sup>vt</sup>-preferred to V<sup>hz</sup>-preferred neurons.”

      (14) The statement in Figure 2g and line 166/167 could be supported by a statistical test (chi-square?).

      Thank you for the suggestion. However, we believe that a statistical test is not required in this case, as the patterns observed are clearly represented in Figure 2g. The qualitative differences between the groups are evident and sufficiently supported by the data.

      (15) Line 168, it is unclear in what sense 'dominant' is meant. Is audition perceived as a dominant sensory modality in a behavioral sense (e.g. Song et al. 2017), or are auditory signals the dominant sensory signal locally in the auditory cortex?

      Thank you for the clarification. To address your question, by "dominant," we are referring to the fact that auditory inputs are the most prominent and influential among the sensory signals feeding into the auditory cortex. This reflects the local dominance of auditory signals within the auditory cortex, rather than a behavioral dominance of auditory perception. We have revised the sentence as follows:

      “We propose that the auditory input, which dominates within the auditory cortex, acts as a 'teaching signal' that shapes visual processing through the selective reinforcement of specific visual pathways during associative learning.”

      (16) Line 180: "we discriminated between auditory, visual, and multisensory cues." This phrasing indicated that the SVMs were trained to discriminate sensory modalities (as is done later in the manuscript), rather than what was done: discriminate stimuli within different categories of trials.

      Thank you for your comment. We have revised the sentence for clarity. Please see the updated version below:

      “Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates stimulus identity within the same modality (e.g., A<sup>3k</sup> vs. A<sup>10k</sup> for auditory stimuli, V<sup>hz</sup> vs. V<sup>vt</sup> for visual stimuli, A<sup>3k</sup>V<sup>hz</sup> vs. A<sup>10k</sup>V<sup>vt</sup> for multisensory stimuli).”

      (17) Line 185: "a deeply accurate incorporation of visual processing in the auditory cortex." the phrasing is a bit excessive for a binary classification performance.

      Thank you for pointing this out. We have revised the sentence to better reflect the findings without overstating them:

      “Interestingly, AC neurons could discriminate between two visual targets with around 80% accuracy (Fig. 2j), demonstrating a meaningful incorporation of visual information into auditory cortical processing.”

      (18) Figure 3, title. An article is missing (a,an/the).

      Done. Please see below:

      Fig. 3 Auditory and visual integration in the multisensory discrimination task

      (19) Line 209, typo pvalue: p<-0.00001.

      Done (p<0.00001).

      (20) Line 209, the pattern is not weaker. The pattern is the same, but more weakly expressed.

      Thank you for your valuable feedback. We appreciate your clarification and agree that our phrasing could be improved for accuracy. The observed pattern under anesthesia is indeed the same but less strongly expressed compared to the task engagement. We have revised the sentence to better reflect this distinction:

      “A similar pattern, albeit less strongly expressed, was observed under anesthesia (Supplementary Fig. 3c-3f), suggesting that multisensory perceptual learning may induce plastic changes in AC.”

      (21) Line 211: choice-free group → free-choice group.

      Done.

      (22) Line 261: wrong → incorrect (to maintain consistent terminology).

      Done.

      (23) Line 265: why 'likely'? Are incorrect choices on the A<sup>3k</sup>-V<sup>hz</sup> trials not by definition contralateral and vice versa? Or are there other ways to have incorrect trials?

      We deleted the word of ‘likely’. Please see below:

      “…, correct choices here correspond to ipsilateral behavioral selection, while incorrect choices correspond to contralateral behavioral selection.”

      (24) Typo legend Fig 3a-c (tasks → task). (only one task performed).

      Done.

      (25) Line 400: typo: Like → like.

      Done.

      (26) Line 405: What is meant by a cohesive visual stimulus? Congruent? Rephrase.

      Done. Please see the below:

      “…layer 2/3 neurons of the primary visual cortex(7), and a congruent visual stimulus can enhance sound representation…”

      (27) Line 412: Very general statement and obviously true: depending on the task, different sensory elements need to be combined to guide adaptive behavior.

      We really appreciate the reviewer and used this sentence (see second paragraph in discussion section).

      (28) Line 428: within → between (?).

      Done.

      (29) Figure 3L is not referenced in the main text. By going through the figures and legends my understanding is that this shows that most neurons have a multisensory response that lies between 2 z-scores of the predicted response in the case of 83% of the sum of the auditory and the visual response. However, how was the 0.83 found? Empirically? Figure S3 shows a neuron that does follow a 100% summation. Perhaps the authors could quantitatively support their estimate of 83% of the A + V sum, by varying the fraction of the sum (80%, 90%, 100% etc.) and showing the distribution of the preferred fraction of the sum across neurons, or by showing the percentage of neurons that fall within 2 z-scores for each of the fractions of the sum.

      Thank you for your detailed feedback and suggestions regarding Figure 3L and the 83% multiplier.

      (1) Referencing Figure 3L:

      Figure 3L is referenced in the text. To enhance clarity, we have revised the text to explicitly highlight its relevance:

      “Specifically, as illustrated in Fig. 3k, the observed multisensory response approximated 83% of the sum of the auditory and visual responses in most cases, as quantified in Fig. 3L.”

      (2) Determination of the 0.83 Multiplier:

      The 0.83 multiplier was determined empirically by comparing observed audiovisual responses with the predicted additive responses (i.e., the sum of auditory and visual responses). For each neuron, we calculated the auditory, visual, and audiovisual responses. We then compared the observed audiovisual response with scaled sums of auditory and visual responses (Fig. 3k), expressed as fractions of the additive prediction (e.g., 0.8, 0.83, 0.9, etc.). We found that when the scaling factor was 0.83, the population-wide difference between predicted and observed multisensory responses, expressed as z-scores, was minimized. Specifically, at this value, the mean z-score across the population was approximately zero (-0.0001±1.617), indicating the smallest deviation between predicted and observed responses.

      (30) Figure 5e: how come the diagonal has 0.5 decoding accuracy within a category? Shouldn't this be high within-category accuracy? If these conditions were untested and it is an issue of the image display it would be informative to test the cross-validated performance within the category as well as a benchmark to compare the across-category performance to. Aside, it is unclear which conventions from Figure 2 are meant by the statement that conventions were the same.

      The diagonal values (~0.5 decoding accuracy) within each category reflect chance-level performance. This occurs because the decoder was trained and tested on the same category conditions in a cross-validated manner, and within-category stimulus discrimination was not the primary focus of our analysis. Specifically, the stimuli within a category shared overlapping features, leading to reduced discriminability for the decoder when distinguishing between them. Our primary objective was to assess cross-category performance rather than within-category accuracy, which may explain the observed pattern in the diagonal values.

      Regarding the reference to Figure 2, we appreciate the reviewer pointing out the ambiguity. To avoid any confusion, we have removed the sentence referencing "conventions from Figure 2" in the legend for Figure 5e, as it does not contribute meaningfully to the understanding of the results.

      (31) Line 473: "movement evoked response", what is meant by this?

      Thank the reviewer for highlighting this point. To clarify, by "movement-evoked response," we are referring to neural activity that is driven by the animal's movements, rather than by sensory inputs. This type of response is typically stereotyped, meaning that it has a consistent, repetitive pattern associated with specific movements, such as whisking, running, or other body or facial movements.

      In our study, we propose that the visually-evoked responses observed within the 150 ms time window after cue onset primarily reflect sensory inputs from the visual stimulus rather than movement-related activity. This interpretation is supported by the response timing: visual-evoked activity occurs within 100 ms of the light flash onset, a timeframe too rapid to be attributed to body or orofacial movements. Additionally, unlike stereotyped movement-evoked responses, the visual responses we observed are discriminative, varying based on specific visual features—a hallmark of sensory processing rather than motor-driven activity.

      We have revised the manuscript as follows (eighth paragraph in discussion section):

      “There is ongoing debate about whether cross-sensory responses in sensory cortices predominantly reflect sensory inputs or are influenced by behavioral factors, such as cue-induced body movements. A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(49). Several studies have demonstrated sensory neurons can encode signals associated with whisking(50), running(51), pupil dilation(52) and other movements(53). In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset. suggests that visual information reaches the AC through rapid pathways. Potential candidates include direct or fast cross-modal inputs, such as pulvinar-mediated pathways(8) or corticocortical connections(5,54), rather than slower associative mechanisms. This early timing suggests that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55). ”

      (32) Line 638-642: It is stated that a two-tailed permutation test is done. The cue selectivity can be significantly positive and negative, relative to a shuffle distribution. This is excellent. But then it is stated that if the observed ROC value exceeds the top 5% of the distribution it is deemed significant, which corresponds to a one-tailed test. How were significantly negative ROC values detected with p<0.05?

      Thank you for pointing this out. We confirm that a two-tailed permutation test was indeed used to evaluate cue selectivity. In this approach, significance is determined by comparing the observed ROC value to both tails of the shuffle distribution. Specifically, if the observed ROC value exceeds the top 2.5% or falls below the bottom 2.5% of the distribution, it is considered significant at p< 0.05. This two-tailed test ensures that both significantly positive and significantly negative cue selectivity values are identified.

      To clarify this in the manuscript, we have revised the text as follows:

      “This generated a distribution of values from which we calculated the probability of our observed result. If the observed ROC value exceeds the top 2.5% of the distribution or falls below the bottom 2.5%, it was deemed significant (i.e., p < 0.05).”

      (33) Line 472: the cited paper (reference 52) actually claims that motor-related activity in the visual cortex has an onset before 100ms and thus does not support your claim that the time window precludes any confound of behaviorally mediated activity. Furthermore, that study and reference 47 show that sensory stimuli could be discriminated based on the cue-evoked body movements and are discriminative. A stronger counterargument would be that both studies show very fast auditory-evoked body movements, but only later visually-evoked body movements.

      We appreciate the reviewer’s comments. As Lohuis et al. (reference 55) demonstrated, activity in the visual cortex (V1) can reflect distinct visual, auditory, and motor-related responses, with the latter often dissociable in timing. In their findings, visually-evoked movement-related activity arises substantially later than the sensory visual response, generally beginning around 200 ms post-stimulus onset. In contrast, auditory-evoked activity in A1 occurs relatively early.

      We have revised the manuscript as follows (eighth paragraph in discussion section):

      “A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(49). ...This early timing suggests that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55). ”

      (34) The training order (multisensory cue first) is important to briefly mention in the main text.

      We appreciate the reviewer’s suggestion and have added this information to the main text. The revised text now reads:

      “The training proceeded in two stages. In the first stage, which typically lasted 3-5 weeks, rats were trained to discriminate between two audiovisual cues. In the second stage, an additional four unisensory cues were introduced, training the rats to discriminate a total of six cues.”

      (35) Line 542: As I understand the multisensory rats were trained using the multisensory cue first, so different from the training procedure in the unisensory task rats where auditory trials were learned first.

      Thank you for pointing this out. You are correct that, in the unisensory task, rats were first trained to discriminate auditory cues, followed by visual cues. To improve clarity and avoid any confusion, we have removed the sentence "Similar to the multisensory discrimination task" from the revised text.

      (36) Line 546: Can you note on how the rats were motivated to choose both ports, or whether they did so spontaneously?

      Thank you for your insightful comment. The rats' port choice was spontaneous in this task, as there was no explicit motivation required for choosing between the ports. We have clarified this point in the text to address your concern. The revised sentence now reads:

      “They received a water reward at either port following the onset of the cue, and their port choice was spontaneous.”

      (37) It is important to mention in the main text that the population decoding is actually pseudopopulation decoding. The interpretation is sufficiently important for interpreting the results.

      Thank you for this valuable suggestion. We have revised the text to specify "pseudo-population" instead of "population" to clarify the nature of our decoding analysis. The revised text now reads:

      “Our multichannel recordings enabled us to decode sensory information from a pseudo-population of AC neurons on a single-trial basis. Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates between stimuli.”

      (38) The term modality selectivity for the description of the multisensory interaction is somewhat confusing. Modality selectivity suggests different responses to the visual or auditory trials. The authors could consider a different terminology emphasizing the multisensory interaction effect.

      Thank you for your insightful comment. We have replaced " modality selectivity " with " multisensory interactive index " (MSI). This term more accurately conveys a tendency for neurons to favor multisensory stimuli over individual sensory modalities (visual or auditory alone).

      (39) In Figures 3 e and g the color code is different from adjacent panels b and c and is to be deciphered from the legend. Consider changing the color coding, or highlight to the reader that the coloring in Figures 3b and c is different from the color code in panels 3 e and g.

      We appreciate the reviewer’s observation. However, we believe that a change in the color coding is not necessary. Figures 3e and 3g differentiate symbols by both shape and color, ensuring accessibility and clarity. This is clearly explained in the figure legend to guide readers effectively.

      (40) Figure S2b: was significance tested here?

      Yes, we did it.

      (41) Figure S2d: test used?

      Yes, test used.

      (42) Line 676: "as appropriate", was a normality test performed prior to statistical test selection?

      In our analysis, we assessed normality before choosing between parametric (paired t-test) and non-parametric (Wilcoxon signed-rank test) methods. We used the Shapiro-Wilk test to evaluate the normality of the data distributions. When data met the assumption of normality, we applied the paired t-test; otherwise, we used the Wilcoxon signed-rank test.

      Thank you for pointing this out. We confirm that a normality test was performed prior to the selection of the statistical test. Specifically, we used the Shapiro-Wilk test to assess whether the data distributions met the assumption of normality. Based on this assessment, we applied the paired t-test for normally distributed data and the Wilcoxon signed-rank test for non-normal data.

      To ensure clarity, we update the "Statistical Analysis" section of the manuscript with the following revised text:

      “For behavioral data, such as mean reaction time differences between unisensory and multisensory trials, cue selectivity and mean modality selectivity across different auditory-visual conditions, comparisons were performed using either the paired t-test or the Wilcoxon signed-rank test. The Shapiro-Wilk test was conducted to assess normality, with the paired t-test used for normally distributed data and the Wilcoxon signed-rank test for non-normal data.”

      (43) Line 679: incorrect, most data is actually represented as mean +- SEM.

      Thank you for pointing this out. In the Results section, we report data as mean ± SD for descriptive statistics, while in the figures, the error bars typically represent the standard error of the mean (SEM) to visually indicate variability around the mean. We have specified in each figure legend whether the error bars represent SD or SEM.

      Reviewer #2 (Recommendations for the authors):

      (1) Line 182 - here it sounds like you mean your classifier was trained to decode the modality of the stimulus, when I think what you mean is that you decoded the stimulus contingencies using A/V/AV cues?

      Thank you for pointing out this potential misunderstanding. We would like to clarify that the classifier was trained to decode the stimulus identity (e.g., A<sup>3k</sup> vs. A<sup>10k</sup> for auditory stimuli, V<sup>hz</sup> vs. V<sup>vt</sup> for visual stimuli, and A<sup>3k</sup>V<sup>hz</sup> vs. A<sup>10k</sup>V<sup>vt</sup> for multisensory stimuli) rather than the modality of the stimulus. The goal of the analysis was to determine how well the pseudo-population of AC neurons could distinguish between individual stimuli within the same modality. We have revised the relevant text in the revised manuscript to ensure this distinction is clear. Please see the following:

      “Our multichannel recordings enabled us to decode sensory information from a pseudo-population of AC neurons on a single-trial basis. Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates stimulus identity (e.g.,  A<sup>3k</sup> vs. A<sup>10k</sup> for auditory stimuli, V<sup>hz</sup> vs. V<sup>vt</sup> for visual stimuli,  A<sup>3k</sup>V<sup>hz</sup> vs. A<sup>10k</sup>V<sup>vt</sup> for multisensory stimuli).”

      (2) Lines 256 - here the authors look to see whether incorrect trials diminish audiovisual integration. I would probably seek to turn the causal direction around and ask are AV neurons critical for behaviour - nevertheless, since this is only correlational the causal direction cannot be unpicked. However, the finding that contralateral responses per se do not result in enhancement is a key control. Showing that multisensory enhancement is less on error trials is a good first step to linking neural activity and perception, but I wonder if the authors could take this further however by seeking to decode choice probabilities as well as stimulus features in an attempt to get a little closer to addressing the question of whether the animals are using these responses for behaviour.

      Thank you for your comment and for highlighting the importance of understanding whether audiovisual (AV) neurons are critical for behavior. As you noted, the causal relationship between AV neural activity and behavioral outcomes cannot be directly determined in our current study due to its correlational nature. We agree that this is an important topic for future exploration. In our study, we examined how incorrect trials influence multisensory enhancement. Our findings show that multisensory enhancement is less pronounced during error trials, providing an initial link between neural activity and behavioral performance. To address your suggestion, we conducted an additional analysis comparing auditory and multisensory selectivity between correct and incorrect choice trials. As shown in Supplementary Fig. 7, both auditory and multisensory selectivity were significantly lower during incorrect trials. This result highlights the potential role of these neural responses in decision-making, suggesting they may extend beyond sensory processing to influence choice selection. We have cited this figure in the Results section as follows: ( the paragraph regarding Impact of incorrect choices on audiovisual integration):

      “Overall, these findings suggest that the multisensory perception reflected by behavioral choices (correct vs. incorrect) might be shaped by the underlying integration strength. Furthermore, our analysis revealed that incorrect choices were associated with a decline in cue selectivity, as shown in Supplementary Fig. 7.”

      We acknowledge your suggestion to decode choice probabilities alongside stimulus features as a more direct approach to exploring whether animals actively use these neural responses for behavior. Unfortunately, in the current study, the low number of incorrect trials limited our ability to perform such analyses reliably. Nonetheless, we are committed to pursuing this direction in subsequent work. We plan to use techniques such as optogenetics in future studies to causally test the role of AV neurons in driving behavior.

      (3) Figure 5E - the purple and red are indistinguishable - could you make one a solid line and keep one dashed?

      We thank the reviewer for pointing out that the purple and red lines in Figure 5E were difficult to distinguish. To address this concern, we modified the figure by making two lines solid and changing the color of one square, as suggested. These adjustments enhance visual clarity and improve the distinction between them.

      (4) The unisensory control training is a really nice addition. I'm interested to know whether behaviourally these animals experienced an advantage for audiovisual stimuli in the testing phase? This is important information to include as if they don't it is one step closer to linking audiovisual responses in AC to improved behavioural performance (and if they do, we must be suitably cautious in interpretation!).

      Thank you for raising this important point. To address this, we have plotted the behavioral results for each animal (see Author response image 2). The data indicate that performance with multisensory cues is slightly better than with the corresponding unisensory cues. However, given the small sample size (n=3) and the considerable variation in behavioral performance across individuals, we remain cautious about drawing definitive conclusions on this matter. We recognize the need for further investigation to establish a robust link between audiovisual responses in the auditory cortex and improved behavioral performance. In future studies, we plan to include a larger number of animals and more thoroughly explore this relationship to provide a comprehensive understanding.

      Author response image 2.

      (5) Line 339 - I don't think you can say this leads to binding with your current behaviour or neural responses. I would agree there is a memory trace established and a preferential linking in AC neurons.

      We thank the reviewer for raising this important point. In the revised manuscript, we have clarified that our data suggest the formation of a memory trace and preferential linking in AC neurons. The text has been updated to emphasize this distinction. Please see the revised section below (first paragraph in Discussion section).

      “Interestingly, a subset of auditory neurons not only developed visual responses but also exhibited congruence between auditory and visual selectivity. These findings suggest that multisensory perceptual training establishes a memory trace of the trained audiovisual experiences within the AC and enhances the preferential linking of auditory and visual inputs. Sensory cortices, like AC, may act as a vital bridge for communicating sensory information across different modalities.”

    1. eLife Assessment

      This work provides a simple, rapid and useful protocol for the isolation of small extracellular vesicles from small volumes of plasma, using two well-known methodologies, in tandem: size exclusion chromatography (SEC) and density gradient ultracentrifugation (DGUC). The authors exhaustively test these methodologies separately and in combination, showing superior results for the SEC-DGUC in terms of purity and yield. The results obtained in this work are convincing, using multiple state-of-art methodologies for the characterization of the isolates that support their conclusions.

    2. Reviewer #2 (Public review):

      Summary:

      In this work, the authors manage to optimize a simple and rapid protocol using SEC followed by DGCU to isolate sEVs with adequate purity and yield from small volumes of plasma. Isolated fractions containing sEVs using SEC, DGCU, SEC-DGCU and DGCU-SEC are compared in terms of their yield, purity surface protein profile and RNA content. Although the combined use of these methodologies has already been evaluated in previous works, the authors manage to adapt them for the use of small volumes of plasma, which allows working in 1.5 mL tubes and reducing the centrifugation time to 2 hours.<br /> The authors finally find that although both the SEC-DGCU and DGCU-SEC combinations achieve isolates with high purity, the SEC-DGCU combination results in higher yields.<br /> This work provides an interesting tool for the rapid obtention of sEVs with sufficient yield and purity for detailed characterization which could be very useful in research and clinical therapy.

      Strengths:

      The work is well written and organized.<br /> The authors clearly state the problem they want to address, that is, optimizing a method that allows sEV to be isolated from small volumes of plasma.<br /> Although these methodologies have been tested in previous works, the authors manage to isolate sEVs of high purity and good performance through a simple and fast methodology.<br /> The characteristics of all isolated fractions are exhaustively analyzed through various state-of-the-art methodologies.<br /> They present a good interpretation of the results obtained through the methodologies used.

      Weaknesses:

      Although this work focuses on comparing different techniques and their combinations to find an optimal option, the authors could strengthen their analysis by using statistical methods that reliably show the differences between the explored techniques.

      Comments on revisions:

      Although superiority of the proposed method was demonstrated by other techniques, it is always advisable to calculate the differences between different methodologies through different statistical methods, whenever possible, to strengthen the obtained results.

    3. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public Review):

      Summary:

      In their manuscript, Kong Fang et al describe a robust pipeline for the isolation of small extracellular vesicles through a combination of size exclusion chromatography and miniaturized density gradient separation. Subsequently, they prove that the method is reproducible and suitable for small-volume operations while at the same time not compromising the quality of vesicles.

      Strengths:

      The paper narrates a robust method for purifying high-quality sEVs from small amounts of blood plasma. They also demonstrate that through this approach, they can derive sEVs without compromising the protein composition, integrity of the vesicles, or contamination with other proteins or lipids.

      Weaknesses:

      The paper is a nice summary of how to enrich sEVs from blood samples. Although well performed and substantiated with data, the paper primarily deals with method development and optimisation.

      We agree with the reviewer's assessment that this paper primarily focuses on the development and optimization of a method. Using this robust technique for isolating small extracellular vesicles (sEVs) from small blood volumes, our future research will investigate sEVs isolated from clinical samples, with a particular focus on their role in various diseases.

      Reviewer #2 (Public Review):

      Summary:

      In this work, the authors manage to optimize a simple and rapid protocol using SEC followed by DGCU to isolate sEVs with adequate purity and yield from small volumes of plasma. Isolated fractions containing sEVs using SEC, DGCU, SEC-DGCU, and DGCU-SEC are compared in terms of their yield, purity surface protein profile, and RNA content. Although the combined use of these methodologies has already been evaluated in previous works, the authors manage to adapt them for the use of small volumes of plasma, which allows working in 1.5 mL tubes and reducing the centrifugation time to 2 hours.

      The authors finally find that although both the SEC-DGCU and DGCU-SEC combinations achieve isolates with high purity, the SEC-DGCU combination results in higher yields.

      This work provides an interesting tool for the rapid obtention of sEVs with sufficient yield and purity for detailed characterization which could be very useful in research and clinical therapy.

      Strengths:

      - The work is well-written and organized.

      - The authors clearly state the problem they want to address, that is, optimizing a method that allows sEV to be isolated from small volumes of plasma.

      - Although these methodologies have been tested in previous works, the authors manage to isolate sEVs of high purity and good performance through a simple and fast methodology.

      - The characteristics of all isolated fractions are exhaustively analyzed through various state-of-the-art methodologies.

      - They present a good interpretation of the results obtained through the methodologies used.

      Weaknesses:

      - Lack of references that support some of the results obtained.

      - Although this work focuses on comparing different techniques and their combinations to find an optimal option, the authors do not use any statistical method that reliably shows the differences between these techniques, except when repeatability is measured.

      We appreciate the reviewer's insightful comments and will incorporate the suggested missing references. We acknowledge that we did not perform statistical analyses when comparing the differences among the three methods. Nevertheless, the superiority of the SEC-DGUC method is evident from observations based on several independent characterization methods, including Cryo-EM, TEM, western blot, and total RNA quantification.

      Firstly, repeated Cryo-EM observations consistently confirm that the SEC-alone method shows severe lipoprotein contamination while the SEC-DGUC method drastically reduces such lipoprotein contamination. In comparing the SEC-DGUC and DGUC-SEC methods, multiple independent characterization methods showed that the SEC-DGUC method yields significantly greater quantity of sEVs: 1) The western blot experiment showed much higher signal intensity for all four tested sEV markers (CD9, CD63, CD81, and TSG101), with estimated concentrations approximately 2.1, 2.1, 4.7, and 4.2 times higher than the DGUC-SEC method. 2) The total RNA analysis showed that SEC-DGUC-1 contained more than 4 times the total amount of RNA compared to DGUC-SEC-PF. 3) Establishing the normalization baseline, particle size distributions in SEC-DGUC-1 and DGUC-SEC-PF measured by TEM were found to be similar, suggesting comparable purity and distribution of the captured sEVs. For comparison purposes, within each independent characterization method, the same plasma source and total plasma volume were used, while across different methods, different plasma sources were used. These independent characterization methods have consistently demonstrated the superiority of the SEC-DGUC method over the DGUC-SEC or SEC-alone methods.

      Recommendations for the authors:

      Reviewer #1 (Recommendations For The Authors):

      In my opinion, this work is elegantly designed and supported by data, which would motivate more studies related to blood-derived microvesicles in the context of infectious and systemic diseases. Overall, the manuscript is well-written and explained in sufficient detail. I only have minor comments.

      (1) Recruitment of volunteers for blood/plasma collection: there is a need for a statement that this was in accordance with ethical and biosafety regulations of the Institute/Clinic.

      We added two sentences at the beginning of the Blood Collection section (under Materials and methods): “All procedures involving peripheral blood specimens were approved by the Singapore National Health Group Domain Specific Review Board (the central ethics committee) and were mutually recognized by the Nanyang Technological University Institutional Review Board (IRB#2018/00671). All blood specimens were de-identified prior to their use in the experiments.”

      (2) Since this is a method development and validation article, it would be good to include an image of the iodixanol gradient with the high-density sEV zone, after centrifugation.

      We have incorporated an image after centrifugation in Supplementary Figure 3.

      (3) Although several sEV markers are shown in Figure 7A, flotillin is missing in this figure which was part of Figure 6B. Does flotillin show a different pattern? Flotillin is a DRM-associated marker, and hence may behave differently, would be interesting to add any insights.

      We appreciate the reviewer’s careful observation. In Figure 6B, Flotillin was used to confirm the presence of sEVs in different density zones. However, for the purpose of comparing the yield between the SEC-DGUC and DGUC-SEC methods, as shown in Figure 7A, Flotillin was not included in the western blot analysis. No obvious pattern changes were observed in other sEV markers tested in both Figures 6B and 7A.   

      (4) Methods section of LC/MS analysis- which protein database was used for protein identification?

      We added the following sentence at the end of the LC/MS analysis section: “The protein database used for protein identification was Uniprot Human.”

      Reviewer #2 (Recommendations For The Authors):

      In line 43 some references are needed.

      We added this reference: EL Andaloussi, S., Mäger, I., Breakefield, X. et al. Extracellular vesicles: biology and emerging therapeutic opportunities. Nat Rev Drug Discov 12, 347–357 (2013). https://doi.org/10.1038/nrd3978

      In line 107, please avoid using short forms such as "it's".

      We have revised that to “it is.”

      In line 153: "...separates low-density particles from those of high density, but a considerable amount of..." the word "but" should not be in the sentence.

      We have removed “but” in this sentence.

      In line 181 the authors establish that "Notably, SEC-PF exhibited a high level of ApoB and low expression of sEV markers." Is there any explanation for this?

      SEC-PF represents the eluate from the SEC step, collected before the DGUC step. This fraction contains a mixture of lipoproteins and sEVs. Due to the overwhelming abundance of lipoproteins compared to sEVs, the western blot predictably shows a high level of ApoB with minimal expression of sEV markers. This highlights that SEC alone effectively reduces plasma protein content but does not efficiently remove lipoproteins. Figure 6C further illustrates this point, as cryo-EM images of SEC-PF reveal the presence of sEVs, which are vastly outnumbered by lipoproteins.

      In line 198, the sentence "Theoretically, the DGUC-SEC protocol should also effectively isolate sEVs from plasma" need to be supported by references.

      See for instance:

      - Holcar M, Ferdin J, Sitar S, Tušek-Žnidarič M, Dolžan V, Plemenitaš A, Žagar E, Lenassi M. 2020. Enrichment of plasma extracellular vesicles for reliable quantification of their size and concentration for biomarker discovery. Sci Rep 10:21346. doi:10.1038/s41598-020-78422-y.

      - Jia Y, Yu L, Ma T, Xu W, Qian H, Sun Y, Shi H. 2022. Small extracellular vesicles isolation and separation: Current techniques, pending questions and clinical applications. Theranostics 12:6548-6575. doi:10.7150/thno.74305

      - Vergauwen G, Dhondt B, Van Deun J, De Smedt E, Berx G, Timmerman E, Gevaert K, Miinalainen I, Cocquyt V, Braems G, Van den Broecke R, Denys H, De Wever O, Hendrix A. 2017. Confounding factors of ultrafiltration and protein analysis in extracellular vesicle research. Sci Rep 7:2704. doi:10.1038/s41598-017-02599-y

      We have added this reference: Holcar M, Ferdin J, Sitar S, Tušek-Žnidarič M, Dolžan V, Plemenitaš A, Žagar E, Lenassi M. 2020. Enrichment of plasma extracellular vesicles for reliable quantification of their size and concentration for biomarker discovery. Sci Rep 10:21346. https://doi.org/10.1038/s41598-020-78422-y.  

      In line 309 the authors establish that "NTA measured size distributions displayed well-overlapped histograms of particles". It is possible for the authors to analyze this overlapping using some statistical test as a chi-squared test?

      We have conducted a statistical analysis of the histogram similarities using the Jensen-Shannon Divergence (JSD) method. This is reflected in the manuscript under the results section, “Repeatability and reliability of the SEC-DGUC protocol”, where we state: “We then compared size distributions for each plasma fraction using Jensen-Shannon Divergence (JSD). The JSD values, which are well below 0.1 (Figure 10B), indicate a consistent population of isolated particles, as further supported by Supplementary Figure 8.” Additionally, we included JSD values in the legend of Figure 10B: “JSD values for SEC-DGUC-1 to 4 are 0.015, 0.006, 0.001, and 0.002, indicating strong similarities among the histograms.” These additions demonstrate the robustness and repeatability of the SEC-DGUC protocol.

      In line 360, "lasts ~ 16 hours or more." This statement needs a reference that supports this time.

      We have added this reference: Vergauwen, G. et al. Robust sequential biophysical fractionation of blood plasma to study variations in the biomolecular landscape of systemically circulating extracellular vesicles across clinical conditions. J Extracell Vesicles 10, e12122 (2021).

      In line 399, the reference format is different from the previously used format.

      This is corrected. We thank the reviewer for this careful examination.

      Line 466: This sentence is not quite clear. It can be understood that for every 0.5 mL of plasma, 2 mL of particle fraction are obtained and that for 6 mL of plasma, this method will give a total volume of 24 mL. However, it is not clear what is meant by the fact that it has been concentrated to 6 mL. While one can assume that those final 6 mL concentrates come from the initial 24 mL, perhaps the way this sentence was worded was not appropriate. I would recommend rewriting it for a simpler interpretation of how this method was performed.

      We have changed the sentence to: “For the DGUC experiment using the 12 ml tube, 24 ml of PFs were obtained from 6 ml of plasma and subsequently concentrated to 6 ml. The 6 ml of concentrated PFs were then transferred to a Beckman Coulter ultra-clear centrifuge tube (344059, Beckman Coulter, USA) for further processing.”

      Line 519: The authors established a second dilution to avoid absorbance values above 1.2. Is there any justification for this value, taking into account that the Lambert-Beer law presents more precision in the absorbance range of 0.2 to 0.8?

      We have added this reference: https://diagnostic.serumwerk.com/wp-content/uploads/2021/05/V05-Serumwerk.pdf

      Line 519-520: "Also included were water and 0.25 M sucrose as blanks". Perhaps authors could consider rephrasing this sentence.

      We have changed the sentence to: “The absorbance measurements were made against water and 0.25 M sucrose blanks.”

      In line 520, the sentence must say "each sample was made by triplicate".

      We have changed the sentence to: “Each sample was prepared by triplicate to reduce error.” We thank the reviewer for this suggestion.

      Line 673: The phrase "0.1% formic acid in 100% ACN" would be better, in my opinion, if it said "0.1% formic acid in ACN".

      Yes, these two expressions have the same meaning. However, to ensure clarity, we have updated the description to “0.1% formic acid in ACN.”. We thank reviewer for this suggestion.

      Supplementary Figure 1: in the Figure caption there is an error in the numbering: at the end, where it is written (E), it should be (F). Please, correct this.

      We have made the necessary correction and sincerely appreciate the reviewer’s attentiveness.

      Supplementary Figure 5: Some sEVs are hard to visualize due to poor image resolution. Is there any possibility for the authors to enhance these images?

      We thank the reviewer for this valuable comment. To improve the visual clarity of the images, we have opted to display four sub-figures instead of nine.

    1. eLife Assessment

      This valuable study provides a comprehensive description of the Nematostella vectensis matrisome - the genes encoding the proteins of the extracellular matrix. The authors combine new mass spectrometry data with bioinformatic analyses of previously published genomic and single-cell RNAseq data. Although this work will be of interest to biologists working on the evolution of the matrisome, as well as more broadly those working with non-bilaterian animals, in its current state it is incomplete due to the lack of rigorous criteria for manual curation and comprehensive annotation of the predicted matrisome.

    2. Reviewer #1 (Public review):

      Summary:

      In this manuscript entitled "Molecular dynamics of the matrisome across sea anemone life history", Bergheim and colleagues report the prediction, using an established sequence analysis pipeline, of the "matrisome" - that is, the compendium of genes encoding constituents of the extracellular matrix - of the starlet sea anemone Nematostella vectensis. Re-analysis of an existing scRNA-Seq dataset allowed the authors to identify the cell types expressing matrisome components and different developmental stages. Last, the authors apply time-resolved proteomics to provide experimental evidence of the presence of the extracellular matrix proteins at three different stages of the life cycle of the sea anemone (larva, primary polyp, adult) and show that different subsets of matrisome components are present in the ECM at different life stages with, for example, basement membrane components accompanying the transition from larva to primary polyp and elastic fiber components and matricellular proteins accompanying the transition from primary polyp to the adult stage.

      Strengths:

      The ECM is a structure that has evolved to support the emergence of multicellularity and different transitions that have accompanied the complexification of multicellular organisms. Understanding the molecular makeup of structures that are conserved throughout evolution is thus of paramount importance.

      The in-silico predicted matrisome of the sea anemone has the potential to become an essential resource for the scientific community to support big data annotation efforts and understand better the evolution of the matrisome and of ECM proteins, an important endeavor to better understand structure/function relationships. This study is also an excellent example of how integrating datasets generated using different -omic modalities can shed light on various aspects of ECM metabolism, from identifying the cell types of origins of matrisome components using scRNA-Seq to studying ECM dynamics using proteomics.

      Weaknesses:

      My concerns pertain to the three following areas of the manuscript:

      (1) In-silico definition of the anemone matrisome using sequence analysis:

      a) While a similar computational pipeline has been applied to predict the matrisome of several model organisms, the authors fail to provide a comprehensive definition of the anemone matrisome: In the text, the authors state the anemone matrisome is composed of "551 proteins, constituting approximately 3% of its proteome (see page 6, line 14), but Figure 1 lists 829 entries as part of the "curated" matrisome, Supplementary Table S1 lists the same 829 entries and the authors state that "Here, we identified 829 ECM proteins that comprise the matrisome of the sea anemone Nematostella vectensis" (see page 17, line 10). Is the sea anemone matrisome composed of 551 or 829 genes? If we refer to the text, the additional 278 entries should not be considered as part of the matrisome, but what is confusing is that some are listed as glycoproteins and the "new_manual_annotation" proposed by the authors and that refer to the protein domains found in these additional proteins suggest that in fact, some could or should be classified as matrisome proteins. For example, shouldn't the two lectins encoded by NV2.3951 and NV2.3157 be classified as matrisome-affiliated proteins? Based on what has been done for other model organisms, receptors have typically been excluded from the "matrisome" but included as part of the "adhesome" for consistency with previously published matrisome; the reviewer is left wondering whether the components classified as "Other" / "Receptor" should not be excluded from the matrisome and moved to a separate "adhesome" list.

      In addition to receptors, the authors identify nearly 70 glycoproteins classified as "Other". Here, does other mean "non-matrisome" or "another matrisome division" that is not core or associated? If the latter, could the authors try to propose a unifying term for these proteins? Unfortunately, since the authors do not provide the reasons for excluding these entries from the bona fide matrisome (list of excluding domains present, localization data), the reader is left wondering how to treat these entries.

      Overall, the study would gain in strength if the authors could be more definitive and, if needed, even propose novel additional matrisome annotations to include the components for now listed as "Other" (as was done, for example, for the Drosophila or C. elegans matrisomes).

      b) It is surprising that the authors are not providing the full currently accepted protein names to the entries listed in Supplementary Table S1 and have used instead "new_manual_annotation" that resembles formal protein names. This liberty is misleading. In fact, the "new_manual_annotation" seems biased toward describing the reason the proteins were positively screened for through sequence analysis, but many are misleading because there is, in fact, more known about them, including evidence that they are not ECM proteins. The authors should at least provide the current protein names in addition to their "new_manual_annotations".

      c) To truly serve as a resource, the Table should provide links to each gene entry in the Stowers Institute for Medical Research genome database used and some sort of versioning (this could be added to columns A, B, or D). Such enhancements would facilitate the assessment of the rigor of the list beyond the manual QC of just a few entries.

      d) Since UniProt is the reference protein knowledge database, providing the UniProt IDs associated with the predicted matrisome entries would also be helpful, giving easy access to information on protein domains, protein structures, orthology information, etc.

      e) In conclusion, at present, the study only provides a preliminary draft that should be more rigorously curated and enriched with more comprehensive and authoritative annotations if the authors aspire the list to become the reference anemone matrisome and serve the community.

      (2) Proteomic analysis of the composition of the mesoglea during the sea anemone life cycle:

      a) The product of 287 of the 829 genes proposed to encode matrisome components was detected by proteomics. What about the other ~550 matrisome genes? When and where are they expressed? The wording employed by the authors (see line 11, page 13) implies that only these 287 components are "validated" matrisome components. Is that to say that the other ~550 predicted genes do not encode components of the ECM? This should be discussed.

      b) Can the authors comment on how they have treated zero TMT values or proteins for which a TMT ratio could not be calculated because unique to one life stage, for example?

      c) Could the authors provide a plot showing the distribution of protein abundances for each matrisome category in the main figure 4? In mammals, the bulk of the ECM is composed of collagens, followed by fibrillar ECM glycoproteins, the other matrisome components being more minor. Is a similar distribution observed in the sea anemone mesoglea?

      d) Prior proteomic studies on the ECM of vertebrate organisms have shown the importance of allowing certain post-translational modifications during database search to ensure maximizing peptide-to-spectrum matching. Such PTMs include the hydroxylation of lysines and prolines that are collagen-specific PTMs. Multiple reports have shown that omitting these PTMs while analyzing LC-MS/MS data would lead to underestimating the abundance of collagens and the misidentification of certain collagens. The authors may want to re-analyze their dataset and include these PTMs as part of their search criteria to ensure capturing all collagen-derived peptides.

      e) The authors should ensure that reviewers are provided with access to the private PRIDE repository so the data deposited can also be evaluated. They should also ensure that sufficient meta-data is provided using the SRDF format to allow the re-use of their LC-MS/MS datasets.

      (3) Supplementary tables:

      The supplementary tables are very difficult to navigate. They would become more accessible to readers and non-specialists if they were accompanied by brief legends or "README" tabs and if the headers were more detailed (see, for example, Table S2, what does "ctrl.ratio_Larvae_rep2" exactly refer to? Or Table S6 whose column headers using extensive abbreviations are quite obscure). Similarly, what do columns K to BX in Supplementary Table S1 correspond to? Without more substantial explanations, readers have no way of assessing these data points.

    3. Reviewer #2 (Public review):

      This work set out to identify all extracellular matrix proteins and associated factors present within the starlet sea anemone Nematostella vectensis at different life stages. Combining existing genomic and transcriptomic datasets, alongside new mass spectometry data, the authors provide a comprehensive description of the Nematostella matrisome. In addition, immunohistochemistry and electron microscopy were used to image whole mount and de-cellularized mesoglea from all life stages. This served to validate the de-cellularization methods used for proteomic analyses, but also resulted in a very nice description of mesoglea structure at different life stages. A previously published developmental cell type atlas was used to identify the cell type specificity of the matrisome, indicating that the core matrisome is predominantly expressed in the gastrodermis, as well as cnidocytes. The analyses performed were rigorous and the results were clear, supporting the conclusions made by the authors.

    4. Reviewer #3 (Public review):

      Summary:

      This manuscript by Bergheim et al investigates the molecular and developmental dynamics of the matrisome, a set of gene products that comprise the extra cellular matrix, in the sea anemone Nematostella vectensis using transcriptomic and proteomic approaches. Previous work has examined the matrisome of the hydra, a medusozoan, but this is the first study to characterize the matrisome in an anthozoan. The major finding of this work is a description of the components of the matrisome in Nematostella, which turns out to be more complex than that previously observed in hydra. The authors also describe remodeling of the extra cellular matrix that occurs in the transition from larva to primary polyp, and from primary polyp to adult. The authors interpret these data to support previously proposed (Steinmetz et al. 2017) homology between the cnidarian endoderm with the bilaterian mesoderm.

      Strengths:

      The data described in this work are comprehensive (but see important considerations of reviewer #1) combining both transcriptome and proteomic interrogation of key stages in the life history of Nematostella and are of value to the community.

      Weaknesses:

      The authors offer numerous evolutionary interpretations of their results that I believe are unfounded. The main problem with extending these results, together with previous results from hydra, into an evolutionary synthesis that aims to reconstruct the matrisome of the ancestral cnidarian is that we are considering data from only two species. I agree with the authors' depiction of hydra as "derived" relative to other medusozoans and see it as potentially misleading to consider the hydra matrisome as an exemplar for the medusozoan matrisome. Given the organismal and morphological diversity of the phylum, a more thorough comparative study that compares matrisome components across a selection of anthozoan and medusozoan species using formal comparative methods to examine hypotheses is required.<br /> Specifically, I question the author's interpretation of the evolutionary events depicted in this statement:

      "The observation that in Hydra both germ layers contribute to the synthesis of core matrisome proteins (Epp et al. 1986; Zhang et al. 2007) might be related to a secondary loss of the anthozoan-specific mesenteries, which represent extensions of the mesoglea into the body cavity sandwiched by two endodermal layers."<br /> Anthozoans and medusozoans are evolutionary sisters. Therefore, secondary loss of "anthozoan-like mesenteries" in hydrozoans is at least as likely as the gain of this character state in anthozoans. By extension, there is no reason to prefer the hypothesis that the state observed in Nematostella, where gastroderm is responsible for the synthesis of the core matrisome components, is the ancestral state of the phylum.<br /> Moreover, the fossil evidence provided in support of this hypotheses (Ou et al. 2022)is not relevant here because the material described in that work is of a crown group anthozoan, which diversified well after the origin of Anthozoa. The phylogenetic structure of Cnidaria has been extensively studied using phylogenomic approaches and is generally well supported(Kayal et al. 2018; DeBiasse et al. 2024). Based on these analyses, anthozoans are not on a "basal" branch, as the authors suggest. The structure of cnidarian phylogeny bifurcates with Anthozoa forming one clade and Medusozoa forming the other. From the data reported by Bergheim and co-workers, it is not possible to infer the evolutionary events that gave rise to the different matrisome states observed in Nematostella (an anthozoan) and hydra (a medusozoan).<br /> Furthermore, I take the observation in Fig 5 that anthozoan matrisomes generally exhibit a higher complexity than other cnidarian species to be more supportive of a lineage specific expansion of matrisome components in the Anthozoa, rather than those components being representative of an ancestral state for Cnidaria. Whatever the implication, I take strong issue with the statement that "the acquisition of complex life cycles in medusozoa, that are distinguished by the pelagic medusa stage, led to a secondary reduction in the matrisome repertoire." There is no causal link in any of the data or analyses reported by Bergheim and co-workers to support this statement and, as stated above, while we are dealing with limited data, insufficient to address this question, it seems more likely to me that the matrisome expanded in anthozoans, contrasting with the authors conclusions. While the discussion raises many interesting evolutionary hypotheses related to the origin of the cnidarian matrisome, which is of vital interest if we are to understand the origin of the bilaterian matrisome, a more thorough comparative analysis, inclusive of a much greater cnidarian species diversity, is required if we are to evaluate these hypotheses.

      DeBiasse MB, Buckenmeyer A, Macrander J, Babonis LS, Bentlage B, Cartwright P, Prada C, Reitzel AM, Stampar SN, Collins A, et al. 2024. A Cnidarian Phylogenomic Tree Fitted With Hundreds of 18S Leaves. Bulletin of the Society of Systematic Biologists [Internet] 3. Available from: https://ssbbulletin.org/index.php/bssb/article/view/9267

      Epp L, Smid I, Tardent P. 1986. Synthesis of the mesoglea by ectoderm and endoderm in reassembled hydra. J Morphol [Internet] 189:271-279. Available from: https://pubmed.ncbi.nlm.nih.gov/29954165/

      Kayal E, Bentlage B, Sabrina Pankey M, Ohdera AH, Medina M, Plachetzki DC, Collins AG, Ryan JF. 2018. Phylogenomics provides a robust topology of the major cnidarian lineages and insights on the origins of key organismal traits. BMC Evol Biol [Internet] 18:1-18. Available from: https://bmcecolevol.biomedcentral.com/articles/10.1186/s12862-018-1142-0

      Ou Q, Shu D, Zhang Z, Han J, Van Iten H, Cheng M, Sun J, Yao X, Wang R, Mayer G. 2022. Dawn of complex animal food webs: A new predatory anthozoan (Cnidaria) from Cambrian. The Innovation 3:100195.

      Steinmetz PRH, Aman A, Kraus JEM, Technau U. 2017. Gut-like ectodermal tissue in a sea anemone challenges germ layer homology. Nature Ecology & Evolution 2017 1:10 [Internet] 1:1535-1542. Available from: https://www.nature.com/articles/s41559-017- 0285-5

      Zhang X, Boot-Handford RP, Huxley-Jones J, Forse LN, Mould AP, Robertson DL, Li L, Athiyal M, Sarras MP. 2007. The collagens of hydra provide insight into the evolution of metazoan extracellular matrices. J Biol Chem [Internet] 282:6792-6802. Available from: https://pubmed.ncbi.nlm.nih.gov/17204477/

    5. Author response:

      We appreciate the effort the reviewers have put into evaluating our work, and will take the opportunity to revise and improve our submission. In response to the reviewer's comments, we will carefully revisit our manuscript to address the concerns they have raised. Specifically, we will ensure that our revised version is coherent with our annotations and public databases, clarify any discrepancy between the investigated proteins and gene models, and re-examine our discussion of the evolutionary implications in light of their suggestions. We are confident that these revisions will strengthen our work and provide a clearer understanding of our research findings.

    1. eLife Assessment

      This important study utilizes single-cell RNA sequencing to reveal the heterogeneity of trans-sialidase-like superfamily gene expression in Trypanosoma cruzi populations. The approach is highly convincing, as it successfully assigns cells to specific developmental forms and highlights the variability in surface protein expression among trypomastigotes. However, while the findings are solid and contribute to the understanding of immune evasion mechanisms, the study would benefit from a more detailed exploration of the regulatory factors governing trans-sialidase expression. Strengthening this aspect would further enhance its impact on researchers studying T. cruzi pathogenesis and host-parasite interactions.

    2. Reviewer #1 (Public review):

      Summary:

      The authors aimed to assess the variability in the expression of surface protein multigene families between amastigote and trypomastigote Trypanosoma cruzi, as well as between individuals within each population. The analysis presented shows higher expression of multigene family transcripts in trypomastigotes compared to amastigotes and that there is variation in which copies are expressed between individual parasites. Notably, they find no clear subpopulations expressing previously characterised trans-sialidase groups. The mapping accuracy to these multicopy genes requires demonstration to confirm this, and the analysis could be extended further to probe the features of the top expressed genes and the other multigene families also identified as variable.

      Strengths:

      The authors successfully process methanol-fixed parasites with the 10x Genomics platform. This approach is valuable for other studies where using live parasites for these methods is logistically challenging.

      Weaknesses:

      The authors describe a single experiment, which lacks controls or complementation with other approaches and the investigation is limited to the trans-sialidase transcripts.

      It would be more convincing to show either bioinformatically or by carrying out a controlled experiment, that the sequencing generated has been mapped accurately to different members of multigene families to distinguish their expression. If mapping to the multigene families is inaccurate, this will impact the transcript counts and downstream analysis.

    3. Reviewer #2 (Public review):

      Summary:

      This manuscript presents a valuable single-cell RNA-seq study on Trypanosoma cruzi, an important human parasite. It investigates the expression heterogeneity of surface proteins, particularly those from the trans-sialidase-like (TcS) superfamily, within amastigote and trypomastigote populations. The findings suggest a previously underappreciated level of diversity in TcS expression, which could have implications for understanding parasite-host interactions and immune evasion strategies. The use of single-cell approaches to delve into population heterogeneity is strong. However, the study does have some limitations that need to be addressed.

      The focus on single-cell transcriptional heterogeneity in surface proteins, especially the TcS family, in T. cruzi is novel. Given the important role of these proteins in parasite biology and host interaction, the findings have potential significance.

      Strengths:

      The key finding of heterogeneous TcS expression in trypomastigotes is well-supported. The analysis comparing multigene families, single-copy genes, and ribosomal proteins highlights the unusual nature of the variation in surface protein-coding genes.

      Weaknesses:

      While the manuscript identifies TcS heterogeneity, the functional implications of the different expression profiles remain speculative. The authors state it may reflect differences in infectivity, but no direct experimental evidence supports this.

      The manuscript lacks any functional validation of the single-cell findings. For instance, do the trypomastigote subpopulations identified based on TcS expression exhibit differences in infectivity, host cell tropism, or immune evasion? Such experiments would greatly strengthen the study.

      The authors identify a subpopulation of TcS genes that are highly expressed in many cells. However, it is unclear if these correspond to previously characterized TcS members with specific functions.<br /> The authors hypothesize that observed heterogeneity may relate to chromatin regulation. However, the study does not directly address these mechanisms. There are interesting connections to be made with what they identify as the colocalization of genes within chromatin folding domains, but the authors do not fully explore this. It would be insightful to address these mechanisms in future work.

      The merging of technical replicates needs further justification and explanation as they were not processed through separate experimental conditions. While barcodes were retained, it would be informative to know how well each technical replicate corresponds with the other. If both datasets were sequenced on the same lane, the inclusion of technical replicates adds noise to the analysis.<br /> While the number of cells sequenced (3192) seems reasonable, it's not clear how much the conclusions are affected by the depth of sequencing. A more detailed description of the sequencing depth and its impact on gene detection would be valuable.

      While most of the methods are clear, the way in which the subsampled gene lists were generated could be more thoroughly described, as some details are not clear for the subsampling of single-copy genes.

      Some of the figures are difficult to interpret. For example, the color scaling in the heatmap of Supplementary Figure 3B is not self-explanatory and it is hard to extract meaningful conclusions from the graph.

    4. Reviewer #3 (Public review):

      The study aimed to address a fundamental question in T. cruzi and Chagas disease biology - how much variation is there in gene expression between individual parasites? This is particularly important with respect to the surface protein-encoding genes, which are mainly from massive repetitive gene families with 100s to 1000s of variant sequences in the genome. There is very little direct evidence for how the expression of these genes is controlled. The authors conducted a single-cell RNAseq experiment of in vitro cultured parasites with a mixture of amastigotes and trypomastigotes. Most of the analysis focused on the heterogeneity of gene expression patterns amongst trypomastigotes. They show that heterogeneity was very high for all gene classes, but surface-protein encoding genes were the most variable. In the case of the trans-sialidase gene family, many sequence variants were only detected in a small minority of parasites. The biology of the parasite (e.g. extensive post-transcriptional regulation) and potential technical caveats (e.g. high dropout rates across the genome) make it difficult to infer what this might mean for actual protein expression on the parasite surface.

      (1) Limit of detection and gene dropouts

      An average of ~1100 genes are detected per parasite which indicates a dropout rate of over 90%. It appears that RNA for the "average" single copy 'core' gene is only detected in around 3% of the parasites sampled (Figure 2c: ~100 / 3192). This may be comparable with some other trypanosome scRNAseq studies, but this still seems to be a major caveat to the interpretation that high cell-to-cell variability in gene expression is explained by biological rather than technical factors. The argument would be more convincing if the dropout rates and expression heterogeneity were minimal for well-known highly expressed genes e.g. tubulin, GAPDH, and ribosomal RNAs. Admittedly, in their Final Remarks, the authors are very cautious in their interpretation, but it would be good to see a more thorough discussion of technical factors that might explain the low detection rates and how these could be tested or overcome in future work.

      (2) Heterogeneity across the board

      The authors focus on the relative heterogeneity in RNA abundance for surface proteins from the multicopy gene families vs core genes. While multicopy gene sequences do show more cell-to-cell variability, the differences (Figure 2D) are roughly average Gini values of 0.99 vs 0.97 (single copy) or 0.95 (ribosomal). Other studies that have applied similar approaches in other systems describe Gini values of < 0.2-0.25 for evenly expressed "housekeeping" genes (PMIDs 29428416, 31784565). Values observed here of >0.9 indicate that the distribution for all gene classes is extremely skewed and so the biological relevance of the comparison is uncertain.

      Nevertheless, this study does provide some tantalising evidence that the expression of surface genes may vary substantially between individual parasites in a single clonal population. The study is also amongst the very first to apply scRNAseq to T. cruzi, so the broader data set will be an important resource for researchers in the field.

    5. Author response:

      We sincerely thank all three reviewers for their time, comments, and valuable suggestions, which will help improve our manuscript. Below, we provide preliminary remarks addressing some of the key issues that have been raised.

      Reviewer 1:

      We agree with the reviewer on the challenge of accurately mapping reads to multigene families. We carefully considered this issue and addressed it by evaluating the performance of multiple aligners using simulated RNA-seq reads. Our results indicate that kallisto performs particularly well in this context, outperforming widely used aligners such as Bowtie2 and STAR. This is likely due to kallisto’s expectation-maximization (EM) algorithm (described in the Materials and Methods section), which employs a probabilistic model to assign reads from similar transcripts. Previous studies have demonstrated the effectiveness of this approach in quantifying highly repetitive sequences, such as transposons (doi.org/10.1093/bioinformatics/btv422). In the revised manuscript, we are considering the inclusion of a supplementary figure to further support the selection of the mapping algorithm.

      Reviewer 2:

      We believe that obtaining experimental evidence on the influence of multiple multigene families would represent a significant advancement in the field. However, we would like to emphasize that this is a short communication centered on a specific and biologically relevant observation within a single multigene family. The manuscript is not intended to comprehensively address all aspects of the experiment but rather to highlight what we consider an important biological phenomenon with potential functional implications.

      The influence of phenotypic heterogeneity and its possible advantages under environmental pressures has been previously proposed for Trypanosoma cruzi, related trypanosomatids, and other biological systems, ranging from bacteria to tumors (Seco-Hidalgo 2015, doi: 10.1098/rsob.150190 and Luzak 2021, doi: 10.1146/annurev-micro-040821-012953, for a comprehensive review on this topic). While the reviewer is correct in noting that our model does not demonstrate a functional role for TcS heterogeneity, the experimental approaches required to address this question in a large multigene family are highly complex and beyond the scope of this study. However, we acknowledge the importance of clarifying that the proposed functional implications remain speculative, so we will revise the manuscript accordingly.

      As the reviewer suggests, in the revised version of the manuscript, we will include additional analyses on the characteristics of frequently expressed TcS genes to identify common features that may explain their expression patterns.

      We appreciate the reviewer’s comments and suggestions regarding the clarity of methodological choices and the explanation of key concepts. Accordingly, we will refine the description of our methodology and ensure that our figures are more intuitive and self-explanatory.

      Reviewer 3:

      We recognize the limitations imposed by gene dropout in our data, as highlighted by the reviewer. In the manuscript, we have aimed to be transparent about this issue and discussed its impact in two separate sections (lines 110–121 and 175–181). To enhance clarity, we will revise these paragraphs to provide a more comprehensive discussion of this limitation. Unfortunately, gene dropout is an inherent limitation of 10x genomics data. Trypanosomatids are not an exception in this regard, and the general metrics of the single-cell RNA-seq data in other reports are equivalent to those obtained in our experiment.

      Despite this important limitation, we believe that our comparative analyses (the contrast between TcS and ribosomal protein expression) provide valuable insights into a biological phenomenon with potential functional relevance for the parasite. Furthermore, we are actively working on generating single-cell RNA-seq data using alternative methodologies that improve gene dropout rates. We anticipate that these future studies will help clarify the extent of the phenomenon described in this work.

    1. eLife Assessment

      The study is useful for advancing spatial transcriptomics through its novel regression-based linear model (glmSMA) that integrates single-cell RNA-seq with spatial reference atlases, though its methodological framework remains incomplete regarding spatial communication applications and feature dependence. The approach demonstrates notable utility by enabling higher-resolution cell mapping across multiple biological systems and spatial platforms compared to existing tools.

    2. Reviewer #1 (Public review):

      Liu et al., present glmSMA, a network-regularized linear model that integrates single-cell RNA-seq data with spatial transcriptomics, enabling high-resolution mapping of cellular locations across diverse datasets. Its dual regularization framework (L1 for sparsity and generalized L2 via a graph Laplacian for spatial smoothness) demonstrates robust performance of their model and offers novel tools for spatial biology, despite some gaps in fully addressing spatial communication.

      Overall, the manuscript is commendable for its comprehensive benchmarking across different spatial omics platforms and its novel application of regularized linear models for cell mapping. I think this manuscript can be improved by addressing method assumptions, expanding the discussion on feature dependence and cell type-specific biases, and clarifying the mechanism of spatial communication.

      The conclusions of this paper are mostly well supported by data, but some aspects of model development and performance evaluation need to be clarified and extended.

      (1) What were the assumptions made behind the model? One of them could be the linear relationship between cellular gene expression and spatial location. In complex biological tissues, non-linear relationships could be present, and this would also vary across organ systems and species. Similarly, with regularization parameters, they can be tuned to balance sparsity and smoothness adequately but may not hold uniformly across different tissue types or data quality levels. The model also seems to assume independent errors with normal distribution and linear additive effects - a simplification that may overlook overdispersion or heteroscedasticity commonly observed in RNA-seq data.

      (2) The performance of glmSMA is likely sensitive to the number and quality of features used. With too few features, the model may struggle to anchor cells correctly due to insufficient discriminatory power, whereas too many features could lead to overfitting unless appropriately regularized. The manuscript briefly acknowledges this issue, but further systematic evaluation of how varying feature numbers affect mapping accuracy would strengthen the claims, particularly in settings where marker gene availability is limited. A simple way to show some of this would be testing on multiple spatial omics (imaging-based) platforms with varying panel sizes and organ systems. Related to this, based on the figures, it also seems like the performance varies by cell type. What are the factors that contribute to this? Variability in expression levels, RNA quantity/quality? Biases in the panel? Personally, I am also curious how this model can be used similarly/differently if we have a FISH-based, high-plex reference atlas. Additional explanation around these points would be helpful for the readers.

      (3) Application 3 (spatial communication) in the graphical abstract appears relatively underdeveloped. While it is clear that the model infers spatial proximities, further explanation of how these mappings translate into insights into cell-cell communication networks would enhance the biological relevance of the findings.

      (4) What is the final resolution of the model outputs? I am assuming this is dictated by the granularity of the reference atlas and the imposed sparsity via the L1 norm, but if there are clear examples that would be good. In figures (or maybe in practice too), cells seem to be assigned to small, contiguous patches rather than pinpoint single-cell locations, which is a pragmatic compromise given the inherent limitations of current spatial transcriptomics technologies. Clarification on the precise spatial scale (e.g., pixel or micrometer resolution) and any post-mapping refinement steps would be beneficial for the users to make informed decisions on the right bioinformatic tools to use.

    3. Reviewer #2 (Public review):

      Summary:

      The author proposes a novel method for mapping single-cell data to specific locations with higher resolution than several existing tools.

      Strengths:

      The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus.

      Weaknesses:

      (1) Although the researchers claim that glmSMA seamlessly accommodates both sequencing-based and image-based spatial transcriptomics (ST) data, their testing primarily focused on sequencing-based ST data, such as Visium and Slide-seq. To demonstrate its versatility for spatial analysis, the authors should extend their evaluation to imaging-based spatial data.

      (2) The definition of "ground truth" for spatial distribution is unclear. A more detailed explanation is needed on how the "ground truth" was established for each spatial dataset and how it was utilized for comparison with the predicted distribution generated by various spatial mapping tools.

      (3) In the analysis of spatial mapping results using intestinal villus tissue, only Figure 3d supports their findings. The researchers should consider adding supplemental figures illustrating the spatial distribution of single cells in comparison to the ground truth distribution to enhance the clarity and robustness of their investigation.

      (4) The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus. However, the original anatomical regions are not displayed, making it difficult to directly compare them with the predicted mapping results. Providing ground truth distributions for each tested tissue would enhance clarity and facilitate interpretation. For instance, in Figure 2a and Supplementary Figures 1 and 2, only the predicted mapping results are shown without the corresponding original spatial distribution of regions in the mouse cortex. Additionally, in Figure 3c, four anatomical regions are displayed, but it is unclear whether the figure represents the original spatial regions or those predicted by glmSMA. The authors are encouraged to clarify this by incorporating ground truth distributions for each tissue.

      (5) The cell assignment results from the mouse hippocampus (Supplementary Figure 6) lack a corresponding ground truth distribution for comparison. DG and CA cells were evaluated solely based on the gene expression of specific marker genes. Additional analyses are needed to further validate the robustness of glmSMA's mapping performance on Slide-seq data from the mouse hippocampus.

      (6) The tested spatial datasets primarily consist of highly structured tissues with well-defined anatomical regions, such as the brain and intestinal villus. It remains unclear whether glmSMA can be effectively applied to tissue types where anatomical regions are not distinctly separated, such as liver tissue. Further evaluation of such tissues would help determine the method's broader applicability.

    4. Reviewer #3 (Public review):

      Summary:

      The authors aim to develop glmSMA, a network-regularized linear model that accurately infers spatial gene expression patterns by integrating single-cell RNA sequencing data with spatial transcriptomics reference atlases. Their goal is to reconstruct the spatial organization of individual cells within tissues, overcoming the limitations of existing methods that either lack spatial resolution or sensitivity.

      Strengths:

      (1) Comprehensive Benchmarking:

      Compared against CellTrek and Novosparc, glmSMA consistently achieved lower Kullback-Leibler divergence (KL divergence) scores, indicating better cell assignment accuracy.

      Outperformed CellTrek in mouse cortex mapping (90% accuracy vs. CellTrek's 60%) and provided more spatially coherent distributions.

      (2) Experimental Validation with Multiple Real-World Datasets:

      The study used multiple biological systems (mouse brain, Drosophila embryo, human PDAC, intestinal villus) to demonstrate generalizability.

      Validation through correlation analyses, Pearson's coefficient, and KL divergence support the accuracy of glmSMA's predictions.

      Weaknesses:

      (1) The accuracy of glmSMA depends on the selection of marker genes, which might be limited by current FISH-based reference atlases.

      (2) glmSMA operates under the assumption that cells with similar gene expression profiles are likely to be physically close to each other in space which not be true under various heterogeneous environments.

    5. Author response:

      Reviewer #1 (Public review):

      Summary:

      Liu et al., present glmSMA, a network-regularized linear model that integrates single-cell RNA-seq data with spatial transcriptomics, enabling high-resolution mapping of cellular locations across diverse datasets. Its dual regularization framework (L1 for sparsity and generalized L2 via a graph Laplacian for spatial smoothness) demonstrates robust performance of their model and offers novel tools for spatial biology, despite some gaps in fully addressing spatial communication.

      Overall, the manuscript is commendable for its comprehensive benchmarking across different spatial omics platforms and its novel application of regularized linear models for cell mapping. I think this manuscript can be improved by addressing method assumptions, expanding the discussion on feature dependence and cell type-specific biases, and clarifying the mechanism of spatial communication.

      The conclusions of this paper are mostly well supported by data, but some aspects of model development and performance evaluation need to be clarified and extended.

      We thank the reviewer for their thoughtful comments. We will clarify the model assumptions and the feature selection process to make it more understandable. To clarify, the performance of glmSMA does not depend on cell type. For some rare cell types, the small number of cells can lead to a drop in performance. To better illustrate our results and reduce cell type-specific biases, we will shuffle and randomly sample the cell types.

      (1) What were the assumptions made behind the model? One of them could be the linear relationship between cellular gene expression and spatial location. In complex biological tissues, non-linear relationships could be present, and this would also vary across organ systems and species. Similarly, with regularization parameters, they can be tuned to balance sparsity and smoothness adequately but may not hold uniformly across different tissue types or data quality levels. The model also seems to assume independent errors with normal distribution and linear additive effects - a simplification that may overlook overdispersion or heteroscedasticity commonly observed in RNA-seq data.

      Thank you for this comment. We acknowledge that the non-linear relationships can be present in complex tissues and may not be fully captured by a linear model. 

      Our choice of a linear model was guided by an investigation of the relationship in the current datasets, which include intestinal villus, mouse brain, and fly embryo.

      There is a linear correlation between expression distance and physical distance [Nitzan et al]. Within a given anatomical structure, cells in closer proximity exhibit more similar expression patterns. In tissues where non-linear relationships are more prevalent—such as the human PDAC sample—our mapping results remain robust. We acknowledge that we have not yet tested our algorithm in highly heterogeneous regions like the liver, and we plan to include such analyses in future work if necessary. Regarding the regularization parameters, we agree that the balance between sparsity and smoothness is sensitive to tissue-specific variation and data quality. In our current implementation, we explored a range of values to find robust defaults.

      (2) The performance of glmSMA is likely sensitive to the number and quality of features used. With too few features, the model may struggle to anchor cells correctly due to insufficient discriminatory power, whereas too many features could lead to overfitting unless appropriately regularized. The manuscript briefly acknowledges this issue, but further systematic evaluation of how varying feature numbers affect mapping accuracy would strengthen the claims, particularly in settings where marker gene availability is limited. A simple way to show some of this would be testing on multiple spatial omics (imaging-based) platforms with varying panel sizes and organ systems. Related to this, based on the figures, it also seems like the performance varies by cell type. What are the factors that contribute to this? Variability in expression levels, RNA quantity/quality? Biases in the panel? Personally, I am also curious how this model can be used similarly/differently if we have a FISH-based, high-plex reference atlas. Additional explanation around these points would be helpful for the readers.

      Thank you for this thoughtful comment. The performance of our method is indeed sensitive to the number and quality of selected features. To optimize feature selection, we employed multiple strategies, including Moran’s I statistic, identification of highly variable genes, and the Seurat pipeline to detect anchor genes linking the spatial transcriptomics data with the reference atlas. The number of selected markers depends on the quality of the data. For high-quality datasets, fewer than 100 markers are typically sufficient for accurate prediction. To address this more clearly, we will revise the manuscript to include detailed descriptions of our feature selection process and demonstrate how varying the number of selected features impacts performance.

      We evaluated our method across diverse tissue types and platforms—including Slide-seq, 10x Visium, and Virtual-FISH—which represent both sequencing-based and imaging-based spatial transcriptomics technologies. Our model consistently achieved strong performance across these settings. It's worth noting that the performance of other methods, such as CellTrek [Wei et al] and novoSpaRc [Nitzan et al], also depends heavily on feature selection. In particular, performance degrades substantially when fewer features are used.

      We do not believe that the observed performance is directly influenced by cell type composition. Major cell types are typically well-defined, and rare cell types comprise only a small fraction of the dataset. For these rare populations, a single misclassification can disproportionately impact metrics like KL divergence due to small sample size. However, this does not necessarily indicate a systematic cell type–specific bias in the mapping. To mitigate this issue, we will implement shuffling and sampling procedures to reduce potential bias introduced by rare cell types.

      (3) Application 3 (spatial communication) in the graphical abstract appears relatively underdeveloped. While it is clear that the model infers spatial proximities, further explanation of how these mappings translate into insights into cell-cell communication networks would enhance the biological relevance of the findings.

      Thank you for this valuable feedback. We agree that further elaboration on the connection between spatial proximity and cell–cell communication would enhance the biological interpretation of our results. While our current model focuses on inferring spatial relationships, we may provide some cell-cell communications in the future.

      (4) What is the final resolution of the model outputs? I am assuming this is dictated by the granularity of the reference atlas and the imposed sparsity via the L1 norm, but if there are clear examples that would be good. In figures (or maybe in practice too), cells seem to be assigned to small, contiguous patches rather than pinpoint single-cell locations, which is a pragmatic compromise given the inherent limitations of current spatial transcriptomics technologies. Clarification on the precise spatial scale (e.g., pixel or micrometer resolution) and any post-mapping refinement steps would be beneficial for the users to make informed decisions on the right bioinformatic tools to use.

      Thank you for the comment. For each cell, our algorithm generates a probability vector that indicates its likely spatial assignment along with coordinate information. We will include the resolution and the number of cells assigned to each spot in future versions. In our framework, each cell is mapped to one or more spatial locations with associated probabilities. Depending on the amount of regularization through L1 and L2 norms, a cell may be localized to a small patch or distributed over a broader domain. For the 10x Visium data, we applied a repelling algorithm to enhance visualization [Wei et al]. If a cell’s original location is already occupied, it is reassigned to a nearby neighborhood to avoid overlap. The users can also see the entire regularization path by varying the penalty terms. 

      Nitzan M, Karaiskos N, Friedman N, Rajewsky N. Gene expression cartography. Nature. 2019;576(7785):132-137. doi:10.1038/s41586-019-1773-3

      Wei, R. et al. (2022) ‘Spatial charting of single-cell transcriptomes in tissues’, Nature Biotechnology, 40(8), pp. 1190–1199. doi:10.1038/s41587-022-01233-1. 

      Reviewer #2 (Public review):

      Summary:

      The author proposes a novel method for mapping single-cell data to specific locations with higher resolution than several existing tools.

      Thank you for recognizing our contribution. Our goal was to develop a method that achieves higher spatial resolution in mapping single-cell data compared to existing tools. We are encouraged by the results and will continue to refine the approach to improve accuracy and generalizability across platforms and tissue types.

      Strengths:

      The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus.

      Thank you for this comment. We believe that evaluating our method across diverse tissue types—such as the mouse cortex, human PDAC, and intestinal villus—demonstrates its robustness and broad applicability. We plan to continue expanding these evaluations to additional tissue contexts and species to further validate the method’s generalizability.

      Weakness:

      (1) Although the researchers claim that glmSMA seamlessly accommodates both sequencing-based and image-based spatial transcriptomics (ST) data, their testing primarily focused on sequencing-based ST data, such as Visium and Slide-seq. To demonstrate its versatility for spatial analysis, the authors should extend their evaluation to imaging-based spatial data.

      Thank you for the comment. We have tested our algorithm on the virtual FISH dataset from the fly embryo, which serves as an example of image-based spatial omics data. However, such datasets often contain a limited number of available genes. To address this, we will conduct additional testing on image-based data if needed. The Allen Brain Atlas provides high-quality ISH data, and we can select specific brain regions from this resource to further evaluate our algorithm if necessary [Lein et al]. Currently, we plan to focus more on the 10x Visium platform, as it supports whole-transcriptome profiling and offers a wide range of tissue samples for analysis.

      (2) The definition of "ground truth" for spatial distribution is unclear. A more detailed explanation is needed on how the "ground truth" was established for each spatial dataset and how it was utilized for comparison with the predicted distribution generated by various spatial mapping tools.

      Thank you for the comment. To clarify how ground truth is defined across different tissues, we provide the following details. Direct ground truth for cell locations is often unavailable in scRNA-seq data due to experimental constraints. To address this, we adopted alternative strategies for estimating ground truth in each dataset:

      - 10x Visium Data: We used the cell type distribution derived from spatial transcriptomics (ST) data as a proxy for ground truth. We then computed the KL divergence between this distribution and our model's predictions for performance assessment.

      - Slide-seq Data: We validated predictions by comparing the expression of marker genes between the reconstructed and original spatial data.

      - Fly Embryo Data: We used predicted cell locations from novoSpaRc as a reference for evaluating our algorithm.

      These strategies allowed us to evaluate model performance even in the absence of direct cell location data. In addition, we can apply multiple evaluation strategies within a single dataset.

      (3) In the analysis of spatial mapping results using intestinal villus tissue, only Figure 3d supports their findings. The researchers should consider adding supplemental figures illustrating the spatial distribution of single cells in comparison to the ground truth distribution to enhance the clarity and robustness of their investigation.

      Thank you for the comment. We will include additional details for this dataset in the supplementary figures. As the intestinal villus is a relatively simple tissue, most existing algorithms performed well on it. For this reason, we did not initially provide extensive details in the main text.

      (4) The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus. However, the original anatomical regions are not displayed, making it difficult to directly compare them with the predicted mapping results. Providing ground truth distributions for each tested tissue would enhance clarity and facilitate interpretation. For instance, in Figure 2a and Supplementary Figures 1 and 2, only the predicted mapping results are shown without the corresponding original spatial distribution of regions in the mouse cortex. Additionally, in Figure 3c, four anatomical regions are displayed, but it is unclear whether the figure represents the original spatial regions or those predicted by glmSMA. The authors are encouraged to clarify this by incorporating ground truth distributions for each tissue.

      Thank you for the comment. To improve visualization, we will include anatomical structures alongside the mapping results in the next version, wherever such structures are available (e.g., mouse brain cortex, human PDAC sample, etc.). Regions will be color-coded to enhance clarity and make the spatial organization easier to interpret.

      (5) The cell assignment results from the mouse hippocampus (Supplementary Figure 6) lack a corresponding ground truth distribution for comparison. DG and CA cells were evaluated solely based on the gene expression of specific marker genes. Additional analyses are needed to further validate the robustness of glmSMA's mapping performance on Slide-seq data from the mouse hippocampus.

      Thank you for the comment. The ground truth for DG and CA cells was not available. To better evaluate the model's performance, we will compute the KL divergence between the original and predicted cell type distributions, following the same approach used for the 10x Visium dataset.

      (6) The tested spatial datasets primarily consist of highly structured tissues with well-defined anatomical regions, such as the brain and intestinal villus. Anatomical regions are not distinctly separated, such as liver tissue. Further evaluation of such tissues would help determine the method's broader applicability.

      Thank you for the comment. We have already tested our algorithm on the fly embryo, where anatomical structures are not well defined or clearly separated. If needed, we can further apply glmSMA to more complex tissues such as the liver. To clarify the role of anatomical structures in our model: glmSMA does not require anatomical information as input. Instead, it leverages a distance matrix between cells to apply L2 norm regularization. Despite the absence of anatomical information, the model still demonstrates strong performance. We will include results to illustrate its effectiveness without anatomical input. Additionally, we plan to evaluate the model on tissues where anatomical regions are not clearly delineated.

      Lein, E., Hawrylycz, M., Ao, N. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168–176 (2007). https://doi.org/10.1038/nature05453

      Reviewer #3 (Public review):

      Summary:

      The authors aim to develop glmSMA, a network-regularized linear model that accurately infers spatial gene expression patterns by integrating single-cell RNA sequencing data with spatial transcriptomics reference atlases. Their goal is to reconstruct the spatial organization of individual cells within tissues, overcoming the limitations of existing methods that either lack spatial resolution or sensitivity.

      Strengths:

      (1) Comprehensive Benchmarking:

      Compared against CellTrek and Novosparc, glmSMA consistently achieved lower Kullback-Leibler divergence (KL divergence) scores, indicating better cell assignment accuracy.

      Outperformed CellTrek in mouse cortex mapping (90% accuracy vs. CellTrek's 60%) and provided more spatially coherent distributions.

      (2) Experimental Validation with Multiple Real-World Datasets:

      The study used multiple biological systems (mouse brain, Drosophila embryo, human PDAC, intestinal villus) to demonstrate generalizability.

      Validation through correlation analyses, Pearson's coefficient, and KL divergence support the accuracy of glmSMA's predictions.

      We thank reviewer #3 for their positive feedback and thoughtful recommendations.

      Weaknesses:

      (1) The accuracy of glmSMA depends on the selection of marker genes, which might be limited by current FISH-based reference atlases.

      We agree that the accuracy of glmSMA is influenced by the selection of marker genes, and that current FISH-based reference atlases may offer a limited gene set. To address this, we incorporate multiple feature selection strategies, including highly variable genes and spatially informative genes (e.g., via Moran’s I), to optimize performance within the available gene space. As more comprehensive reference atlases become available, we expect the model’s accuracy to improve further.

      (2) glmSMA operates under the assumption that cells with similar gene expression profiles are likely to be physically close to each other in space which not be true under various heterogeneous environments.

      While this assumption effectively captures spatial continuity in many cases, we acknowledge that it may not hold across all biological contexts. To address this, we plan to refine our regularization strategy and evaluate the model's performance in heterogeneous tissue regions.

    1. eLife Assessment

      Kwon et al. present an important paper using a novel approach to estimating rotavirus vaccine efficacy using data from a passive surveillance network in the US. They provide convincing evidence to support their conclusion that using the whole genome, rather than previous use of two surface proteins, enhances our understanding of strain-specific vaccine efficacy. These findings have implications for this vaccine specifically as well as type-specific vaccine evaluation more generally.

    2. Reviewer #1 (Public review):

      Summary:

      Kwon et al present a very well-conducted and well-written sieve analysis of rotavirus infections in a passive surveillance network in the US, considering how relative vaccine efficacy changes with genetic distance from the vaccine strains including the whole genome. The results are compelling, supported by a number of sensitivity analyses, and the manuscript is generally easy to follow.

      Strengths:

      (1) The underlying study base, a surveillance network across multiple sites in the US.

      (2) The use of a test-negative design, which is well established for rotavirus, to estimate vaccine efficacy.

      (3) The use of genetic distance to measure differences between infecting and vaccine strains, and the innovative use of k-means clustering to make results more interpretable.

      (4) The secondary and sensitivity analyses that provide additional context and support for the primary findings.

      Weaknesses:

      (1) As identified by the authors, there is a limited sample size for the analysis of RV1 (monovalent rotavirus vaccine).

      (2) Sieve analyses were originally designed for randomized trials, in which setting their key assumptions are more likely to be met. There is little discussion in this paper of how those assumptions might be violated and what effect that might have on the results. The authors have access to some important confounders, but I believe some more discussion on potential biases in this observational study is warranted.

    3. Reviewer #2 (Public review):

      Summary:

      This study introduces a new metric for assessing the efficacy of rotavirus vaccines through the genetic distance clustering of strains. The authors analyzed variations in vaccine protection using whole genome sequencing.

      Strengths:

      Evaluating vaccine efficacy using whole genome sequencing can enhance our understanding of how pathogen evolution influences disease transmission and control.

      Weaknesses:

      While the study proposed a new method for evaluating vaccine efficacy using genetic information, its weaknesses arise from the insufficient evidence that analyses based on whole genome sequencing are more reliable than those that rely solely on VP7 and VP4 genotypes.

      Though most cases received the RV5 vaccine (n=119 compared to n=30 for RV1), Figure 2 and the primary focus of the paper concentrate on RV1, as the authors identified a stronger association with genetic distance.

      Additionally, it is unclear whether the difference between the two groups (j=0 versus j=1) is statistically significant for the analysis based on genetic distance to the RV1 strain, as well as for that based on minimum genetic distance to any of the RV5 vaccine strains. In both cases, the confidence intervals show substantial overlap.

      The authors do not seem to have used a criterion for model selection based on the number of clusters; therefore, k=2 may not represent the optimal number of clusters, particularly in relation to the genetic distance associated with the RV5 vaccine (Figure 1B), which does not appear to show a bimodal distribution.

      Finally, outcomes for RV1 are highly associated with both homotypic and heterotypic antibody responses (Supplemental Figure 10), which have already been shown to impact vaccine effectiveness (The Pediatric Infectious Disease Journal 40(12):p 1135-1143, 2021, doi:10.1097/INF.0000000000003286). Given this strong association, the benefit of using genetic distance is unclear, as the GxPx genotype serves as a good proxy for genetic similarity.

    4. Reviewer #3 (Public review):

      Overall, this is an outstanding paper. It presents a novel approach to estimating rotavirus vaccine efficacy; is clearly written and presented; and has implications for this vaccine specifically as well as type-specific vaccine evaluation more generally. The analytical framework is a creative and there is rigorous use of data and statistical approaches. It has long been argued that rotavirus immunity/vaccine performance operates beyond the scale of G/P genotyping. This paper is the first to demonstrate that convincingly, using data on all 11 viral genes and whole genome sequence analysis. I have only minor comments that I recommend should be addressed.

    5. Author response:

      Public Reviews

      Reviewer #1 (Public review):

      Summary:

      Kwon et al present a very well-conducted and well-written sieve analysis of rotavirus infections in a passive surveillance network in the US, considering how relative vaccine efficacy changes with genetic distance from the vaccine strains including the whole genome. The results are compelling, supported by a number of sensitivity analyses, and the manuscript is generally easy to follow.

      Strengths:

      (1) The underlying study base, a surveillance network across multiple sites in the US.

      (2) The use of a test-negative design, which is well established for rotavirus, to estimate vaccine efficacy.

      (3) The use of genetic distance to measure differences between infecting and vaccine strains, and the innovative use of k-means clustering to make results more interpretable.

      (4) The secondary and sensitivity analyses that provide additional context and support for the primary findings.

      Weaknesses:

      (1) As identified by the authors, there is a limited sample size for the analysis of RV1 (monovalent rotavirus vaccine).

      (2) Sieve analyses were originally designed for randomized trials, in which setting their key assumptions are more likely to be met. There is little discussion in this paper of how those assumptions might be violated and what effect that might have on the results. The authors have access to some important confounders, but I believe some more discussion on potential biases in this observational study is warranted.

      We appreciate the reviewer’s positive comments and the opportunity to discuss the application of sieve analysis in observational vaccine effectiveness studies, contrasting it with its traditional use in clinical trials assessing vaccine efficacy. We fully acknowledge the reviewer's point that sieve analysis was originally developed for, and is most frequently employed in, randomized controlled trials (RCTs).

      Sieve analysis, as defined by Gilbert et al. (2001), has the following core assumptions: (A1) uniform susceptibility to infection for all participants except for vaccine-induced strain-specific effects; (A2) equal exposure (for each strain s = 1,…,K ) distribution between vaccine groups; and (A3), constant strain prevalence. RCTs ensure these through randomization. However, our observational design is vulnerable to violating these assumptions, especially A1 and A3. To address A1 and A3, we adjusted for age (in years), sample collection year, and clinical setting (i.e., outpatient, inpatient, ED), aiming to account for both individual-level and temporal variations.

      A2 is particularly challenging in observational settings. We found that study site was correlated with both vaccination status (main predictor) and the strain distribution, potentially violating A2. However, adjusting for study site reversed the expected association. Upon further reflection, we realized that the site-specific differences in strain distributions likely reflect the population-level effect of vaccination, which we believe outweighs the potential confounding by study site as an independent cause of both individual-level vaccination status and strain distributions irrespective of vaccination. Thus, adjusting for site would have obscured this genuine population-level effect, and therefore we elected not to do so. We will include further discussion of this point in the revised manuscript.

      Our study demonstrates the unique capacity of sieve analysis to disentangle individual- and population-level effects on vaccine effectiveness in observational settings. We will expand on these considerations, including the potential biases inherent to observational studies and the rationale for our analytical choices, within the discussion section of the revised manuscript.

      Reviewer #2 (Public review):

      Summary:

      This study introduces a new metric for assessing the efficacy of rotavirus vaccines through the genetic distance clustering of strains. The authors analyzed variations in vaccine protection using whole genome sequencing.

      Strengths:

      Evaluating vaccine efficacy using whole genome sequencing can enhance our understanding of how pathogen evolution influences disease transmission and control.

      Weaknesses:

      While the study proposed a new method for evaluating vaccine efficacy using genetic information, its weaknesses arise from the insufficient evidence that analyses based on whole genome sequencing are more reliable than those that rely solely on VP7 and VP4 genotypes.

      Though most cases received the RV5 vaccine (n=119 compared to n=30 for RV1), Figure 2 and the primary focus of the paper concentrate on RV1, as the authors identified a stronger association with genetic distance.

      Additionally, it is unclear whether the difference between the two groups (j=0 versus j=1) is statistically significant for the analysis based on genetic distance to the RV1 strain, as well as for that based on minimum genetic distance to any of the RV5 vaccine strains. In both cases, the confidence intervals show substantial overlap

      The authors do not seem to have used a criterion for model selection based on the number of clusters; therefore, k=2 may not represent the optimal number of clusters, particularly in relation to the genetic distance associated with the RV5 vaccine (Figure 1B), which does not appear to show a bimodal distribution.

      Finally, outcomes for RV1 are highly associated with both homotypic and heterotypic antibody responses (Supplemental Figure 10), which have already been shown to impact vaccine effectiveness (The Pediatric Infectious Disease Journal 40(12):p 1135-1143, 2021, doi:10.1097/INF.0000000000003286). Given this strong association, the benefit of using genetic distance is unclear, as the GxPx genotype serves as a good proxy for genetic similarity. 

      We sincerely appreciate reviewer's careful consideration of our manuscript and their constructive suggestions for improvement.

      Regarding the comparison of whole-genome sequencing with traditional VP7/VP4 genotyping, we concur that a more explicit comparison would strengthen our findings. To this end, we plan to incorporate the direct comparison of genetic distance (GD) and genotype-specific vaccine effectiveness (VE) analyses into the main text. Additionally, we will conduct an analysis of VE based on homotypic, partially heterotypic, and fully heterotypic genotype groupings. This will provide a clearer demonstration of the potential added value of GD in refining VE estimates, particularly for future applications. Given the potential for reassortment among the rotavirus gene segments, our analysis highlights that relying solely on the VP7/VP4 genotype can at times be misleading. 

      Regarding k-means clustering, we wish to clarify that the selection of k=2 was not arbitrary. It was determined using the elbow method on the total within-sum-of-squares (using the fviz_nbclust function in the factoextra R package, with n=5000 bootstrapping). While we acknowledge that other methods, such as silhouette and gap statistics, may yield different optimal cluster numbers, we prioritized maximizing group sample sizes. We will explicitly state this model selection criterion within the methods section of the revised manuscript.

      We acknowledge the reviewer’s concern regarding the overlapping confidence intervals and the statistical significance of the differences between the VE for the j=0 and j=1 groups. One way to address this would be to modify our analysis. Instead of two separate logistic regression models (controls vs j=0 cases, and controls vs j=1 cases), we could employ a multinomial logistic regression model with three categories: controls (reference), j=0 cases, and j=1 cases, then conduct Wald test to directly compare the regression slopes for the j=0 and j=1 cases against controls. We intend to explore this approach in the revised manuscript, which will provide a more rigorous assessment of differences in VE by accounting for the relationship between groups within a single model.

      Reviewer #3 (Public review):

      Overall, this is an outstanding paper. It presents a novel approach to estimating rotavirus vaccine efficacy; is clearly written and presented; and has implications for this vaccine specifically as well as type-specific vaccine evaluation more generally. The analytical framework is a creative and there is rigorous use of data and statistical approaches. It has long been argued that rotavirus immunity/vaccine performance operates beyond the scale of G/P genotyping. This paper is the first to demonstrate that convincingly, using data on all 11 viral genes and whole genome sequence analysis. I have only minor comments that I recommend should be addressed.

      We sincerely thank the reviewer for their highly positive assessment of our manuscript. We will carefully address their minor comments and incorporate their recommendations in the revised manuscript, which we believe will further enhance the clarity and impact of our study.

    1. eLife Assessment

      This study reanalyzed previously published scRNA-seq and TCR-seq data to examine the proportion and characteristics of dual-TCR-expressing Treg cells in mice, presenting some useful insights into TCR diversity and immune regulation. However, the evidence is incomplete, particularly with respect to data interpretation, statistical rigor, and the functionality of dual -TCR Treg cells. The study is potentially of interest to immunologists studying T-cell biology.

    2. Reviewer #1 (Public review):

      Summary:

      This study presents findings on dual TCR regulatory T cells (Tregs) using previously published single-cell RNA and TCR sequencing datasets. The authors aimed to quantify dual TCR Tregs in different tissues and analyze their characteristics. Rather than perform the difficult experiments needed to ascertain the functional role of dual receptors, this study relies entirely on scRNA-VDJ-seq data published by two other groups. The findings primarily confirm prior work rather than provide new insights, and the methodology has significant weaknesses that limit the study's impact. We have concerns about the scientific integrity of this work.

      Strengths:

      (1) The use of single-cell RNA and TCR sequencing is appropriate for addressing potential relationships between gene expression and dual TCR.

      (2) The data confirm the presence of dual TCR Tregs in various tissues, with proportions ranging from 10.1% to 21.4%, aligning with earlier observations in αβ T cells.

      (3) Tissue-specific patterns of TCR gene usage are reported, which could be of interest to researchers studying T cell adaptation, although these were more rigorously analyzed in the original works.

      Weaknesses

      (1) Lack of Novelty: The primary findings do not substantially advance our understanding of dual TCR expression, as similar results have been reported previously in other contexts.

      (2) Incomplete Evidence: The claims about tissue-specific differences lack sufficient controls (e.g., comparison with conventional T cells) and functional validation (e.g., cell surface expression of dual TCRs).

      (3) Methodological Weaknesses: The diversity analysis does not account for sample size differences, and the clonal analysis conflates counts and clonotypes, leading to potential misinterpretation.

      (4) Insufficient Transparency: The sequence analysis pipeline is inadequately described, and the study lacks reproducibility features such as shared code and data.

      (5) Weak Gene Expression Analysis: No statistical validation is provided for differential gene expression, and the UMAP plots fail to reveal meaningful clustering patterns.

      (6) A quick online search reveals that the same authors have repeated their approach of reanalysing other scientists' publicly available scRNA-VDJ-seq data in six other publications:

      (1) Peng, Q., Xu, Y. & Yao, X. scRNA+ TCR-seq revealed dual TCR T cells antitumor response in the TME of NSCLC. J Immunother Cancer 12 (2024). https://doi.org:10.1136/jitc-2024-009376

      (2) Wang, H., Li, J., Xu, Y. & Yao, X. scRNA + BCR-seq identifies proportions and characteristics of dual BCR B cells in the peritoneal cavity of mice and peripheral blood of healthy human donors across different ages. Immun Ageing 21, 90 (2024). https://doi.org:10.1186/s12979-024-00493-6

      (3) Xu, Y. et al. scRNA+TCR-seq reveals the pivotal role of dual receptor T lymphocytes in the pathogenesis of Kawasaki disease and during IVIG treatment. Front Immunol 15, 1457687 (2024). https://doi.org:10.3389/fimmu.2024.1457687

      (4) Yuanyuanxu, Qipeng, Qingqingma & Yao, X. scRNA + TCR-seq revealed the dual TCR pTh17 and Treg T cells involvement in autoimmune response in ankylosing spondylitis. Int Immunopharmacol 135, 112279 (2024). https://doi.org:10.1016/j.intimp.2024.112279

      (5) Zhu, L. et al. scRNA-seq revealed the special TCR beta & alpha V(D)J allelic inclusion rearrangement and the high proportion dual (or more) TCR-expressing cells. Cell Death Dis 14, 487 (2023). https://doi.org:10.1038/s41419-023-06004-7

      (6) Zhu, L., Peng, Q., Wu, Y. & Yao, X. scBCR-seq revealed a special and novel IG H&L V(D)J allelic inclusion rearrangement and the high proportion dual BCR expressing B cells. Cell Mol Life Sci 80, 319 (2023). https://doi.org:10.1007/s00018-023-04973-8

      In other words, the approach used here seems to be focused on quick re-analyses of publicly available data without further validation and/or exploration

      Appraisal of the Study's Aims and Conclusions:

      The authors set out to analyze dual TCR Tregs across tissues, but the lack of robust controls, incomplete analyses, and insufficient novelty limit the study's ability to achieve its aims. The results confirm prior findings but do not provide compelling evidence to support the broader claims about the characteristics or significance of dual TCR Tregs.

      Impact and Utility:

      While the study provides a descriptive analysis of dual TCR Tregs, its limited novelty and methodological weaknesses reduce its likely impact on the field. The methods and data could have utility for researchers interested in tissue-specific TCR gene usage, but additional rigor is required to make the findings broadly applicable.

    3. Reviewer #2 (Public review):

      Summary:

      The manuscript, "scRNA+TCR-seq Reveals the Proportion and Characteristics of Dual TCR Treg Cells in Mouse Lymphoid and Non-lymphoid Tissues" by Xu and Peng, et al. investigates whether co-expression of 2 T cell receptor (TCR) clonotypes can be detected in FoxP3+ regulatory CD4+ T cells (Tregs) and if it is associated with identifiable phenotypic effects. This paper presents data reanalyzing publicly available single-cell TCR sequencing and transcriptional analysis, convincingly demonstrating that dual TCR co-expression can be detected in Tregs, both in peripheral circulation as well as among Tregs in tissues. They then compare metrics of TCR diversity between single-TCR and dual TCR Tregs, as well as between Tregs in different anatomic compartments, finding the TCR repertoires to be generally similar though with dual TCR Tregs exhibiting a less diverse repertoire and some moderate differences in clonal expansion in different anatomic compartments. Finally, they examine the transcriptional profile of dual TCR Tregs in these datasets, finding some potential differences in the expression of key Treg genes such as Foxp3, CTLA4, Foxo3, Foxo1, CD27, IL2RA, and Ikzf2 associated with dual TCR-expressing Tregs, which the authors postulate implies a potential functional benefit for dual TCR expression in Tregs.

      Strengths:

      This report examines an interesting and potentially biologically significant question, given recent demonstrations that dual TCR co-expression is a much more common phenomenon than previously appreciated (approximately 15-20% of T cells) and that dual TCR co-expression has been associated with significant effects on the thymic development and antigenic reactivity of T cells. This investigation leverages large existing datasets of single-cell TCRseq/RNAseq to address dual TCR expression in Tregs. The identification and characterization of dual TCR Tregs is rigorously demonstrated and presented, providing convincing new evidence of their existence.

      Weaknesses:

      The existence of dual TCR expression by Tregs has previously been demonstrated in mice and humans (Reference #18 and Tuovinen. 2006. Blood. 108:4063; Schuldt. 2017. J Immunol. 199:33, both omitted from references). The presented results should be considered in the context of these prior important findings.

      This demonstration of dual TCR Tregs is notable, though the authors do not compare the frequency of dual TCR co-expression by Tregs with non-Tregs. This limits interpreting the findings in the context of what is known about dual TCR co-expression in T cells.

      Comparison of gene expression by single- and dual TCR Tregs is of interest, but as presented is difficult to interpret. Statistical analyses need to be performed to provide statistical confidence that the observed differences are true.

      The interpretations of the gene expression analyses are somewhat simplistic, focusing on the single-gene expression of some genes known to have a function in Tregs. However, the investigators miss an opportunity to examine larger patterns of coordinated gene expression associated with developmental pathways and differential function in Tregs (Yang. 2015. Science. 348:589; Li. 2016. Nat Rev Immunol. Wyss. 2016. 16:220; Nat Immunol. 17:1093; Zenmour. 2018. Nat Immunol. 19:291).

    4. Reviewer #3 (Public review):

      Summary:

      This study addressed the TCR pairing types and CDR3 characteristics of Treg cells. By analyzing scRNA and TCR-seq data, it claims that 10-20% of dual TCR Treg cells exist in mouse lymphoid and non-lymphoid tissues and suggests that dual TCR Treg cells in different tissues may play complex biological functions.

      Strengths:

      The study addresses an interesting question of how dual-TCR-expressing Treg cells play roles in tissues.

      Weaknesses:

      This study is inadequate, particularly regarding data interpretation, statistical rigor, and the discussion of the functional significance of Dual TCR Tregs.

      Major Comments:

      (1) Definition of Dual TCR and Validity of Doublet Removal<br /> This study analyzes Treg cells with Dual TCR, but it is not clearly stated how the possibility of doublet cells was eliminated. The authors mention using DoubletFinder for detecting doublets in scRNA-seq data, but is this method alone sufficient?<br /> We strongly recommend reporting the details of doublet removal and data quality assessment in the Supplementary Data.

      (2) Inconsistency in the Proportion of Dual TCR T Cells in the Skin Across Figures<br /> In Figure 3D, the proportion of Dual TCR T cells (A1+A2+B1+B2) in the skin is reported to be very high compared to other tissues. However, in Figure 4C, the proportion appears lower than in other tissues, which may be due to contamination by non-Tregs. The authors should clarify why it was necessary to include non-Tregs as a target for analysis in this study. Additionally, the sensitivity of scRNA-seq and TCR-seq may vary between tissues and may also be affected by RNA quality and sequencing depth in skin samples, so the impact of measurement bias should be assessed.

      (3) Issue of Cell Contamination<br /> In Figure 2A, the data suggest a high overlap between blood, kidney, and liver samples, likely due to contamination. Can the authors effectively remove this effect? If the dataset allows, distinguishing between blood-derived and tissue-resident Tregs would significantly enhance the reliability of the findings. Otherwise, it would be difficult to separate biological signals from contamination noise, making interpretation challenging.

      (4) Inconsistency Between CDR3 Overlap and TCR Diversity<br /> The manuscript states that Single TCR Tregs have a higher CDR3 overlap, but this contradicts the reported data that Dual TCR Tregs exhibit lower TCR diversity (higher 1/DS score). Typically, when TCR diversity is low (i.e., specific clones are concentrated), CDR3 overlap is expected to increase. The authors should carefully address this discrepancy and discuss possible explanations.

      (5) Functional Evaluation of Dual TCR Tregs<br /> This study indicates gene expression differences among tissue-resident Dual TCR T cells, but there is no experimental validation of their functional significance. Including functional assays, such as suppression assays or cytokine secretion analysis, would greatly enhance the study's impact.

      (6) Appropriateness of Statistical Analysis<br /> When discussing increases or decreases in gene expression and cell proportions (e.g., Figure 2D), the statistical methods used (e.g., t-test, Wilcoxon, FDR correction) should be explicitly described. They should provide detailed information on the statistical tests applied to each analysis.

    1. eLife Assessment

      This important study shows, for the first time, the structure and snapshots of the dynamics of the full-length soluble Angiotensin-I converting enzyme dimer. The combination of structural and computational approaches elucidates with convincing evidence the conformational dynamics of the complex and key regions mediating the conformational change. This work provides an example of how conformational heterogeneity can be used to gain insights into protein function.

    2. Reviewer #1 (Public review):

      Summary:

      The authors report four cryoEM structures (2.99 to 3.65 Å resolution) of the 180 kDa, full-length, glycosylated, soluble Angiotensin-I converting enzyme (sACE) dimer, with two homologous catalytic domains at the N- and C-terminal ends (ACE-N and ACE-C). ACE is a protease capable of effectively degrading Aβ. The four structures are C2 pseudo-symmetric homodimers and provide insight into sACE dimerization. These structures were obtained using discrete classification in cryoSPARC and show different combinations of open, intermediate, and closed states of the catalytic domains, resulting in varying degrees of solvent accessibility to the active sites.

      To deepen the understanding of the gradient of heterogeneity (from closed to open states) observed with discrete classification, the authors performed all-atom MD simulations and continuous conformational analysis of cryo-EM data using cryoSPARC 3DVA, cryoDRGN, and RECOVAR. cryoDRGN and cryoSPARC 3DVA revealed coordinated open-closed transitions across four catalytic domains, whereas RECOVAR revealed independent motion of two ACE-N domains, also observed with cryoSPARC-focused classification. The authors suggest that the discrepancy in the results of the different methods for continuous conformational analysis in cryo-EM could result from different approaches used for dimensionality reduction and trajectory generation in these methods.

      Strengths:

      This is an important study that shows, for the first time, the structure and the snapshots of the dynamics of the full-length sACE dimer. Moreover, the study highlights the importance of combining insights from different cryo-EM methods that address questions difficult or impossible to tackle experimentally while lacking ground truth for validation.

      Weaknesses:

      The open, closed, and intermediate states of ACE-N and ACE-C in the four cryo-EM structures from discrete classification were designated quantitatively (based on measured atomic distances on the models fitted into cryo-EM maps, Figure 2D). Unfortunately, atomic models were not fitted into cryo-EM maps obtained with cryoSPARC 3DVA, cryoDRGN, and RECOVAR, and the open/closed states in these cases were designated based on qualitative analysis. As the authors clearly pointed out, there are many other methods for continuous conformational heterogeneity analysis in cryo-EM. Among these methods, some allow analyzing particle images in terms of atomic models, like MDSPACE (Vuillemot et al., J. Mol. Biol. 2023, 435:167951), which result in one atomic model per particle image and can help in analyzing cooperativity of domain motions through measuring atomic distances or angular differences between different domains (Valimehr et al., Int. J. Mol. Sci. 2024, 25: 3371). This could be discussed in the article.

    3. Reviewer #2 (Public review):

      Summary:

      The manuscript presents a valuable contribution to the field of ACE structural biology and dynamics by providing the first complete full-length dimeric ACE structure in four distinct states. The study integrates cryo-EM and molecular dynamics simulations to offer important insights into ACE dynamics. The depth of analysis is commendable, and the combination of structural and computational approaches enhances our understanding of the protein's conformational landscape. However, the strength of evidence supporting the conclusions needs refinement, particularly in defining key terms, improving structural validation, and ensuring consistency in data analysis. Addressing these points through major revisions will significantly improve the clarity, rigor, and accessibility of the study to a broader audience, allowing it to make a stronger impact in the field.

      Strengths:

      The integration of cryo-EM and MD simulations provides valuable insights into ACE dynamics, showcasing the authors' commitment to exploring complex aspects of protein structure and function. This is a commendable effort, and the depth of analysis is appreciated.

      Weaknesses:

      Several aspects of the manuscript require further refinement to improve clarity and scientific rigor as detailed in my recommendations for the authors.

    4. Reviewer #3 (Public review):

      Summary:

      Mancl et al. report four Cryo-EM structures of glycosylated and soluble Angiotensin-I converting enzyme (sACE) dimer. This moves forward the structural understanding of ACE, as previous analysis yielded partially denatured or individual ACE domains. By performing a heterogeneity analysis, the authors identify three structural conformations (open, intermediate open, and closed) that define the openness of the catalytic chamber and structural features governing the dimerization interface. They show that the dimer interface of soluble ACE consists of an N-terminal glycan and protein-protein interaction region, as well as C-terminal protein-protein interactions. Further heterogeneity mining and all-atom molecular dynamic simulations show structural rearrangements that lead to the opening and closing of the catalytic pocket, which could explain how ACE binds its substrate. These studies could contribute to future drug design targeting the active site or dimerization interface of ACE.

      Strengths:

      The authors make significant efforts to address ACE denaturation on cryo-EM grids, testing various buffers and grid preparation techniques. These strategies successfully reduce denaturation and greatly enhance the quality of the structural analysis. The integration of cryoDRGN, 3DVA, RECOVAR, and all-atom simulations for heterogeneity analysis proves to be a powerful approach, further strengthening the overall experimental methodology.

      Weaknesses:

      In general, the findings are supported by experimental data, but some experimental details and approaches could be improved. For example, CryoDRGN analysis is limited to the top 5 PCA components for ease of comparison with cryoSPARC 3DVA, but wouldn't an expansion to more components with CryoDRGN potentially identify further conformational states? The authors also say that they performed heterogeneity analysis on both datasets but only show data for one. The results for the first dataset should be shown and can be included in supplementary figures. In addition, the authors mention that they were not successful in performing cryoSPARC 3DFLex analysis, but they do not show their data or describe the conditions they used in the methods section. These data should be added and clearly described in the experimental section.

      Some cryo-EM data processing details are missing. Please add local resolution maps, box sizes, and Euler angle distributions and reference the initial PDB model used for model building.

    1. eLife Assessment

      This study presents a useful contribution to understanding zinc regulation of sperm physiology, specifically its inhibitory effects on the sperm-specific potassium channel Slo3. However, the evidence supporting the claims is incomplete, as critical experimental controls are lacking, key mechanistic aspects remain insufficiently explored, and experimental descriptions are often inadequate, making it difficult to fully assess the findings. Strengthening the study with additional electrophysiological recordings in sperm cells, improved imaging controls, and clearer methodological descriptions would enhance its impact and rigor.

    2. Reviewer #1 (Public review):

      Summary:

      In their manuscript, Andriani et al. show intracellular zinc is exported from sperm during capacitation and suppresses the alkalinization-induced hyperpolarization in sperm. Intracellular zinc inhibits Slo3 current, which is enhanced by the co-expression of gamma subunit Lrrc52. Computational studies reveal that the Zn binding site on mSlo3 is located near E169 and E205, which are involved in the sustained zinc inhibition of mSlo3 current. The authors propose that intracellular zinc plays a key role in sperm capacitation by inhibiting the Slo3 channel.

      Strengths:

      Overall, the work appears well-designed (e.g., oocyte patch-clamp experiments), and clearly presented. Three-dimensional structural modeling and flooding simulations are executed.

      Weaknesses:

      The simple mutagenesis analysis of E169 and E205 showed partial abolishment, but the molecular mechanism by which zinc inhibits Slo3 current is not yet fully shown. The authors should consider performing more extensive experiments, such as creating double mutants or combination mutants involving other residues. Additionally, could other mechanisms explain the role of zinc in regulating the Slo3 current?

      While elucidating the mechanism of Slo3 is interesting, there is substantial literature indicating how zinc regulates channel functions at a molecular level. Given this, the manuscript should provide a deeper understanding by clearly elucidating the molecular mechanism of the regulation of Slo3 current by zinc.<br /> The manuscript includes no experimental data on the mechanism of intracellular zinc export during sperm capacitation, despite being crucial for the regulation of sperm function.

    3. Reviewer #2 (Public review):

      Summary:

      In this paper, Andriani and colleagues are examining the potential role of Zn flux in sperm and its effect on Slo3 channels. This is an interesting question that is likely critical to how sperm function properly and Slo3 channels are a possible candidate for a downstream molecule that is impacted by Zn. In this paper, the authors use Zn imaging, sperm motility assays, and electrophysiology to show that Zn flux impacts sperm function. They then go on to look at the impact Zn has on Slo3 current and propose a binding site based on MD simulations. While the ideas are interesting, the experiments are not well described in many places making understanding the results very difficult. In addition, critical controls are missing throughout the paper.

      Strengths:

      The question of how Zn flux impacts membrane potential and sperm motility is an important one. Moreover, Slo3 presents an interesting candidate or the target of Zn regulation. The combination of methods used here also has the potential to uncover mechanisms of Zn regulation of Slo3.

      Weaknesses:

      Much of the paper lacks experimental description which makes interpretation quite difficult, or a detailed discussion is missing. Examples include:

      (1) Figure 1, particularly the Zn imaging, is not sufficiently described. How is the fluorescence intensity measured? A representative ROI? The whole tail and head? Are the sperm immobile? If not, there is evidence that motion artifacts can significantly distort these sorts of measures from Calcium measurements in Cilia. Were there controls done? Is the small amount of Zn seen in the tail above the background?

      (2) The second half of Figure 1 is also not well described. What is the extracellular solution in the recordings? When you apply the Zn ionophore, do you expect influx or efflux? I assume efflux is based on the conclusions but this should be discussed explicitly.

      (3) Figure 2H labels the Y axis, "normalized current". Normalized to what? Why do neither of the curves end at 1? A better description of what this figure represents is needed.

      (4) The alpha fold simulations are not well described. How many Zn binding sites were found? Are all of the histidine mutations in Figure 4 Supplement 1 the ones that were found?

      (5) There is no discussion of physiological intracellular Zn concentration. How much Zn is inside the sperm? How much if likely Free vs buffered? Is 100uM a reasonable physiological concentration?

      There are a number of areas where the interpretation is not well supported by the data including:

      (6) You say in the Figure 4 supplement, that "we did not observe any significant decrease in the percentage of current inhibition." But that is a pretty misleading statement. There are large changes (increases) in the amount of zinc inhibition. These might be allosteric changes but I don't think you can safely eliminate these as relevant Zn binding sites. Also, some of these mutations appear to allow at least some unbinding of Zn.

      (7) Following up on the above point, it seems unfair to conclude that the D162S, E169A, and E205 mutants are part of the inhibitory binding site for Zn when the mutation has no effect on inhibition and only an effect on the washout. The mutations on the intracellular side also had an impact on the washout so it seems equally likely that they are the critical residues based on your data.

      (8) Nowhere in the paper do you make the specific link between Zn flux and membrane hyperpolariation via Slo3. You show that Zn flux changes the ability of the sperm to hyperpolarize and you show that Slo3 is inhibited by Zn but the connection between the two is not demonstrated. There appears to be a specific Slo3 blocker. If you use this in sperm, do you no longer see the Zn effect?

      (9) In the second half of Figure 1, the authors suggest that there is "no hyperpolization in 100uM Zn. That is not really true. It is reduced but not absent.

      (10) The claim that Lrcc52 with Slo3 shows a higher current inhibition at pH 7.5 than pH 8 is not well supported because there are only 3 replicates in the 7.5 case. In addition, the claim is made in the test that 100uM ZnCl2 "already inhibited mSlo3+Lrcc52 at pH7.5", contrasted with mSlo3 alone, is not tested statistically.

      In a number of places, better controls are needed.

      (11) How specific is this effect for Zn? Mg2+, for instance, is also a divalent cation that is in the hundreds of uM range inside the cell. Does it exert the same effect? Each ion certainly has unique preferred coordination geometries, does your predicted binding with MD show what you might expect for tetrahedral coordination with Zn? Did you test other divalent cations functionally or in silicon?

      (12) For the VCF experiments, a significantly higher concentration of Zn was used (10mM). What is the reason for this? There is no discussion of how much a "puff" is. Assuming you are using the RNA injector it is probably on the order of 50nL or less. Assuming the volume of an oocyte is 1uL that would argue that the final concentration is 500uM or higher. But this is also complicated by potential local effects of high Zn at the injection site, artifacts of injecting that much metal, and the fact that a great deal of the Zn will likely be bound to other things inside the cell. Better controls are needed for this experiment.

    4. Reviewer #3 (Public review):

      Summary:

      The study titled "Zinc is a Key Regulator of the Sperm-Specific K+ Channel (Slo3) Function" aims to investigate the role of intracellular zinc in sperm capacitation and its regulation of the sperm-specific Slo3 potassium channel. Capacitation is a crucial physiological process that enables sperm to fertilize an egg, and membrane hyperpolarization through Slo3 activation is a well-established event in this process. The authors propose that intracellular zinc dynamically decreases during capacitation and inhibits Slo3-mediated K⁺ currents, thereby playing a regulatory role in sperm function.

      Strengths:

      (1) Novel Contribution to Sperm Physiology.

      The study provides new insights into how zinc dynamics contribute to sperm capacitation, specifically through its direct inhibition of Slo3 activity.<br /> Previous research has focused primarily on extracellular zinc's effect on sperm function; this work expands the discussion to intracellular zinc regulation, an area with limited prior investigation.

      (2) Strong Electrophysiological Evidence.

      The study employs inside-out patch-clamp recordings in Xenopus oocytes to demonstrate zinc's direct inhibition of Slo3 currents.<br /> The observed slow dissociation of zinc from Slo3 suggests a long-lasting regulatory effect, adding to the understanding of ion channel modulation in sperm cells.

      (3) Molecular Mechanistic Insights

      Using Molecular Dynamics (MD) simulations and mutagenesis, the authors identify potential zinc-binding sites within Slo3's voltage-sensing domain (VSD), particularly E169 and E205.

      These computational predictions are supported by electrophysiological recordings, strengthening the argument that zinc directly binds and inhibits Slo3.

      (4) Physiological Relevance and Functional Implications

      The study suggests that zinc inhibition of Slo3 could contribute to sperm motility regulation during capacitation.

      The authors provide sperm motility assays as supporting evidence, showing that zinc chelation affects motility only after capacitation has begun, suggesting a dynamic role of intracellular zinc in the capacitation process.

      Weaknesses:

      While the study presents compelling electrophysiological data and molecular insights, there are several critical gaps that must be addressed before fully supporting the physiological relevance of the findings.

      (1) The authors should measure the effects in sperm cells using the patch-clamp technique to directly record Slo3 currents. By normalizing Slo3 currents to cell capacitance at different intracellular zinc concentrations, the authors can quantitatively assess the extent of Slo3 inhibition by zinc and strengthen the physiological relevance of their findings.

      (2) Lack of Controls in Non-Capacitated Sperm

      The claim that zinc is exported from sperm during capacitation needs stronger experimental validation.

      The authors did not include a control group of non-capacitated sperm in key fluorescence imaging experiments, making it difficult to confirm that the observed zinc decrease is capacitation-specific rather than a general zinc redistribution process.

      To strengthen this conclusion, experiments should be performed in non-capacitating conditions to determine whether intracellular zinc levels remain unchanged.

      (3) Unclear Role of Zinc in Physiological Capacitation

      The study clearly demonstrates zinc inhibition of Slo3 but does not sufficiently establish how this affects capacitation at a functional level.

      Additional motility and capacitation markers should be analyzed to confirm that zinc influences sperm behavior beyond Slo3 inhibition.

      (4) Insufficient Data on Zinc-Slo3 Specificity

      The authors should consider using quinidine, a known washable Slo3 inhibitor, to confirm that zinc acts specifically on Slo3 channels rather than other endogenous ion channels.

      The study would benefit from including washout controls in the inside-out patch-clamp recordings, as seen in Figure 3-Supplement 1, to confirm that zinc inhibition is reversible or long-lasting.

      (5) Missing Discussion of Zinc's Role in CatSper Regulation

      The study focuses solely on Slo3 but does not mention CatSper, the principal Ca²⁺ channel essential for sperm capacitation.

      Zinc has been reported to inhibit CatSper activity, which could significantly impact sperm function.

      The discussion should address whether zinc's effect on Slo3 represents a broader regulatory mechanism influencing multiple ion channels during capacitation.

      Final Assessment

      This work presents important findings on zinc regulation of Slo3 channels, supported by strong electrophysiological and molecular analyses. However, the physiological relevance of these findings remains unclear due to missing controls, and needs additional functional assays. Addressing these issues would significantly enhance the manuscript's scientific rigor and impact.

    1. eLife Assessment

      This important study addresses a topic that is frequently discussed in the literature but is under-assessed, namely correlations among genome size, repeat content, and pathogenicity in fungi. Contrary to previous assertions, the authors found that repeat content is not associated with pathogenicity. Rather, pathogenic lifestyle was found to be better explained by the number of protein-coding genes, with other genomic features associated with insect association status. While the results are considered solid, confidence in the results would be deepened if the authors were to comprehensively account for potential biases stemming from the underlying data quality of the analyzed genomes.

    2. Reviewer #1 (Public review):

      Summary:

      The manuscript "Lifestyles shape genome size and gene content in fungal pathogens" by Fijarczyk et al. presents a comprehensive analysis of a large dataset of fungal genomes to investigate what genomic features correlate with pathogenicity and insect associations. The authors focus on a single class of fungi, due to the diversity of lifestyles and availability of genomes. They analyze a set of 12 genomic features for correlations with either pathogenicity or insect association and find that, contrary to previous assertions, repeat content does not associate with pathogenicity. They discover that the number of protein-coding genes, including the total size of non-repetitive DNA does correlate with pathogenicity. However, unique features are associated with insect associations. This work represents an important contribution to the attempts to understand what features of genomic architecture impact the evolution of pathogenicity in fungi.

      Strengths:

      The statistical methods appear to be properly employed and analyses thoroughly conducted. The manuscript is well written and the information, while dense, is generally presented in a clear manner.

      Weaknesses:

      My main concerns all involve the genomic data, how they were annotated, and the biases this could impart to the downstream analyses. The three main features I'm concerned with are sequencing technology, gene annotation, and repeat annotation.

      The collection of genomes is diverse and includes assemblies generated from multiple sequencing technologies including both short- and long-read technologies. Not only has the impact of the sequencing method not been evaluated, but the technology is not even listed in Table S1. From the number of scaffolds it is clear that the quality of the assemblies varies dramatically. This is going to impact many of the values important for this study, including genome size, repeat content, and gene number. Additionally, since some filtering was employed for small contigs, this could also bias the results.

      I have considerable worries that the gene annotation methods could impart biases that significantly affect the main conclusions. Only 5 reference training sets were used for the Sordariomycetes and these are unequally distributed across the phylogeny. Augusts obviously performed less than ideally, as the authors reported that it under-annotated the genomes by 10%. I suspect it will have performed worse with increasing phylogenetic distance from the reference genomes. None of the species used for training were insect-associated, except for those generated by the authors for this study. As this feature was used to split the data it could impact the results. Some major results rely explicitly on having good gene annotations, like exon length, adding to these concerns. Looking manually at Table S1 at Ophiostoma, it does seem to be a general trend that the genomes annotated with Magnaporthe grisea have shorter exons than those annotated with H294. I also wonder if many of the trends evident in Figure 5 are also the result of these biases. Clades H1 and G each contain a species used in the training and have an increase in genes for example.

      Unfortunately, the genomes available from NCBI will vary greatly in the quality of their repeat masking. While some will have been masked using custom libraries generated with software like Repeatmodeler, others will probably have been masked with public databases like repbase. As public databases are again biased towards certain species (Fusarium is well represented in repbase for example), this could have significant impacts on estimating repeat content. Additionally, even custom libraries can be problematic as some software (like RepeatModeler) will include multicopy host genes leading to bona fide genes being masked if proper filtering is not employed. A more consistent repeat masking pipeline would add to the robustness of the conclusions.

      To a lesser degree, I wonder what impact the use of representative genomes for a species has on the analyses. Some species vary greatly in genome size, repeat content, and architecture among strains. I understand that it is difficult to address in this type of analysis, but it could be discussed.

    3. Reviewer #2 (Public review):

      Summary:

      In this paper, the authors report on the genomic correlates of the transition to the pathogenic lifestyle in Sordariomycetes. The pathogenic lifestyle was found to be better explained by the number of genes, and in particular effectors and tRNAs, but this was modulated by the type of interacting host (insect or not insect) and the ability to be vectored by insects.

      Strengths:

      The main strength of this study lies in the size of the dataset, and the potentially high number of lifestyle transitions in Sordariomycetes.

      Weaknesses:

      The main strength of the study is not the clarity of the conclusions.

      (1) This is due firstly to the presentation of the hypotheses. The introduction is poorly structured and contradictory in some places. It is also incomplete since, for example, fungus-insect associations are not mentioned in the introduction even though they are explicitly considered in the analyses.

      (2) The lack of clarity also stems from certain biases that are challenging to control in microbial comparative genomics. Indeed, defining lifestyles is complicated because many fungi exhibit different lifestyles throughout their life cycles (for instance, symbiotic phases interspersed with saprotrophic phases). In numerous fungi, the lifestyle referenced in the literature is merely the sampling substrate (such as wood or dung), which doesn't mean that this substrate is a crucial aspect of the life cycle. This issue is discussed by the authors, but they do not eliminate the underlying uncertainties.

    4. Reviewer #3 (Public review):

      Summary:

      This important study combines comparative genomics with other validation methods to identify the factors that mediate genome size evolution in Sordariomycetes fungi and their relationship with lifestyle. The study provides insights into genome architecture traits in this Ascomycete group, finding that, rather than transposons, the size of their genomes is often influenced by gene gain and loss. With an excellent dataset and robust statistical support, this work contributes valuable insights into genome size evolution in Sordariomycetes, a topic of interest to both the biological and bioinformatics communities.

      Strengths:

      This study is complete and well-structured.

      Bioinformatics analysis is always backed by good sampling and statistical methods. Also, the graphic part is intuitive and complementary to the text.

      Weaknesses:

      The work is great in general, I just had issues with the Figure 1B interpretation.

      I struggled a bit to find the correspondence between this sentence: "Most genomic features were correlated with genome size and with each other, with the strongest positive correlation observed between the size of the assembly excluding repeats and the number of genes (Figure 1B)." and the Figure 1B. Perhaps highlighting the key p values in the figure could help.

    1. eLife Assessment

      The IBEX Knowledge-Base is an important tool that will enhance scientific collaboration by providing a centralized, community-driven resource for immunofluorescence imaging and reagent validation. Its detailed use cases, open-source design, and transparent reporting offer compelling evidence of its broad utility and impact in the life sciences. Overall, the resource sets a high standard as a blueprint for future community initiatives in reproducibility and standardization.

    2. Reviewer #1 (Public review):

      IBEX Knowledge Database

      Here, Anidi and colleagues present the IBEX knowledge base. A community tool developed to centralize knowledge and help its adoption by more users. The authors have done a fantastic job, and there is careful consideration of the many aspects of data management and FAIR principles. The manuscript needs no further work, as it is very well written and has detailed descriptions for data contribution as well as describing the KB itself. Overall, it is a great initiative, especially the aim to inform about negative data and non-recommended reagents, which will positively affect the user community and scientific reproducibility.

      As such amount of work has been put into developing this community tool, it would be worth thinking about how it could serve other multiplex-immunofluorescence methods (such as immunoSABER, 4i, etc). Adding an extra tab where the particular method that uses those reagents is mentioned. This would also help as IBEX itself and related methods evolve in the future.

      It has a rather minimal description of the software. In particular, there is software that has not been developed for IBEX specifically but that could be used for IBEX datasets (ASHLAR, WSIReg, VALIS, WARPY, and QuPath, etc). It would be nice if there was mention of those.

      There is a concern about how the negative data information will be added, as no publication or peer-review process can back it up. Perhaps the particular conditions of the experiment should be very well described to allow future users to assess the validity. The proposed scheme where a reagent can be validated or recommended against by up to 4 different labs should be good. It may be good to make sure that researchers who validate belong to different labs and are not only different ORCID that belong to the same group. Similar to making a case of recommendations against a reagent.

      It is very interesting to keep track of the protocol versions used. Perhaps users should be able to validate independent versions and it will be important to know how information is kept.

      The final point I would make is that the need to form a GitHub repository may deter some people from submitting data. For sporadic contributions, authors could think that users could either reach out to main developers and/or provide a submission form that can help less experienced users of command-line and GitHub programming, but still promote the contribution from the community.

      I am keen to see how the KB evolves and how it helps disseminate the use of this and other great techniques.

    3. Reviewer #2 (Public review):

      Summary:

      The paper introduces the IBEX Knowledge-Base (KB), a shared online resource designed to help scientists working with immunofluorescence imaging. It acts as a central hub where researchers can find and share information about reagents, protocols, and imaging methods. The KB is not static like traditional publications; instead, it evolves as researchers contribute new findings and refinements. A key highlight is that it includes results of both successful and unsuccessful experiments, helping scientists avoid repeating failed experiments and saving time and resources. The platform is built on open-access tools ensuring that the information remains available to everyone. Overall, the KB aims to collaboratively accelerate research, improve reproducibility, and reduce wasted effort in imaging experiments.

      Strengths:

      (1) The IBEX KB is built entirely on open-source tools, ensuring accessibility and long-term sustainability. This approach aligns with FAIR data principles and ensures that the KB remains adaptable to future advancements.

      (2) The KB also follows strict data organization standards, ensuring that all information about reagents and protocols is clearly documented and easy to find with little ambiguity.

      (3) The KB allows scientists to report both positive and negative results, reducing duplication of effort and speeding up the research process.

      (4) The KB is helpful for all researchers, but even more so for scientists in resource-limited settings. It provides guidance on finding affordable alternatives to expensive or discontinued reagents, making it easier for researchers with fewer resources to perform high-quality experiments.

      (5) The KB includes a community discussion forum where scientists can ask for advice, share troubleshooting tips, and collaborate with others facing similar challenges.

      Weaknesses:

      (1) The potential impact of IBEX KB is very clear. However, the paper would benefit by also discussing more on KB maintenance and outreach, and how higher participation could be incentivized.

      (2) Use of resources like GitHub may limit engagement from non-coding members of the scientific community. Will there be alternative options like a user-friendly web interface to contribute more easily?

    4. Reviewer #3 (Public review):

      Summary:

      The authors have developed an interactive knowledge-base that uses crowdsourcing information on antibodies and reagents for immunofluorescence imaging.

      Strengths:

      The authors provide an extremely relevant and needed interphase for a community-based IF reagent and protocol knowledgebase, and a well-built interface. All the links on their website work, the information provided, reagents, datasets, videos, and protocols are very informative. The instructions for the community researchers to contribute are clear and they provide detailed instructions on how to technically proceed.

      Weaknesses:

      Reporting of the validation of antibodies could be improved. To increase public participation they suggest reducing the amount of details that one needs to submit to claim that something does not work. However, in our experience, this information is critical to be shared with the community.

    1. eLife Assessment

      This manuscript demonstrates that Oct4 overexpression synergizes with Notch inhibition (Rbpj knockout) to promote the conversion of adult murine Müller glia (MG) into bipolar cells. These findings are important as the authors used rigorous genetic lineage tracing (GLAST-CreER; Sun-GFP) to confirm that neurogenesis indeed originates from MGs, addressing a key issue in the field. The single-cell multiomic analyses are convincing, and while functional studies of MG-derived bipolar cells would strengthen the conclusions, they are beyond the scope of this study.

    2. Reviewer #1 (Public review):

      Summary:

      In this study, Le et al.. aimed to explore whether AAV-mediated overexpression of Oct4 could induce neurogenic competence in adult murine Müller glia, a cell type that, unlike its counterparts in cold-blooded vertebrates, lacks regenerative potential in mammals. The primary goal was to determine whether Oct4 alone, or in combination with Notch signaling inhibition, could drive Müller glia to transdifferentiate into bipolar neurons, offering a potential strategy for retinal regeneration.

      The authors demonstrated that Oct4 overexpression alone resulted in the conversion of 5.1% of Müller glia into Otx2+ bipolar-like neurons by five weeks post-injury, compared to 1.1% at two weeks. To further enhance the efficiency of this conversion, they investigated the synergistic effect of Notch signaling inhibition by genetically disrupting Rbpj, a key Notch effector. Under these conditions, the percentage of Müller glia-derived bipolar cells increased significantly to 24.3%, compared to 4.5% in Rbpj-deficient controls without Oct4 overexpression. Similarly, in Notch1/2 double-knockout Müller glia, Oct4 overexpression increased the proportion of GFP+ bipolar cells from 6.6% to 15.8%.

      To elucidate the molecular mechanisms driving this reprogramming, the authors performed single-cell RNA sequencing (scRNA-seq) and ATAC-seq, revealing that Oct4 overexpression significantly altered gene regulatory networks. They identified Rfx4, Sox2, and Klf4 as potential mediators of Oct4-induced neurogenic competence, suggesting that Oct4 cooperates with endogenously expressed neurogenic factors to reshape Müller glia identity.

      Overall, this study aimed to establish Oct4 overexpression as a novel and efficient strategy to reprogram mammalian Müller glia into retinal neurons, demonstrating both its independent and synergistic effects with Notch pathway inhibition. The findings have important implications for regenerative therapies as they suggest that manipulating pluripotency factors in vivo could unlock the neurogenic potential of Müller glia for treating retinal degenerative diseases.

      Strengths:

      (1) Novelty: The study provides compelling evidence that Oct4 overexpression alone can induce Müller glia-to-bipolar neuron conversion, challenging the conventional view that mammalian Müller glia lacks neurogenic potential.

      (2) Technological Advances: The combination of Muller glia-specific labeling and modifying mouse line, AAV-GFAP promoter-mediated gene expression, single-cell RNA-seq, and ATAC-seq provides a comprehensive mechanistic dissection of glial reprogramming.

      (3) Synergistic Effects: The finding that Oct4 overexpression enhances neurogenesis in the absence of Notch signaling introduces a new avenue for retinal repair strategies.

      Weaknesses:

      (1) In this study, the authors did not perform a comprehensive functional assessment of the bipolar cells derived from Müller glia to confirm their neuronal identity and functionality.

      (2) Demonstrating visual recovery in a bipolar cell-deficiency disease model would significantly enhance the translational impact of this work and further validate its therapeutic potential.

    3. Reviewer #2 (Public review):

      Summary:

      The authors harness single-cell RNAseq data from zebrafish and mice to identify Oct4 as a candidate driver of neurogenesis. They then use adeno-associated virus vectors to show that while Oct4 overexpression alone converts rare adult Müller glia (MG) to bipolar cells, it synergizes with Notch pathway inhibition to cause this neurogenesis (achieved by Cre-mediated knockout of Rbpj floxed allele). Importantly, they genetically lineage-mark adult MG using a GLAST-CreER transgene and a Sun-GFP reporter, so that any non-MG cells that convert can be identified unambiguously. This is crucial because several high-profile papers made erroneous claims using short promoters in the viral delivery vector itself to mark MG, but those promoters are leaky and mark other non-MG cell types, making it impossible to definitively state whether manipulations studied were actually causing neurogenesis, or were merely the result of expression in pre-existing neurons. Once the authors establish Oct4 + RbpjKO synergy they use snRNAseq/ATACseq to identify known and novel transcription factors that could play a role in driving neurogenesis.

      Strengths:

      The system to mark MG is stringent, so the authors are studying transdifferentiation, not artifactual effects due to leaky viral promoters. The synergy between Oct4 and Notch pathway blockade is notable. The single-cell results add the potential involvement of new players such as Rfx4 in adult-MG-neurogenesis.

      Weaknesses:

      The existing version is difficult to read due to an unusually high number of text errors (e.g. references to the wrong figure panels etc.). A fuller explanation for the fraction of non-MG cells seen in control scRNAseq assays is required, particularly because the neurogenic trajectory which is enhanced in the Oct4/Rbpj-KO context is also evident in the control retina. Claims regarding the involvement of transcription factors in adult neurogenesis (such as Rfx4) need to be toned down unless they are backed up with functional data. It is possible that such factors are important, but equally, they may have no role or a redundant role, and without functional tests, it's impossible to say one way or the other.

      Overall, the authors achieved what they set out to do, and have made new insights into how neurogenesis can be stimulated in MG. Ultimately, a major long-term goal in the field is to replace lost photoreceptors as this is most relevant to many human visual disorders, and while this paper (like all others before it) does not generate rods or cones, it opens new strategies to coax MG to form a related neuronal cell type. Their approach underscores the benefits of using a gold-standard approach for lineage tracing.

    1. eLife Assessment

      This manuscript presents important information as to how adolescent alcohol exposure (AIE) alters pain behavior and relevant neurocircuits, with convincing data. The manuscript focuses on how AIE alters the basolateral amygdala, to the PFC (PV-interneurons), to the periaquaductal gray circuit, resulting in feed-forward inhibition. The manuscript is a detailed study of the role of alcohol exposure in regulating the circuit and reflexive pain, however, the role of the PV interneurons in mechanistically modulating this feed-forward circuit could be more strongly supported.

    1. eLife Assessment

      This manuscript presents important information as to how adolescent alcohol exposure (AIE) alters pain behavior and relevant neurocircuits, with convincing data. The manuscript focuses on how AIE alters the basolateral amygdala, to the PFC (PV-interneurons), to the periaquaductal gray circuit, resulting in feed-forward inhibition. The manuscript is a detailed study of the role of alcohol exposure in regulating the circuit and reflexive pain, however, the role of the PV interneurons in mechanistically modulating this feed-forward circuit could be more strongly supported.

    2. Reviewer #1 (Public review):

      Summary:

      In this manuscript by Obray et al., the authors show that adolescent ethanol exposure increases mechanical allodynia in adulthood. Additionally, the show that BLA mediated inhibition of prelimbic cortex is reduced, resulting in increased excitability in neurons that then project to vlPAG. This effect was mediated by BLA inputs onto PV interneurons. The primary finding of the manuscript is that these AIE induced changes further impact acute pain processing in the BLA-PrL-vlPAG circuit, albeit behavioral readouts after inducing acute pain were not different between AIE rats and controls. These results provide novel insights into how AIE can have long lasting effects on pain-related behaviors and neurophysiology.In this manuscript by Obray et al., the authors show that adolescent ethanol exposure increases mechanical allodynia in adulthood. Additionally, the show that BLA mediated inhibition of prelimbic cortex is reduced, resulting in increased excitability in neurons that then project to vlPAG. This effect was mediated by BLA inputs onto PV interneurons. The primary finding of the manuscript is that these AIE induced changes further impact acute pain processing in the BLA-PrL-vlPAG circuit, albeit behavioral readouts after inducing acute pain were not different between AIE rats and controls. These results provide novel insights into how AIE can have long lasting effects on pain-related behaviors and neurophysiology.

      The manuscript was very well written and the experiments were rigorously conducted. The inclusion of both behavioral and neurophysiological circuit recordings was appropriate and compelling. The authors analyzed their data extensively, and consider how many different factors may influence physiological activity and downstream behavior. The attention to SABV and appropriate controls was well thought out. The Discussion provided novel ideas for how to think about AIE and chronic pain, and proposed several interesting mechanisms. This was a very well executed set of experiments.

      Comments on revisions:

      The authors have addressed the concerns raised by the reviewers. Excellent work!

    3. Reviewer #2 (Public review):

      Summary:

      The study by Obray et al. entitled "Adolescent alcohol exposure promotes mechanical allodynia and alters synaptic function at inputs from the basolateral amygdala to the prelimbic cortex" investigated how adolescent intermittent ethanol exposure (AIE) affects the BLA -> PL circuit, with an emphasis on PAG projecting PL neurons, and how AIE changes mechanical and thermal nociception. The authors found that AIE increased mechanical, but not thermal nociception, and an injection of an inflammatory agent did not produce changes in an ethanol-dependent manner. Physiologically, a variety of AIE-specific effects were found in PL neuron firing at BLA synapses, suggestive of AIE-induced alterations in neurotransmission at BLA-PVIN synapses.

      Strengths:

      This was a comprehensive examination of the effects of AIE on this neural circuit, with an in-depth dissection of the various neuronal connections within the PL.

      Sex was included as a biological variable, yet, there were little to no sex differences in AIE's effects, suggestive of similar adaptations in males and females.

      Comments on revisions:

      The authors addressed the reviews from the first submission which has substantially strengthened the conclusions of the study, including acknowledgement of unanswered questions for future studies to address.

    4. Reviewer #3 (Public review):

      Summary:

      Obray et al. investigate the long-lasting effects of adolescent intermittent ethanol (AIE) in rats, a model of alcohol dependence, on a neural circuit within prefrontal cortex. The studies are focused on inputs from the basolateral amygdala (BLA) onto parvalbumin (PV) interneurons and pyramidal cells that project to the periaqueductal gray (PAG). The authors found that AIE increased BLA excitatory drive onto parvalbumin interneurons and increased BLA feedforward inhibition onto PAG-projecting neurons.

      Strengths:

      Fully powered cohorts of male and female rodents are used, and the design incorporates both AIE and an acute pain model. The authors used several electrophysiological techniques to assess synaptic strength and excitability from a few complimentary angles. The design and statistical analysis are sound, and the evidence supporting synaptic changes following AIE results is convincing. The authors have also revised the Discussion to assimilate the findings within prior work out of their lab and others.

      Weaknesses:

      (1) There is incomplete evidence supporting some of the conclusions drawn in this manuscript. The authors claim the changes in feedforward inhibition onto pyramidal cells are due to the changes in parvalbumin interneurons; however, the authors did not determine that PV cells mediate the feedforward BLA op-IPSCs and changes following AIE (this would require a manipulation to reduce/block PV-IN activity). This limitation in results and interpretation is important because prior work shows BLA-PFC feedforward IPSCs can be driven by somatostatin cells. Cholecystokinin cells are also abundant basket cells in PFC and have been recently shown to mediate feedforward inhibition from thalamus and ventral hippocampus, so it's also possible that CCK cells are involved in the effects observed here

      (2) The authors conclude that the changes in this circuit likely mediate long-lasting hyperalgesia, but this is not addressed experimentally. In some ways, the focused nature of the study is a benefit in this regard, as there is extensive prior literature linking this circuit with pain behaviors in alternative models (e.g., SNI), but it should be noted that these studies have not assessed hyperalgesia stemming from prior alcohol exposure. While the current studies do not include a causative behavioral manipulation, the strength of the association between BLA-PL-PAG function and hyperalgesia could be bolstered by with current data if there were relationships detected between electrophysiological properties and hyperalgesia.

      (3) It should be noted that asEPSC frequency can also reflect changes in number of functional/detectable synapses. This measurement is also fairly susceptible to differences in inter-animal differences in ChR2 expression. There are other techniques for assessing presynaptic release probability (e.g., PPR, MK-801 sensitivity) that would improve the interpretation of these studies if that is intended to be a point of emphasis.

    5. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1:

      Major Concerns/Public Review

      Comment 1: There is a mild disconnect between behavioral readout (reflexive pain) and neural circuits of interest (emotional). Considering that this circuit is likely engaged in the aversiveness of pain, it would have been interesting to see how carrageenan and/or AIE impacted non-reflexive pain measures. Perhaps this would reveal a potentiated or dysregulated phenotype that matches the neurophysiological changes reported. However, this critique does not take away from the value of the paper or its conclusions.

      We agree that including measures of non-reflexive pain would enhance future studies and potentially reveal a phenotype that is closely related to the observed changes in neurophysiology.

      Minor Concerns/Recommendations

      Comment 1: There are a few minor grammatical errors in the text, mostly in the captions. A close read should be able to identify these errors.

      We have fixed what grammatical errors we found.

      Reviewer #2:

      Major Concerns/Public Review

      No major concerns.

      Minor Concerns/Recommendations

      Comment 1: If pain sensitivity was assessed at 3 time points post carrageenan administration, why were these data averaged? Were there no differences between the time points? The data from the 3 time points should be presented, either in a figure, table, or supplementary materials.

      We averaged the pain sensitivity data across the 3 time points following carrageenan administration because we were trying to present this data in a more concise manner. Pain sensitivity did change over time following carrageenan administration. We have now included the unaveraged data in figure 2 (panels D, F, H, and J).

      Comment 2: For the optically-evoked EPSCs and IPSCs, were the peak amplitudes the max responses that could be obtained? If not, how were levels of ChR2 expression or light intensity controlled for?

      The peak amplitudes for EPSCs and IPSCs were half the maximal response that could be evoked by optical stimulation. The AMPA and NMDA currents were maximal responses as prior literature indicated some PVINs have small NMDA currents, and we wanted to ensure these currents would be detected reliably. We updated our methods section to include this information in the voltage clamp recordings section.

      Comment 3: In the example traces for the aEPSC experiment, the figure legend states that the "+" symbol indicates an asynchronous event. However, there are several "|" or "-" symbols in the figure. Perhaps this is an issue with the resolution of the figure and those are supposed to be "+"s.

      We have increased the resolution of the figures to ensure that the markings of the asynchronous events display properly. We apologize for not noticing that these symbols were not displayed correctly in the original figures included in the manuscript.

      Comment 4: For the von Frey and the Hargreaves test, were animals acclimated to the apparatus in the days leading up to the first test, or was the 5-minute pre-test the only acclimation that was done? This information needs to be provided. If the latter, there is concern that the animals did not fully acclimate to the apparatus and handling prior to testing, which should be taken into consideration in the interpretation of the behavioral analyses.

      The rats underwent handling once a day for three days prior to the first von Frey and Hargreaves tests. On the day prior to the first test, rats were acclimated to the von Frey and Hargreaves apparatuses. The acclimation period consisted of a 15-min exposure to the von Frey apparatus and a 30-min exposure to the Hargreaves apparatus for each animal. This information has been added to the revised methods section under the assessment of mechanical and thermal sensitivity heading.

      Reviewer #3:

      Major Concerns/Public Review

      Comment 1: There is incomplete evidence supporting some of the conclusions drawn in this manuscript. The authors claim that the changes in feedforward inhibition onto pyramidal cells are due to the changes in parvalbumin interneurons, but evidence is not provided to support that idea. PV cells do not spontaneously fire action potentials spontaneously in slices (nor do they receive high levels of BLA activity while at rest in slices). It is possible that spontaneous GABA release from PV cells is increased after AIE but the authors did not report sIPSC frequency. Second, the authors did not determine that PV cells mediate the feedforward BLA op-IPSCs and changes following AIE (this would require manipulation to reduce/block PV-IN activity). This limitation in results and interpretation is important because prior work shows BLA-PFC feedforward IPSCs can be driven by somatostatin cells. Cholecystokinin cells are also abundant basket cells in PFC and have been recently shown to mediate feedforward inhibition from the thalamus and ventral hippocampus, so it's also possible that CCK cells are involved in the effects observed here.

      The hypothesis that adolescent alcohol exposure could change spontaneous GABA release from PVINs is an interesting one that merits future exploration. Unfortunately, as the focus of this manuscript was on circuit-specific alterations in synaptic function, this experiment is somewhat outside the scope of the paper as sIPSCs and mIPSCs are not circuit specific measures of GABA activity and would not reflect spontaneous release from only GABA interneurons receiving input from the BLA. Despite this, a future study investigating spontaneous GABA release from PVINs in the PrL would be a valuable complement to the present study.

      While we did not directly manipulate PVINs to demonstrate that decreased oIPSC amplitude at PrL<sup>PAG</sup> neurons following AIE is due solely to changes in PVINs, it is notable that both the intrinsic excitability of PVINs and the BLA-driven E/I balance at PVINs were reduced following AIE. These changes would be consistent with decreased PVIN output onto PrL<sup>PAG</sup> neurons. However, we agree that this does not preclude the possibility that changes in SST or CCK interneurons contribute to the observed decrease in BLA-driven inhibition at PrL<sup>PAG</sup> neurons following AIE. As such, we have altered the wording in the discussion to indicate that reduced BLA-driven feedforward inhibition of PrL<sup>PAG</sup> neurons may be related, at least in part, to the observed changes in PVINs.

      Comment 2: The authors conclude that the changes in this circuit likely mediate long-lasting hyperalgesia, but this is not addressed experimentally. In some ways, the focused nature of the study is a benefit in this regard, as there is extensive prior literature linking this circuit with pain behaviors in alternative models (e.g., SNI), but it should be noted that these studies have not assessed hyperalgesia stemming from prior alcohol exposure. While the current studies do not include a causative behavioral manipulation, the strength of the association between BLA-PL-PAG function and hyperalgesia could be bolstered by current data if there were relationships detected between electrophysiological properties and hyperalgesia. Have the authors assessed this? In addition, this study is limited by not addressing the specificity of synaptic adaptations to the BLA-PL-PAG circuit. For instance, PL neurons send reciprocal projections to BLA and send direct projections to the locus coeruleus (which the authors note is an important downstream node of the PAG for regulating pain).

      We have not assessed correlations between the electrophysiological properties and hyperalgesia. We feel that future studies using DREADDs to perform cell-type and circuit-specific manipulations can better address the involvement of this circuitry in long-lasting hyperalgesia following AIE. With respect to the circuit specificity of the observed changes, we have previously evaluated the effects of AIE on pyramidal neurons projecting from the PrL to the BLA (PrL<sup>BLA</sup>). We found that following AIE exposure there was no change in the intrinsic excitability of these neurons. In addition, the amplitude and frequency of sEPSCs and sIPSCs onto PrL<sup>BLA</sup> neurons was unchanged. While these results did not assess whether the BLA-PrL-BLA circuit undergoes synaptic adaptations similar to those observed in the BLA-PrL-vlPAG circuit, it is notable that the intrinsic excitability of PrL<sup>BLA</sup> neurons was unchanged following AIE exposure. This indicates that the effects of AIE on the intrinsic excitability of pyramidal neurons in the PrL may be circuit specific. We agree that it would be interesting to study the effect of AIE on PrL neurons that project to the locus coeruleus, however due to the well-defined role of the BLA-PrL-vlPAG circuit in pain we chose to evaluate this circuit first.

      Comment 3: I have some concerns about methodology. First, 5-ms is a long light pulse for optogenetics and might induce action-potential independent release. Does TTX alone block op-EPSCs under these conditions? Second, PV cells express a high degree of calcium-permeable AMPA receptors, which display inward rectification at positive holding potentials due to blockade from intracellular polyamines. Typically, this is controlled/promoted by including spermine in the internal solution, but I do not believe the authors did that. Nonetheless, the relatively low A/N ratios for this cell type suggest that CP-AMPA receptors were not sampled with the +40/+40 design of this experiment, raising concerns that the majority of AMPA receptors in these cells were not sampled during this experiment. Finally, it should be noted that asEPSC frequency can also reflect changes in a number of functional/detectable synapses. This measurement is also fairly susceptible to differences in inter-animal differences in ChR2 expression. There are other techniques for assessing presynaptic release probability (e.g., PPR, MK-801 sensitivity) that would improve the interpretation of these studies if that is intended to be a point of emphasis.

      When we included TTX but not 4-AP we did not observe any optically evoked responses, so we don’t believe that the 5-ms pulse induced action-potential independent release in these experiments. With respect to the second point, we did not include spermine in the internal solution for the AMPA/NMDA recordings in PVINs, and it is possible that endogenous polyamines interfered with recording CP-AMPA receptors in the +40/+40 design. To address this concern, we recalculated the AMPA/NMDA ratio for PVINs using data from an optically evoked AMPA current that was collected while holding the cell at -70 mV. This data was collected at the end of the +40/+40 recording protocol as we were interested in assessing whether there would be any difference in the ratio of the +40/-70 AMPA current across treatment conditions. As there were no observed difference in the +40/-70 AMPA current ratio across treatment groups, we had originally used the +40 AMPA current for calculating the AMPA/NMDA ratio for PVINs to make the methods for calculating this ratio uniform for both PVINs and PrL<sup>PAG</sup> neurons. The methods, results, and Fig. 10 have been updated to reflect the recalculated AMPA/NMDA ratio for PVINs. Notably, only the significance of the AIE x carrageenan interaction was altered by the change in the way the AMPA/NMDA ratio was calculated. Originally, this interaction displayed a trend toward significance (p = 0.0501), however when the recalculated AMPA/NMDA ratio was analyzed this interaction term became significant (p = 0.0131). We have also added the +40/-70 AMPA ratio to figure 10 as it might be of interest.

      Finally, the point regarding aEPSC frequency reflecting not only release probability but also the number of functional/detectable synapses is an important consideration. For this manuscript, we intentionally selected aEPSC frequency for this reason. As the BLA to PrL projection continues to mature during adolescence, the number of BLA contacts onto GABA neurons in the PrL increases. Thus, we thought that it was possible that AIE would alter the number of detectable BLA inputs onto PVINs. We acknowledge that as this measure is sensitive to differences in ChR2 expression between animals/slices it can be difficult to interpret. We also agree that in the future it would be beneficial to include either PPR or MK-801 sensitivity to improve interpretability.

      Comment 4: In a few places in the manuscript, results following voluntary drinking experiments (especially Salling et al. and Sicher et al.) are discussed without clear distinction from prior work in vapor models of dependence.

      We have altered the manuscript to specifically note where voluntary drinking was used rather than vapor models.

      Comment 5: Discussion (lines 416-420). The authors describe some differing results with the literature and mention that the maximum current injection might be a factor. To me, this does not seem like the most important factor and potentially undercuts the relevance of the findings. Are the cells undergoing a depolarization block? Did the authors observe any changes in the rheobase or AP threshold? On the other hand, a more likely difference between this and previous work is that the proportion of PAG-projecting cells is relatively low, so previous work in L5 likely sampled many types of pyramidal cells that project to other areas. This is a key example where additional studies by the current group assessing a distinct or parallel set of pyramidal cells would aid in the interpretation of these results and help to place them within the existing literature. Along these lines, PAG-projecting neurons are Type A cells with significant hyperpolarization sag. Previous studies showed that adolescent binge drinking stunts the development of HCN channel function and ensuing hyperpolarization sag. Have the authors observed this in PAG-projecting cells? Another interesting membrane property worth exploring with the existing data set is the afterhyperpolarization / SK channel function.

      In discussing the maximum current injection as a factor in differing results on intrinsic excitability, we were principally considering how the additional data points increase the power of the analysis and thus the likelihood of detecting an effect. In focusing on this, however, we ignored other relevant and interesting factors that we should also have discussed. Additional analyses examining HCN and SK channel function have now been added to the manuscript and incorporated into the results section under the heading Adolescent Intermittent Ethanol Exposure and Carrageenan Enhanced the Intrinsic Excitability of Prelimbic Neurons Projecting to the Ventrolateral Periaqueductal Gray. We have also modified the third paragraph in the discussion to add additional context. Additional information on the biophysical properties of the neurons has been added to Figure 4.

      Minor Concerns/Recommendations

      Comment 1: Subheadings are vague. "Analysis of..." Should be rephrased to use active voice to describe key findings.

      The subheadings have been rephrased to describe key findings.

      Comment 2: Consider altering or consolidating the figure layout for clarity. For instance, it would be helpful for aEPSCs to be near the AMPA and NMDA experiments. The feedforward IPSCs could also be with the PV-IN recordings. This would be helpful in developing a cohesive picture of key findings. To that end, a working model or graphical abstract would be helpful.

      It doesn’t appear that this journal allows graphical abstracts, but we have added a model that summarizes the principal findings in the discussion.

      Comment 3: There are a lot of statistics punctuating the text in the Results. It can be hard to parse at times.

      We considered moving the statistics to tables, but this became unwieldy.

      Comment 4: The Discussion is quite long (10 paragraphs). Suggest consolidating to 3-4 most salient points.

      We appreciate this comment and have made some edits to the discussion, albeit without consolidating it to only 3-4 points.

    1. eLife Assessment

      This study provides a novel and critically important insight into the long-term use of DREADDs to modulate neuronal activity in nonhuman primates. The methods are compelling, demonstrating the peak dynamics and the subsequent stability of chemogenetic effects for 1.5 years, informing experimental designs and interpretation of highly impactful chemogenetic studies in macaques. The protocols, data, and outcomes can serve as guidelines for future experiments. Therefore, the findings will be of significant interest to the field of chemogenetics and may also be of broader interest to researchers and clinicians who seek to utilize viral vectors and/or related genetic technologies.

    2. Reviewer #1 (Public review):

      Summary:

      Inhibitory hM4Di and excitatory hM3Dq DREADDs are currently the most commonly utilized chemogenetic tools in the field of nonhuman primate research, but there is a lack of available information regarding the temporal aspects of virally-mediated DREADD expression and function. Nagai et al. investigated the longitudinal expression and efficacy of DREADDs to modulate neuronal activity in the macaque model. The authors demonstrate that both hM4Di and hM3Dq DREADDs reach peak expression levels after approximately 60 days and are stably expressed for a period of at least 1.5 years in the macaque brain. During this period, DREADDs effectively modulated neuronal activity, as evidenced by a variety of measures, including behavioural testing, functional imaging, and/or electrophysiological recording. Notably, some of the data suggest that DREADD expression may decline after two years. This is a novel finding and has important implications for the utilization of this technology for long-term studies, as well as its potential therapeutic applications. Lastly, the authors highlight that peak DREADD expression may be significantly influenced by the choice of viral titer and the expressed protein tag, emphasizing the importance of careful design and selection of viral constructs for neuroscientific research. This study represents a critical step in the field of chemogenetics, setting the scene for future development and optimization of this technology.

      Strengths:

      The longitudinal approach of this study provides important preliminary insights into the long-term utility of chemogenetics, which has not yet been thoroughly explored.

      The data presented are novel and inclusive, relying on well-established in vivo imaging methods, as well as behavioral and immunohistochemical techniques. The conclusions made by the authors are generally supported by a combination of these techniques. In particular, the utilization of in vivo imaging as a non-invasive method is translationally relevant and likely to make an impact in the field of chemogenetics, such that other researchers may adopt this method of longitudinal assessment in their own experiments. Rigorous standards have been applied to the datasets, and the appropriate controls have been included where possible.

      The number of macaque subjects (20) from which data was available is also notable. Behavioral testing was performed in 11 subjects, FDG-PET in 5, electrophysiology in 1, and [11C]DCZ-PET in 15. This is an impressive accumulation of work that will surely be appreciated by the growing community of researchers using chemogenetics in nonhuman primates.

      The implication that chemogenetic effects can be maintained for up to 1.5-2 years, followed by a gradual decline beyond this period, is an important development in knowledge. The limited duration of DREADD expression may present an obstacle in the translation of chemogenetic technology as a potential therapeutic tool, and it will be of interest for researchers to explore whether this limitation can be overcome. This study therefore represents a key starting point upon which future research can build.

      Weaknesses:

      Overall, the conclusions of the paper are mostly supported by the data but may be overstated in some cases, and some details are also missing or not easily recognizable within the figures. The provision of additional information and analyses would be valuable to the reader and may even benefit the authors' interpretation of the data.

      The conclusion that DREADD expression gradually decreases after 1.5-2 years is only based on a select few of the subjects assessed; in Figure 2, it appears that only 3 hM4Di cases and 2 hM3Dq cases are assessed after the 2-year timepoint. The observed decline appears consistent within the hM4Di cases, but not for the hM3Dq cases (see Figure 2C: the AAV2.1-hSyn-hM3Dq-IRES-AcGFP line is increasing after 2 years.)

      Given that individual differences may affect expression levels, it would be helpful to see additional labels on the graphs (or in the legends) indicating which subject and which region are being represented for each line and/or data point in Figure 1C, 2B, 2C, 5A, and 5B. Alternatively, for Figures 5A and B, an accompanying table listing this information would be sufficient.

      While the authors comment on several factors that may influence peak expression levels, including serotype, promoter, titer, tag, and DREADD type, they do not comment on the volume of injection. The range in volume used per region in this study is between 2 and 54 microliters, with larger volumes typically (but not always) being used for cortical regions like the OFC and dlPFC, and smaller volumes for subcortical regions like the amygdala and putamen. This may weaken the claim that there is no significant relationship between peak expression level and brain region, as volume may be considered a confounding variable. Additionally, because of the possibility that larger volumes of viral vectors may be more likely to induce an immune response, which the authors suggest as a potential influence on transgene expression, not including volume as a factor of interest seems to be an oversight.

      The authors conclude that vectors encoding co-expressed protein tags (such as HA) led to reduced peak expression levels, relative to vectors with an IRES-GFP sequence or with no such element at all. While interesting, this finding does not necessarily seem relevant for the efficacy of long-term expression and function, given that the authors show in Figures 1 and 2 that peak expression (as indicated by a change in binding potential relative to non-displaced radioligand, or ΔBPND) appears to taper off in all or most of the constructs assessed. The authors should take care to point out that the decline in peak expression should not be confused with the decline in longitudinal expression, as this is not clear in the discussion; i.e. the subheading, "Factors influencing DREADD expression," might be better written as, "Factors influencing peak DREADD expression," and subsequent wording in this section should specify that these particular data concern peak expression only.

    3. Reviewer #2 (Public review):

      Summary

      This paper reports histological, PET imaging, functional, and behavioural data evaluating the longevity of AAV2 infection in multiple brain areas of macaques in the context of DREADD experiments. The central aim is to provide unprecedented information about how long the expression of HM4di or HM3dq receptors is expressed and efficient in modulating brain functions after vector injections. The data show peak expression after 40 to 60 days of vector injection, and stable expressions for up to 1.5 years for hM4di, and that hM3dq remained mostly at 75% of peak after a year, declining to 50% after 2 years. DREADDs effectively modulated neuronal activity and behaviour for approximately two years, evaluated with behavioral testings, neural recordings, or FDG-PET. A statistical evaluation revealed that vector titers, DREADD type, and tags contribute to the measured peak level of DREADD expression.

      The article presents a thorough discussion of the limitations and specificities of chemogenetic approaches in monkeys.

      Strength

      These are unique data, in non-human primates (NHP), an animal model that not only features physiological and immunological characteristics similar to humans but also contribute to neurobiological functional studies on a long timescale with experiments spanning months or years. This evaluation of the long-term efficacy of DREADDs will be very important for all laboratories using this approach in NHP but also for future use of such approach in experimental therapies. The longevity estimates are based on multiple approaches including behavioural and neurophysiological ones, thus providing information on the functional efficacy of DREADD expression.

      Performing such evaluation requires specific tools like PET imaging that very few monkey labs have access to in the world. This study was done by the laboratory that has developed the radiotracer c11-DCZ used here, a radiotracer binding selectively to DREADDs and providing, using PET, quantitative in vivo measures of DREADD expression. This study and its data should thus be a reference in the field, providing estimates to plan future chemogenetic experiments.

      Publishing databases of experimental outcomes in NHP DREADD experiments is crucial for the community because such experiments are rare, expensive, and long. It contributes to refining experiments and reducing the number of animals overall used in the domain.

      Weaknesses

      This study is a meta-analysis of several experiments performed in one lab. The good side is that it combined a large amount of data that might not have been published individually; the downside is that all things were not planned and equated, creating a lot of unexplained variances in the data. This was yet judiciously used by the authors, but one might think that planned and organized multicentric experiments would provide more information and help test more parameters, including some related to inter-individual variability, and particular genetic constructs.

    4. Reviewer #3 (Public review):

      Summary

      This manuscript, from the developers of the novel DREADD-selective agonist DCZ (Nagai et al., 2020), utilizes a unique dataset where multiple PET scans in a large number of monkeys, including baseline scans before AAV injection, 30-120 days post-injection, and then periodically over the course of the prolonged experiments, were performed to access short- and long-term dynamics of DREADD expression in vivo, and to associate DREADD expression with the efficacy of manipulating the neuronal activity or behavior. The goal was to provide critical insights into the practicality and design of multi-year studies using chemogenetics and to elucidate factors affecting expression stability.

      Strengths are systematic quantitative assessment of the effects of both excitatory and inhibitory DREADDs, quantification of both the short-term and longer-term dynamics, a wide range of functional assessment approaches (behavior, electrophysiology, imaging), and assessment of factors affecting DREADD expression levels, such as serotype, promoter, titer (concentration), tag, and DREADD type.

      Minor weaknesses are related to a few instances of suboptimal phrasing, and some room for improvement in time course visualization and quantification. These would be easily addressed in a revision.

      These findings will undoubtedly have a very significant impact on the rapidly growing but still highly challenging field of primate chemogenetic manipulations. As such, the work represents an invaluable resource for the community.

    5. Author response:

      Public Reviews:

      Reviewer #1 (Public review):

      Overall, the conclusions of the paper are mostly supported by the data but may be overstated in some cases, and some details are also missing or not easily recognizable within the figures. The provision of additional information and analyses would be valuable to the reader and may even benefit the authors' interpretation of the data.

      We thank the reviewer for the thoughtful and constructive feedback. We are pleased that the reviewer found the overall conclusions of our paper to be well supported by the data, and we appreciate the suggestions for improving figure clarity and interpretive accuracy. Below we address each point raised:

      The conclusion that DREADD expression gradually decreases after 1.5-2 years is only based on a select few of the subjects assessed; in Figure 2, it appears that only 3 hM4Di cases and 2 hM3Dq cases are assessed after the 2-year timepoint. The observed decline appears consistent within the hM4Di cases, but not for the hM3Dq cases (see Figure 2C: the AAV2.1-hSyn-hM3Dq-IRES-AcGFP line is increasing after 2 years.)

      We agree that our interpretation should be stated more cautiously, given the limited number of cases assessed beyond the two-year timepoint. In the revised manuscript, we will clarify in both the Results and Discussion that the observed decline is based on a subset of animals. We will also state that while a consistent decline was observed in hM4Di-expressing monkeys, the trajectory for hM3Dq expression was more variable—with at least one case showing increased in signal beyond two years.

      Given that individual differences may affect expression levels, it would be helpful to see additional labels on the graphs (or in the legends) indicating which subject and which region are being represented for each line and/or data point in Figure 1C, 2B, 2C, 5A, and 5B. Alternatively, for Figures 5A and B, an accompanying table listing this information would be sufficient.

      We thank the reviewer for these helpful suggestions. In response, we will revise the relevant figures as noted in the “Recommendations for the authors”, including simplifying visual encodings and improving labeling. We will also provide a supplementary table listing the animal ID and brain regions for each data point shown in the graphs.

      While the authors comment on several factors that may influence peak expression levels, including serotype, promoter, titer, tag, and DREADD type, they do not comment on the volume of injection. The range in volume used per region in this study is between 2 and 54 microliters, with larger volumes typically (but not always) being used for cortical regions like the OFC and dlPFC, and smaller volumes for subcortical regions like the amygdala and putamen. This may weaken the claim that there is no significant relationship between peak expression level and brain region, as volume may be considered a confounding variable. Additionally, because of the possibility that larger volumes of viral vectors may be more likely to induce an immune response, which the authors suggest as a potential influence on transgene expression, not including volume as a factor of interest seems to be an oversight.

      We thank the reviewer for raising this important issue. We agree that injection volume is a potentially confounding variable. In response, we will conduct an exploratory analysis including volume as an additional factor. We will also expand the Discussion to highlight the need for future systematic evaluation of injection volume, especially in relation to immune responses or transduction efficiency in different brain regions.

      The authors conclude that vectors encoding co-expressed protein tags (such as HA) led to reduced peak expression levels, relative to vectors with an IRES-GFP sequence or with no such element at all. While interesting, this finding does not necessarily seem relevant for the efficacy of long-term expression and function, given that the authors show in Figures 1 and 2 that peak expression (as indicated by a change in binding potential relative to non-displaced radioligand, or ΔBPND) appears to taper off in all or most of the constructs assessed. The authors should take care to point out that the decline in peak expression should not be confused with the decline in longitudinal expression, as this is not clear in the discussion; i.e. the subheading, "Factors influencing DREADD expression," might be better written as, "Factors influencing peak DREADD expression," and subsequent wording in this section should specify that these particular data concern peak expression only.

      We appreciate this important clarification. In response, we will revise the title to “Factors influencing peak DREADD expression levels”, and we will specify that our analysis focused on peak ΔBP<sub>ND</sub> values around 60 days post-injection. We will also explicitly distinguish these findings from the later-stage changes in expression seen in the longitudinal PET data in both the Results and Discussion sections.

      Reviewer #2 (Public review):

      Weaknesses

      This study is a meta-analysis of several experiments performed in one lab. The good side is that it combined a large amount of data that might not have been published individually; the downside is that all things were not planned and equated, creating a lot of unexplained variances in the data. This was yet judiciously used by the authors, but one might think that planned and organized multicentric experiments would provide more information and help test more parameters, including some related to inter-individual variability, and particular genetic constructs.

      We thank the reviewer for bringing this important point to our attention. We fully agree that the retrospective nature of our dataset, compiled from multiple studies conducted within a single laboratory, introduces variability due to differences in constructs, injection sites, and timelines. While this reflects the real-world constraints of long-term NHP research, we acknowledge the need for more standardized approaches. We will add a statement in the revised Discussion emphasizing that future multicenter and harmonized studies would be valuable for systematically examining specific parameters and inter-individual variability.

      Reviewer #3 (Public review):

      Minor weaknesses are related to a few instances of suboptimal phrasing, and some room for improvement in time course visualization and quantification. These would be easily addressed in a revision.

      These findings will undoubtedly have a very significant impact on the rapidly growing but still highly challenging field of primate chemogenetic manipulations. As such, the work represents an invaluable resource for the community.

      We thank the reviewer for the positive assessment of our manuscript and for the constructive suggestions noted in the “Recommendations for the authors”. In response, we will carefully review and revise the manuscript to improve visualization and quantification.

    1. eLife Assessment

      This important study provides compelling insights into the differential impact of intrinsic and synaptic conductances on circuit robustness using computational models of the pyloric network from the crustacean stomatogastric ganglion. The results demonstrate that model networks are more sensitive to perturbations in intrinsic conductances than in synaptic conductances, highlighting the critical role of intrinsic plasticity in stabilizing neuronal networks. These findings underscore the importance of intrinsic plasticity, a crucial yet often overlooked factor in neuronal dynamics. The generality of these conclusions should be tested across diverse networks and functions.

    2. Reviewer #1 (Public review):

      The paper by Fournier et al. investigates the sensitivity of neural circuits to changes in intrinsic and synaptic conductances. The authors use models of the stomatogastric ganglion (STG) to compare how perturbations to intrinsic and synaptic parameters impact network robustness. Their main finding is that changes to intrinsic conductances tend to have a larger impact on network function than changes to synaptic conductances, suggesting that intrinsic parameters are more critical for maintaining circuit function.

      The paper is well-written, and the results are compelling. The authors addressed most of the minor comments I had and improved the manuscript.

      However, it remains unclear how general the results are and what the underlying mechanism is. Regarding generality, the authors changed the title and added a sentence in the discussion. At this point, they do not claim generality beyond the specific function they explore in the STG circuit. While this is acceptable, I still believe the paper would be much more insightful if it provided a more general statement and investigated the mechanism behind why, in their hands, synaptic parameters appear more resilient to changes than intrinsic parameters.

    3. Author response:

      The following is the authors’ response to the original reviews

      Public Reviews:

      Reviewer #1 (Public review):

      The paper by Fournier et al. investigates the sensitivity of neural circuits to changes in intrinsic and synaptic conductances. The authors use models of the stomatogastric ganglion (STG) to compare how perturbations to intrinsic and synaptic parameters impact network robustness. Their main finding is that changes to intrinsic conductances tend to have a larger impact on network function than changes to synaptic conductances, suggesting that intrinsic parameters are more critical for maintaining circuit function.

      The paper is well-written and the results are compelling, but I have several concerns that need to be addressed to strengthen the manuscript. Specifically, I have two main concerns:

      (1) It is not clear from the paper what the mechanism is that leads to the importance of intrinsic parameters over synaptic parameters.

      (2) It is not clear how general the result is, both within the framework of the STG network and its function, and across other functions and networks. This is crucial, as the title of the paper appears very general.

      I believe these two elements are missing in the current manuscript, and addressing them would significantly strengthen the conclusions. Without a clear understanding of the mechanism, it is difficult to determine whether the results are merely anecdotal or if they depend on specific details such as how the network is trained, the particular function being studied, or the circuit itself. Additionally, understanding how general the findings are is vital, especially since the authors claim in the title that "Circuit function is more robust to changes in synaptic than intrinsic conductances," which suggests a broad applicability.

      I do not wish to discourage the authors from their interesting result, but the more we understand the mechanism and the generality of the findings, the more insightful the result will be for the neuroscience community.

      Major comments

      (1) Mechanism

      While the authors did a nice job of describing their results, they did not provide any mechanism for why synaptic parameters are more resilient to changes than intrinsic parameters. For example, from Figure 5, it seems that there is mainly a shift in the sensitivity curves. What is the source of this shift? Can something be changed in the network, the training, or the function to control it? This is just one possible way to investigate the mechanism, which is lacking in the paper.

      (2) Generality of the results within the framework of the STG circuit

      (a) The authors did show that their results extend to multiple networks with different parameters (the 100 networks). However, I am still concerned about the generality of the results with respect to the way the models were trained. Could it be that something in the training procedure makes the synaptic parameters more robust than intrinsic parameters? For example, the fact that duty cycle error is weighted as it is in the cost function (large beta) could potentially affect the parameters that are more important for yielding low error on the duty cycle.

      (b) Related to (a), I can think of a training scheme that could potentially improve the resilience of the network to perturbations in the intrinsic parameters rather than the synaptic parameters. For example, in machine learning, methods like dropout can be used to make the network find solutions that are robust to changes in parameters. Thus, in principle, the results could change if the training procedure for fitting the models were different, or by using a different optimization algorithm. It would be helpful to at least mention this limitation in the discussion.

      (3) Generality of the function

      The authors test their hypothesis based on the specific function of the STG. It would be valuable to see if their results generalize to other functions as well. For example, the authors could generate non-oscillatory activity in the STG circuit, or choose a different, artificial function, maybe with different duty cycles or network cycles. It could be that this is beyond the scope of this paper, but it would be very interesting to characterize which functions are more resilient to changes in synapses, rather than intrinsic parameters. In other words, the authors might consider testing their hypothesis on at least another 'function' and also discussing the generality of their results to other functions in the discussion.

      (4) Generality of the circuit

      The authors have studied the STG for many years and are pioneers in their approach, demonstrating that there is redundancy even in this simple circuit. This approach is insightful, but it is important to show that similar conclusions also hold for more general network architectures, and if not, why. In other words, it is not clear if their claim generalizes to other network architectures, particularly larger networks. For example, one might expect that the number of parameters (synaptic vs intrinsic) might play a role in how resilient the function is with respect to changes in the two sets of parameters. In larger models, the number of synaptic parameters grows as the square of the number of neurons, while the number of intrinsic parameters increases only linearly with the number of neurons. Could that affect the authors' conclusions when we examine larger models?

      In addition, how do the authors' conclusions depend on the "complexity" of the non-linear equations governing the intrinsic parameters? Would the same conclusions hold if the intrinsic parameters only consisted of fewer intrinsic parameters or simplified ion channels? All of these are interesting questions that the authors should at least address in the discussion.

      We thank Reviewer #1 for their valuable input. We agree with the reviewer that generality of the results may have been overstated. To address this we changed the title of the manuscript to make it more specific to rhythmic circuits and we included a sentence to this effect in the discussion. 

      (1) We were more interested in knowing which set of conductances is more robust in a population of models, rather than a mechanism. If such a mechanism exists it will be the subject of a different study.

      (2) (a) It is impossible to explore the whole parameter space of these models. Our method to find circuits will leave subsets of circuits out of the study. Our sole goal in constructing the model database was that the activities were similar but the conductances were different.  (b) Of course one could devise a cost function targeting circuits that are more or less robust to changes in one parameter. Whether those exist is a different matter. This is not what we intended to do.

      (3) For this we would need a different circuit that produces non-oscillatory activity. A normal pyloric rhythm circuit always produces oscillatory activity unless it is “crashed"either by temperature or perturbations, but even in this case because we don’t have a proper “control” activity (circuits crash in different ways) we would not be able to utilize the same approach.

      We think it is a valuable idea to perform a similar study in another small circuit with nonoscillatory (or rhythmic) activities. 

      (4) We did not explore the issue of how our results generalize to larger networks as it would be pure speculation. It could be potentially interesting to do a similar sensitivity analysis with a large network trained to perform a simple task. Our understanding is that many large trained networks are extremely sensitive to perturbations in synaptic weights, at the same time that the intrinsic properties of neurons in ANN are typically oversimplified and identical across units. 

      Reviewer #2 (Public review):

      Summary:

      This manuscript presents an important exploration of how intrinsic and synaptic conductances affect the robustness of neural circuits. This is a well-deserved question, and overall, the manuscript is written well and has a logical progression.

      The focus on intrinsic plasticity as a potentially overlooked factor in network dynamics is valuable. However, while the stomatogastric ganglion (STG) serves as a well-characterized and valuable model for studying network dynamics, its simplified structure and specific dynamics limit the generalizability of these findings to more complex systems, such as mammalian cortical microcircuits.

      Strengths:

      Clean and simple model. Simulations are carefully carried out and parameter space is searched exhaustively.

      Weaknesses:

      (1) Scope and Generalizability:

      The study's emphasis on intrinsic conductance is timely, but with its minimalistic and unique dynamics, the STG model poses challenges when attempting to generalize findings to other neural systems. This raises questions regarding the applicability of the results to more complex circuits, especially those found in mammalian brains and those where the dynamics are not necessarily oscillating. This is even more so (as the authors mention) because synaptic conductances in this study are inhibitory, and changes to their synaptic conductances are limited (as the driving force for the current is relatively low).

      (2) Challenges in Comparison:

      A significant challenge in the study is the comparison method used to evaluate the robustness of intrinsic versus synaptic perturbations. Perturbations to intrinsic conductances often drastically affect individual neurons' dynamics, as seen in Figure 1, where such changes result in single spikes or even the absence of spikes instead of the expected bursting behavior. This affects the input to downstream neurons, leading to circuit breakdowns. For a fair comparison, it would be essential to constrain the intrinsic perturbations so that each neuron remains within a particular functional range (e.g., maintaining a set number of spikes). This could be done by setting minimal behavioral criteria for neurons and testing how different perturbation limits impact circuit function.

      (3) Comparative Metrics for Perturbation:

      Another notable issue lies in the evaluation metrics for intrinsic and synaptic perturbations. Synaptic perturbations are straightforward to quantify in terms of conductance, but intrinsic perturbations involve more complexity, as changes in maximal conductance result in variable, nonlinear effects depending on the gating states of ion channels. Furthermore, synaptic perturbations focus on individual conductances, while intrinsic perturbations involve multiple conductance changes simultaneously. To improve fairness in comparison, the authors could, for example, adjust the x-axis to reflect actual changes in conductance or scale the data post hoc based on the real impact of each perturbation on conductance. For example, in Figure 6, the scale of the panels of the intrinsic (e.g., g_na-bar) is x500 larger than the synaptic conductance (a row below), but the maximal conductance for sodium hits maybe for a brief moment during every spike and than most of the time it is close to null. Moreover, changing the sodium conductance over the range of 0-250 for such a nonlinear current is, in many ways, unthinkable, did you ever measure two neurons with such a difference in the sodium conductance? So, how can we tell that the ranges of the perturbations make a meaningful comparison?

      We thank Reviewer #2 for their comments. We agree with both reviewers about scope and generalizability. We changed the title of the manuscript and included a sentence in the discussion to address this. 

      Recommendations for the authors:

      Reviewer #1 (Recommendations for the authors):

      (1) Line 63: Tau_b is tau in Fig 1B? What is the 'network period' tau_n? Both are defined in the methods, but it would be good to clarify here and also in the figure.

      This was fixed. Tau_b is the  bursting period and we indicated it in the figure. Network period means the period of the network activity. This was rewritten.  

      (2) Line 74: "maximal conductances g_i." What is i? I can imagine what you meant, but it would be good to clarify the notation.

      There are multiple different currents. Letter ‘i' is an index over the different types. It now reads as follows,

      "The activity of the network depends on the values of the maximal conductances g ̄ i, where i is an index corresponding to the different current types (Na,CaS,CaT,Kd,KCa,A,H,Leak IMI)"

      (3) Line 78: "conductances are changed by a random amount." How much is the "random amount"? In percentages? 

      We fixed this sentence. This is how it reads now, 

      "The blue trace in Figure 1C corresponds to the activity of the same model when each  of the intrinsic conductances is changed by a random amount within a range between 0  (completely removing the conductance) and twice its starting value, 2×gi, or equivalently, an increment of 100%."

      Similarly, in Line 87: "by a similar percent." Can you provide Figures 1E-F in percentages? Are the percentages the same?

      The phrase "by a similar percent.” Is misleading and unimportant. Thank you, we removed it. 

      (4) Line 113: Why did you add I_MI? Is it important for the results or for the conclusions?

      I_MI was added because the current is known to be there and it is not more or less important for the results or conclusions than any other current. 

      (5) Line 117: "We used a genetic algorithm to generate a database." Confusing. I guess you meant that you used genetic algorithms to optimize the cost function.

      Thank you for this comment. We fixed this sentence, see below. 

      “We used a genetic algorithm to optimize the cost function, and in this way generated a database of N = 100 models with different values of maximal conductances (Holland 88)."

      (6) Line 136: "The models in the database were constrained to produce solutions whose features were similar to the experimental measurements." Why are there differences in the features? Is this an optimization issue? I thought you wanted to claim that there are degenerate solutions, that is, solutions where the parameters are different, but the output is identical. Please clarify.

      The concept of degenerate solutions does not imply that the solutions are mathematically identical. In biology this means that they provide very similar functions, but do so with different underlying parameters (in this case, maximal conductances). The activity of the pyloric network is slightly different across animals, and it also changes over time within the same individual. Variation across models reflects individual variation in the biological circuit, and it is strength of our modeling approach. The function of the circuits are equally good because they produce biologically realistic patterns, although the details of the activity patterns show differences. 

      (7) Line 139: "distributed (p > 0.05)." What test did you use? N? Similarly, at Lines 218, 241, 239, etc. Please be more rigorous when reporting statistical tests.

      Thank you. We now specify the test we utilized every time we report a p value. 

      (8) Line 143: "In this case, it is not possible to identify clusters, suggesting that there are no underlying relationships between the features in the model database." The 2D plot is misleading, as the features are in 11 dimensions. Claims should be about the 11D space, not projections onto 2D. In fact, I don't think you can rule out correlations between the features based on the 2D plots. For example, shouldn't there be correlations between the on and off phases and the burst durations?

      Thank you. These sentences were confusing and were removed. We added the following sentence to the end of that paragraph.

      "Because the feature vectors are similar, their t-SNE projections do not form groups or clusters."

      (9) Related to this, I don't understand this sentence: "Even though the conductances are broadly distributed over many-fold ranges, the output of the circuits results in tight yet uncorrelated distributions.”

      This sentence is confusing and was removed. 

      (10) Line 158: Repetition of Line 152: Figure 3 shows the currentscapes of each cell in two model networks.

      We removed the second instance of the repeated sentences. 

      (11) Line 160: "yet the activity of the networks is similar." Well, they are similar, but not identical. I can also say that the current scapes are 'similar'. This should be better quantified and not left as a qualitative description.

      While this is an interesting point it will not change the results and conclusions of the present study. The network models are different since the values of their maximal conductances are distributed over wide ranges.  

      (12) Line 218: midpoint parameter? Is that b - the sharpness? Please be consistent. Regarding the mechanism (see above) - any ideas what leads to this shift in the sensitivity curves between the two types of parameters?

      Yes, we made a mistake. ‘b’ is the midpoint parameter. This was fixed in the text, thank you.

      (13) Figure 6 illustrates why synaptic parameters are more robust, but it is not quantified. Why not provide a quantitative measure for this claim? For example, calculate the colored area within the white square for each pair, for each cell, and for each model. Show that these measures can predict improved robustness for one model over another and for synaptic vs. intrinsic parameters.

      The ratio of areas of the colored and non-colored regions in the whole hyperboxes (for intrinsic and synaptic conductances) is the number reported in the y-axis of the sensitivity curves when we include all conductances (and not just a pair). 

      We computed the ratios of the colored/noncolored areas in all panels in figure 6 and now report these quantities as follows, 

      "We computed the proportions of areas of the white boxes that correspond to pyloric activity. These values for the intrinsic conductances panels are PD = 0.58, LP = 0.50, PY = 0.49, and the proportions for the synaptic conductances panels are PDPY = 0.62, P DLP = 0.87, and LPPD = 0.94. The occupied areas for synaptic conductances are larger than in the intrinsic conductances panels, consistent with our finding that the circuits’ activities are more robust to changes in synaptic conductances versus changes in intrinsic conductances."

      "As before, we computed the proportion of areas of pyloric activity within the white boxes: PD = 0.61, LP = 0.55, PY = 0.52, and the proportions for the synaptic conductances panels are PDPY = 0.88, PDLP = 0.87, and LPP D = 0.83. These results provide an intuition of the complexities of GP . Not only are these regions hard-to-impossible to characterize in one circuit, but they are also different across circuits.” 

      (14) Does the sign of the synaptic weights affect the conclusions?

      We did not explore this issue because all chemical synapses in this network are inhibitory.

      (15) Line 492: typo: deltai.

      We fixed this.

      Reviewer #2 (Recommendations for the authors):

      (1) Line 301 - you can also add Williams and Fletcher 2019 Neuron.

      We added the reference. Thank you. 

      (2) Line 316 - this is a strange comment as these exact regions that were shown intrinsic plasticity (e.g., Losonczy, Attila, Judit K. Makara, and Jeffrey C. Magee. "Compartmentalized dendritic plasticity and input feature storage in neurons." Nature 452.7186 (2008): 436-441).

      We did not understand this comment. 

      (3) I found only one citation for the work of Turrigiano, the most relevant of which is only mentioned in the Method section. This is odd, as her work directly relates how synaptic conductance perturbation results in changes in intrinsic conductance.

      We included more references to the work of Turrigiano to provide more context. 

      "Desai, Niraj S., Lana C. Rutherford, and Gina G. Turrigiano. "Plasticity in the intrinsic excitability of cortical pyramidal neurons." Nature neuroscience 2, no. 6 (1999): 515-520.” "Desai, Niraj S., Sacha B. Nelson, and Gina G. Turrigiano. "Activity-dependent regulation of excitability in rat visual cortical neurons." Neurocomputing 26 (1999): 101-106.”

      (4) Line 329 - The list of citations is very limited regarding studies of ext/int balance which started really way before 2009. Please give some of the credit to the classics.

      We included the following additional references.

      Van Vreeswijk, Carl, and Haim Sompolinsky. "Chaos in neuronal networks with balanced excitatory and inhibitory activity." Science 274, no. 5293 (1996): 1724-1726.

      Rubin, Ran, L. F. Abbott, and Haim Sompolinsky. "Balanced excitation and inhibition are required for high-capacity, noise-robust neuronal selectivity." Proceedings of the National Academy of Sciences 114, no. 44 (2017): E9366-E9375.

      Wang, Xiao-Jing. "Macroscopic gradients of synaptic excitation and inhibition in the neocortex." Nature reviews neuroscience 21, no. 3 (2020): 169-178.

      Lo, Chung-Chuan, Cheng-Te Wang, and Xiao-Jing Wang. "Speed-accuracy tradeoff by a control signal with balanced excitation and inhibition." Journal of Neurophysiology 114, no. 1 (2015): 650-661.

      (5) In Figure 1B, why does it say 'OFF' when the neuron is spiking?

      The label indicates the interval of time elapsed between the first spike in the PD neuron (taken as a reference), and the last spike in the burst (PD off). 

      Summary of changes to figures:

      Figure 1:

      Fixed labels indicating bursting period and burst duration.

      Figure 5:

      Added labels in panels C and D specifying the symbol corresponding to the sigmoidal parameter.

      Additional changes

      We changed the title of the manuscript as follows:

      "Rhythmic circuit function is more robust to changes in  synaptic than intrinsic conductances." We included the following sentence at the end of the Discussion Section. 

      "We believe our results will hold for other rhythmic circuits and will be relevant for similar studies in other circuits with more complex functions.”

      We realized we made a mistake with the units for maximal conductances. They were incorrectly expressed in nS (nano Siemens) in the figure labels, and correctly expressed in micro Siemens in the methods section. This was fixed and now conductances are expressed in micro Siemens consistently in the manuscript.

    1. eLife Assessment

      This observational study from the UK Biobank provides an important investigation into the associations between menopausal hormone therapy and brain health in a large, population-based cohort of females in the UK. A convincing model of brain aging using an open source algorithm is used. While some modest adverse brain health characteristics were associated with current mHT use and older age at last use, the findings do not support a general neuroprotective effect of mHT nor severe adverse effects on the female brain. This work addresses a topic that is of grave importance since menopausal hormone therapy and its effect on the brain should be better understood in order to provide individualized effective medical support to women going through menopause.

    2. Reviewer #1 (Public review):

      Summary:

      This study takes a detailed approach to understand the effect of menopausal hormone therapy (MHT) in brain aging of females. Neuroimaging data from the UK Biobank is used to explore brain aging and shows an unexpected effect of current MHT use and poorer brain health outcomes relative to never users. There is considerable debate about the benefits of MHT and estrogens in particular for brain health, and this analysis illustrates thta the effects are certainly not straight forward and require greater considerations.

      Strengths:

      (1) The detailed approach to obtain important information about MHT use from primary care records. Prior studies have suggested that factors such as estrogen/progestin type, route of administration, duration, and timing of use relative to menopause onset can contribute to whether MHT benefits brain health.<br /> (2) Consideration of type of menopause (spontaneous, or surgical) in the analysis, as well as sensitivity diagnoses to rule out the effect being driven by those with clinical conditions<br /> (3) The incorporation of the brain age estimate along with hippocampal volume to address brain health<br /> (4) The complex data are also well explained and interpretations are reasonable.<br /> (5) Limitations of the UKbiobank data are acknowledged

      Weaknesses:

      These have since been addressed by the authors in the revision.

    3. Reviewer #2 (Public review):

      Summary:

      In this observational study, Barth et al. investigated the association between menopausal hormone therapy and brain health in middle- to older-aged women from the UK Biobank. The study evaluated detailed MHT data (never, current, or past user), duration of mHT use (age first/last used), history of hysterectomy with or without bilateral oophorectomy, APOEE4 genotype, and brain characteristics in a large, population-based sample. The researchers found that current mHT use (compared to never-users), but not past use, was associated with a modest increase in gray and white matter brain age gap (GM and WM BAG) and decrease in hippocampal volumes. No significant association was found between the age of mHT initiation and brain measures among mHT users. Longer duration of use and older age at last MHT use post-menopause were associated with higher GM and WM BAG, larger WMH volumes, and smaller hippocampal volumes. In a sub-sample, after adjusting for multiple comparisons, no significant associations were found between detailed mHT variables (formulations, route of administration, dosage) and brain measures. The association between mHT variables and brain measures was not influenced by APOEE4 allele carrier status. Women with a history of hysterectomy with or without bilateral oophorectomy had lower GM BAG compared to those without such history. Overall, these observational data suggest that the association between mHT use and brain health in women may vary depending on the duration of use and surgical history.

      Strengths:

      The study has several strengths, including a large, population-based sample of women in the UK, and comprehensive details of demographic variables such as menopausal status, history of oophorectomy/hysterectomy, genetic risk factors for Alzheimer's disease (APOE ε4 status), age at mHT initiation, age at last use, duration of mHT, and brain imaging data (hippocampus and WMH volume).

      In a sub-sample, the study accessed detailed mHT prescription data (formulations, route of administration, dosage, duration), allowing the researchers to study how these variables were associated with brain health outcomes. This level of detail is generally missing in observational studies investigating the association of mHT use with brain health.

      Weaknesses:

      While the study has many strengths, it also has some weaknesses. These weaknesses were properly discussed throughout the article. The manuscript has indicated that the need of mHT use which might be associated with these symptoms may be indicators of preexisting neurological changes, potentially reflecting worse brain health scores, including higher BAG and lower hippocampal volume and/or higher WMH. The authors noted that the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, limiting the study's ability to understand how these variables influence outcomes. The authors also highlighted that these results don't reflect causal relationships. The authors caution that these findings should not guide individual-level decisions regarding the benefits versus risks of mHT use. However, the study raises new questions that should be addressed by randomized clinical trials to investigate the varying effects of MHT on brain health and dementia risk.

    4. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1 (Public review):

      Summary:

      This study takes a detailed approach to understanding the effect of menopausal hormone therapy (MHT) in the brain aging of females. Neuroimaging data from the UK Biobank is used to explore brain aging and shows an unexpected effect of current MHT use and poorer brain health outcomes relative to never users. There is considerable debate about the benefits of MHT and estrogens in particular for brain health, and this analysis illustrates that the effects are certainly not straightforward and require greater consideration.

      Strengths:

      (1) The detailed approach to obtaining important information about MHT use from primary care records. Prior studies have suggested that factors such as estrogen/progestin type, route of administration, duration, and timing of use relative to menopause onset can contribute to whether MHT benefits brain health.

      (2) Consideration of type of menopause (spontaneous, or surgical) in the analysis, as well as sensitivity diagnoses to rule out the effect being driven by those with clinical conditions.

      (3) The incorporation of the brain age estimate along with hippocampal volume to address brain health.

      (4) The complex data are also well explained and interpretations are reasonable.

      (5) Limitations of the UK Biobank data are acknowledged

      We thank the reviewer for their time and the positive evaluation of our manuscript.

      Weaknesses:

      (1) Lifestyle factors are listed and the authors acknowledge group differences (at least between current users and never users of MHT). I was not able to find these analyses showing these differences.

      We highlighted and tested for group differences in lifestyle scores, and the results are shown in Table 1-3, column p-value. As highlighted in the method section (page 9): “The lifestyle score was calculated using a published formula (69), and included data on sleep, physical activity, nutrition, smoking, and alcohol consumption (see supplementary Note 3, Table S2)”. In line with reviewer 1 suggestion to the authors, we now included an additional table testing for group differences in the specific lifestyle factors constituting the lifestyle score in the supplementary materials (Table S2). Please find a more detailed response below (Recommendations for the authors, Response to Comment 1).

      (2) The distribution of women who were not menopausal was unequal across groups, and while the authors acknowledge this, one wonders to what extent this explains the observed findings.

      We agree with the reviewer that the unequal distribution of women across groups can influence the observed findings. We have made minor edits to highlight this important topic more explicitly in the discussion:

      Discussion (page 21): “Current MHT users were significantly younger than past- and never-users, and around 67 % were menopausal relative to over 80% in the past- and never-user groups. The unequal distribution of age and menopausal status across groups may have influenced the observed findings. For instance, a larger proportion of the current users might be in the perimenopausal phase, which is often associated with debilitating neurological and vasomotor symptoms (1). MHT is commonly prescribed to minimize such symptoms. Although MHT initiation during perimenopause has been associated with improved memory and hippocampal function, as well as lower AD risk later in life (15), the need for MHT might in itself be an indicator of neurological changes (71); here potentially reflected in higher BAG and lower hippocampal volumes. After the transition to menopause, symptoms might subside and some perimenopausal brain changes might revert or stabilize in the postmenopausal phase 5. Although the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, our results might be capturing subtle disturbances during perimenopause that later stabilize. This could explain why the largely postmenopausal groups of past MHT users and never-users present with lower GM and WM BAG than the current user group. Considering the critical window hypothesis emphasizing perimenopause as a key phase for MHT action (29,43), future longitudinal studies are crucial to clarify the interplay between neurological changes and MHT use across the menopause transition.”

      Discussion (page 25): “In addition, previous studies highlight that UK Biobank participants are considered healthier than the general population based on several lifestyle and health-related factors (89, 90). This healthy volunteer bias increases with age, likely resulting in a disproportionate number of healthier older adults. Together with the imbalance in age distributions across groups, this might explain the less apparent brain aging in the older MHT user groups. We have previously highlighted that age is negatively associated with the number of APOE ε4 carriers in the UK Biobank (21), which is indicative of survivor bias.”

      (3) While the interpretations are reasonable, and relevant theories (healthy cell & critical window) are mentioned, the discussion is missing a more zoomed-out perspective of the findings. While I appreciate wanting to limit speculation, the reader is left having to synthesize a lot of complex details on their own. A particularly difficult finding to reconcile is under what conditions these women benefit from MHT and when do they not (and why that may be).

      We thank the reviewer for this comment. As the presented data is cross-sectional and does not enable causal inference, we have refrained from a more zoomed-out interpretation of the results to avoid undue speculations. However, where applicable, we have discussed our findings in a broader context such as the effects of MHT use on the brain across the menopausal transition (discussion page 21) and the effects of MHT use on the brain in the presence and absence of bilateral oophorectomy and/or hysterectomy (discussion page 25).

      To best inform the reader about the scope of our paper, we would like to highlight the following sentences in our discussion (page 24):

      “The current work represents the most comprehensive study of detailed MHT data, APOE ε4 genotype, and several brain measures in a large population-based cohort to date. Overall, our findings do not unequivocally support general neuroprotective effects of MHT, nor do they indicate severe adverse effects of MHT use on the female brain. The results suggest subtle yet complex relationships between MHT’s and brain health, highlighting the necessity for a personalized approach to MHT use. Importantly, our analyses provide a broad view of population-based associations and are not designed to guide individual-level decisions regarding the benefits versus risks of MHT use.”

      And the conclusion (page 25): “In conclusion, our findings suggest that associations between MHT use and female brain health might vary depending on duration of use and past surgical history. Although the effect sizes were generally modest, future longitudinal studies and RCTs, particularly focused on the perimenopausal transition window, are warranted to fully understand how MHT use influences female brain health. Importantly, considering risks and benefits, decisions regarding MHT use should be made within the clinical context unique to each individual.”

      Reviewer #1 (Recommendations for the authors):

      Can the authors provide:

      (1) More information about which aspects of lifestyle factors were different between the groups, and how these factors may have contributed to the observed findings (if possible, without burying this information in the supplemental)?

      We thank the reviewer for this suggestion. We now added a table comparing lifestyle factors contained in the lifestyle score by MHT user status using t-tests (continuous variables) or χ2 tests (see Table S2). The results are referred to in the main manuscript result section under “Sample characteristics”, and the table (Table S2) is provided in the supplements not to overburden the main text, in line with input from reviewer 3.

      We updated the main text to refer to Table S2 and updated the supplementary Note 3 (page 2-3) to include the results of the comparison of the lifestyle factors contained in the lifestyle score by MHT user status.

      Methods, page 9:“The lifestyle score was calculated using a published formula (69), and included data on sleep, physical activity, nutrition, smoking, and alcohol consumption (see supplementary Note 3, Table S2).”

      Results, page 13: “Sample demographics including lifestyle score, stratified by MHT user group, surgical history among MHT users, and estrogen only MHT or combined MHT use, are summarized in Table 1, 2 and 3, respectively. MHT user group differences for each lifestyle factor contained in the lifestyle score are shown in Table S2.”

      “Note 3| Lifestyle Score

      The lifestyle score was calculated based on sleep duration, time spent watching television, current and past smoking status, alcohol consumption frequency, physical activity level (number of days per week of moderate/vigorous activity for at least 10 minutes), intake of fruits and vegetables, and intake of oily fish, beef, lamb/mutton, pork and processed meat (for details see (10)). Each unhealthy lifestyle factor was scored with 1 point (e.g., smoking), and participants points were summed to generate an unweighted score (from 0-9): the higher the lifestyle score, the unhealthier the participant’s lifestyle.

      A comparison of the lifestyle factors contained in the lifestyle score by MHT user status is presented in Table S2. In summary, we found that current MHT were more often smokers than never-users, had a higher alcohol intake than never- and past MHT users, reported the lowest fruit and vegetable intake relative to never-users and past MHT users, and stated lower moderate activity levels relative to past MHT users. Past MHT users reported higher alcohol intake than never-users, spend more time watching TV relative to never- and current-users, consumed more beef, pork, lamb/mutton, and processed meat than never-users, and reported lower vigorous activity levels relative to never-users. However, oily fish intake and fruit and vegetable intake was higher among past MHT users relative to never-and current-users. Self-reported sleep duration did not differ between MHT user groups.”

      (2) A greater description of the 2 main theories of MHT effects on the brain (healthy cell vs critical window). Can the authors also provide a more thorough explanation for how the findings fit with these theories.

      We thank the reviewer for this comment. We have described our findings in the context of the critical window hypothesis (discussion, page 21, paragraph 2), the healthy cell bias hypothesis (discussion, page 22, paragraph 3), and healthy user bias hypothesis (discussion, page 22, paragraph 4). We refrained from a more thorough explanation to avoid undue speculations.

      (3) Reflect more on what the findings may indicate as to who benefits from MHT, and why. There are some references that the authors may want to add, particularly related to recent findings from premenopausal bilateral oophortectomies that also speak to when (and for whom) MHT use might benefit.

      We thank the reviewer for this feedback. We have included additional references in the revised manuscript as follows:

      Discussion, page 23: “It is also possible that the timing between MHT use and surgery is more tightly controlled and therefore more beneficial for brain aging (43). For instance, studies suggest that MHT may mitigate the potential long-term adverse effects of bilateral oophorectomy before natural menopause on bone mineral density as well as cardiovascular, cognitive and mental health (79-81). In addition, a 2024 UK Biobank study found that ever used MHT was associated with decreased odds of Alzheimer’s disease in women with bilateral oophorectomy (82).”  

      (79) Blumel JE, Arteaga E, Vallejo MS, et al. Association of bilateral oophorectomy and menopause hormone therapy with mild cognitive impairment: the REDLINC X study. Climacteric 2022;25:195-202.

      (80) Kaunitz AM, Kapoor E, Faubion S. Treatment of Women After Bilateral Salpingo-oophorectomy Performed Prior to Natural Menopause. JAMA 2021;326:1429-1430.

      (81) Stuursma A, Lanjouw L, Idema DL, de Bock GH, Mourits MJE. Surgical Menopause and Bilateral Oophorectomy: Effect of Estrogen-Progesterone and Testosterone Replacement Therapy on Psychological Well-being and Sexual Functioning; A Systematic Literature Review. J Sex Med 2022;19:1778-1789.

      (82) Calvo N, McFall GP, Ramana S, et al. Associated risk and resilience factors of Alzheimer's disease in women with early bilateral oophorectomy: Data from the UK Biobank. J Alzheimers Dis 2024;102:119-128.

      Reviewer #2 (Public review):

      Summary:

      In this observational study, Barth et al. investigated the association between menopausal hormone therapy and brain health in middle- to older-aged women from the UK Biobank. The study evaluated detailed MHT data (never, current, or past user), duration of mHT use (age first/last used), history of hysterectomy with or without bilateral oophorectomy, APOEE4 genotype, and brain characteristics in a large, population-based sample. The researchers found that current mHT use (compared to never-users), but not past use, was associated with a modest increase in gray and white matter brain age gap (GM and WM BAG) and a decrease in hippocampal volumes. No significant association was found between the age of mHT initiation and brain measures among mHT users. Longer duration of use and older age at last MHT use post-menopause were associated with higher GM and WM BAG, larger WMH volumes, and smaller hippocampal volumes. In a sub-sample, after adjusting for multiple comparisons, no significant associations were found between detailed mHT variables (formulations, route of administration, dosage) and brain measures. The association between mHT variables and brain measures was not influenced by APOEE4 allele carrier status. Women with a history of hysterectomy with or without bilateral oophorectomy had lower GM BAG compared to those without such a history. Overall, these observational data suggest that the association between mHT use and brain health in women may vary depending on the duration of use and surgical history.

      Strengths:

      (1) The study has several strengths, including a large, population-based sample of women in the UK, and comprehensive details of demographic variables such as menopausal status, history of oophorectomy/hysterectomy, genetic risk factors for Alzheimer's disease (APOE ε4 status), age at mHT initiation, age at last use, duration of mHT, and brain imaging data (hippocampus and WMH volume).

      (2) In a sub-sample, the study accessed detailed mHT prescription data (formulations, route of administration, dosage, duration), allowing the researchers to study how these variables were associated with brain health outcomes. This level of detail is generally missing in observational studies investigating the association of mHT use with brain health.

      We thank the reviewer for their time and the positive evaluation of our manuscript.

      Weaknesses:

      (1) While the study has many strengths, it also has some weaknesses. As highlighted in an editorial by Kantarci & Manson (2023), women with symptoms such as subjective cognitive problems, sleep disturbances, and elevated vasomotor symptoms combined with sleep disturbances tend to seek mHT more frequently than those without these symptoms. The authors of this study have also indicated that the need of mHT use which might be associated with these symptoms may be indicators of preexisting neurological changes, potentially reflecting worse brain health scores, including higher BAG and lower hippocampal volume and/or higher WMH. However, among current users, how many of these women have these symptoms could not be reported in the study. Women with these vasomotor symptoms who are using mHT are more likely to stay longer in the healthcare system compared with those without these symptoms and no MHT use history. The authors noted that the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, limiting the study's ability to understand how these variables influence outcomes.

      We thank the reviewer for the succint synopsis of the limitations highlighted in discussion, page 21. We have now added the mentioned reference, 2023 editoral by Kantarci & Manson, to the discussion as well (see reference 71).

      Discussion (page 21): “Current MHT users were significantly younger than past- and never-users, and around 67 % were menopausal relative to over 80% in the past- and never-user groups. The unequal distribution of age and menopausal status across groups may have influenced the observed findings. For instance, a larger proportion of the current users might be in the perimenopausal phase, which is often associated with debilitating neurological and vasomotor symptoms (1). MHT is commonly prescribed to minimize such symptoms. Although MHT initiation during perimenopause has been associated with improved memory and hippocampal function, as well as lower AD risk later in life (15), the need for MHT might in itself be an indicator of neurological changes (71); here potentially reflected in higher BAG and lower hippocampal volumes. After the transition to menopause, symptoms might subside and some perimenopausal brain changes might revert or stabilize in the postmenopausal phase 5. Although the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, our results might be capturing subtle disturbances during perimenopause that later stabilize. This could explain why the largely postmenopausal groups of past MHT users and never-users present with lower GM and WM BAG than the current user group. Considering the critical window hypothesis emphasizing perimenopause as a key phase for MHT action (29,43), future longitudinal studies are crucial to clarify the interplay between neurological changes and MHT use across the menopause transition.”

      (2)  Earlier observational studies have reported conflicting results regarding the association between mHT use and the risk of dementia and brain health. Contrary to some observational studies, three randomized trials (WHI, KEEPS, ELITE) (Espeland et al 2013, Gleason et al 2015; Henderson et al 2016) demonstrated neither beneficial nor harmful effects of mHT (with varying doses and formulations) when initiated closer to menopause (<5 years). While strong efforts were made to run proper statistical analyses to investigate the association between mHT use and brain health, these results reflect mainly associations, but not causal relationships as also stated by the authors.

      We thank the reviewer for pointing that out.

      (3)  Furthermore, observational studies have intrinsic limitations, such as a lack of control over switching mHT doses and formulations, a lack of laboratory measures to confirm mHT use, and reliance on self-reported data, which may not always be reliable. The authors caution that these findings should not guide individual-level decisions regarding the benefits versus risks of mHT use. However, the study raises new questions that should be addressed by randomized clinical trials to investigate the varying effects of MHT on brain health and dementia risk.

      We thank the reviewer for making our efforts in providing proper disclaimers in the discussion visible.

      Reviewer #2 (Recommendations for the authors):

      (1) The study could benefit from extending these findings by adding plasma biomarkers of AD and PET imaging markers to further study the association of mHT variables with brain health.

      We agree with the reviewer that such markers would be beneficial for elucidating the association between MHT variables and brain health. Unfortunately, these markers are not readily available in the UK Biobank.

      (2) The study's reliance on a predominantly white cohort limits the generalizability of the findings to more diverse populations. This homogeneity may not capture the full spectrum of responses to MHT across different ethnic and genetic backgrounds.

      We fully agree with the reviewers statement and state this limitation in the discussion (page 25) as follows:

      “In addition to these inherent biases in aging cohorts, the ethnic background of the sample is homogeneous (> 96% white), further reducing the generalizability of the results.”

      (3) The study may benefit by editing the following information in the introduction: "In summary, WHIMS, HERS, and KEEPS mainly relied on orally administered CEE in older-aged or recently postmenopausal females." KEEPS used two routes and formulations (transdermal estradiol and oCEE, both with micronized progesterone).

      We thank the reviewer for catching this oversight. We removed the sentence to avoid ambiguities and revised the sentence specifically refering to the KEEPS study as follows:

      Introduction, page 3: “In contrast, administering oral CEE or transdermal estradiol plus micronized progesterone in recently postmenopausal females did not alter cognition in the Kronos Early Estrogen Prevention Study (KEEPS) (28).”

      (4) The study may benefit by editing the following statement in the introduction: "oral CEE use in combination with MPA seems to increase the risk for AD regardless of timing": I would suggest revising this statement, which is based on review article 29. The statement of the adverse effect of oCEE regardless of the time of start contradicts earlier randomized clinical findings. I think it is important to make a distinction between the outcomes of randomized control trials and observational studies. The WMIHS (Shumaker et al., 2003) (randomized control trial) reported that there was an increased risk of dementia for women (who were more than 10 years from the onset of menopause when the therapy was initiated) in oCEE + MPA compared to placebo. Two other long-duration randomized trials tested the effect of oral oestrogen and progesterone treatment on cognitive function in women who started treatment shortly after menopause (within 3 or 6 years) did not find evidence that treatment benefits or harms cognitive function compared with placebo (Gleason et al., 2015; Henderson et al., 2016). A short-term (4 months) randomized trial (Maki et al 2007 (Maki et al., 2007) (mentioned in ref 29) reported a potential negative effect of CEE/MPA on verbal memory in women who started HT shortly after menopause (within 3 years). The study did not investigate the risk of dementia, and the duration of use of HT was short-term.

      We thank the reviewer for this detailed input. After checking the provided references, we rephrased the sentence as follows:

      Introduction, page 4:“Although emerging evidence supports this hypothesis (30, 31), oral CEE use in combination with MPA has been found to increase the risk for memory decline regardless of timing (26, 29, 32).”

      We believe this formulation is more in line with the evidence provided by Shumaker et al. 2003, Maki et al. 2007 and the other references provided in the review paper by Maki and colleagues (mentioned in ref. 29). The reviewer further refers to Gleason et al. 2015 and Henderson et al. 2016, however both RCTs use micronized progesterone, not MPA, thereby not supporting the statement.

      (26) Shumaker SA, Legault C, Rapp SR, et al. Estrogen plus progestin and the incidence of dementia and mild cognitive impairment in postmenopausal women: the Women's Health Initiative Memory Study: a randomized controlled trial. JAMA 2003;289:2651-2662.

      (29) Maki PM. Critical window hypothesis of hormone therapy and cognition: a scientific update on clinical studies. Menopause 2013;20:695-709.

      (32) Maki PM, Gast MJ, Vieweg AJ, Burriss SW, Yaffe K. Hormone therapy in menopausal women with cognitive complaints: a randomized, double-blind trial. Neurology 2007;69:1322-1330.

      Reviewer #3 (Public review):

      In this study Barth et al. present results of detailed analyses of the relationships between menopausal hormone therapy (MHT), APOE ε4 genotype, and measures of anatomical brain age in women in the UK Biobank. While past studies have investigated the links between some of these variables (including works by the authors themselves), this new study adds more detailed MHT variables, surgical status, and additional brain aging measures. The UK biobank sample is large, but it is a population cohort and many of the MHT measures are self-reported (as the authors point out). However, the authors present a solid analysis of the available information which shows associations between MHT user status, length of MHT use, as well as surgical status with brain age. However, as the authors themselves state, the results do not unequivocally support the neuroprotective or adverse effect of MHT on the brain. I think this work strengthens the case for the need of better-designed longitudinal studies investigating the effect of MHT on the brain in the peri/post-menopausal stage.

      Strengths:

      (1) The authors addressed the statistical analyses rigorously. For example, multiple testing corrections, outlier removal, and sensitivity analysis were performed carefully. Ample background information is provided in the introduction allowing even individuals not familiar with the field to understand the motivation behind the work. The discussion section also does a great job of addressing open questions and limitations. Very detailed results of all statistical tests are provided either in the main text or in the supplementary information.

      We thank the reviewer for their time and the positive evaluation of our manuscript.

      Weaknesses:

      (1) For me, the biggest weakness was the presentation of the results. As many variables are involved and past studies have investigated several of these questions, it would have helped to better clarify the analysis and questions that are addressed by this study in particular and what sets this work apart from past studies. The information is present in the manuscript but better organization might have helped. For example, a figure depicting the key questions near the beginning of the manuscript would have been very helpful for me. The Tables also contain a lot of information but I wonder if there might be a way to capture the most relevant information more succinctly (either in Table format or in a figure) for the main text.

      We thank the reviewer for this comment. We do agree that with the large number of analyses it can be hard to keep an overview. We now added a Figure summarizing the main and sensitity analyses by sample.

      (2) Another concern I had was the linear models investigating the effects of these MHT variables on the brain age gap. The authors have included "age" as one of the parameters in this analysis. I wonder if adding a quadratic age factor age2 in the model might have improved the fit since many brain phenotypes tend to show quadratic brain age effects in the 40 to 80-year age range.

      We thank the reviewer for this suggestion. We have rerun the main analysis in the whole sample (model 1) with age squared as an additional covariate, and compared the gray matter brain age gap model fits using the corrected Akaike Information Criterion (AIC). All models with age squared had a better model fit than models without age squared (see Author response table 1). Hence, in the revised manuscript, we added a sensitivity analysis rerunning the model 1 with age squared to account for potential non-linear effect. The results were largely consistent. The manuscript was revised as follows to reflect the added analysis:

      Sensitivity analysis (Methods, Page 11): “To test whether the results were influenced by the inclusion of participants with ICD-10 diagnosis or by non-linear effects of age, the main analyses (models 1-2) were re-run excluding the sub-sample with diagnosed brain disorders (see supplementary Note 2) or adding age(2) as additional covariate, respectively.”

      Sensitivity analysis (Results, Page 20): “The results were consistent after removing participants with ICD-10 diagnoses known to impact the brain (see Table S9 for model 1 analyses and Table S10 for model 2 analyses), after additionally adjusting for age(2) (see Table S11), and after removing extreme values (see Table S12 for model 1 analyses).”

      Author response table 1.

      Gray matter brain age gap model selection based on corrected Akaike Information Criterion (AICc)

      Abbreviations and explanations of parameters: MHT = menopausal hormone therapy, K = number of estimated parameters for each model, AICc = the information criterion requested for each model, ΔAICc = the appropriate delta AIC component depending on the information criteria selectedModelLik = the relative likelihood of the model given the data, AICcWT = Akaike weights to indicate the level of support in favor of any given model being the most parsimonious among the candidate model sets, LL = log-likelihood of each model.

      Reviewer #3 (Recommendations for the authors):

      (1) Please note typo in Figures 2 and 3 legend "GM WM".

      We thank the reviewer for catching this typo and we changed it to BAG GM and BAG WM for all Figures for consistency.

    1. eLife Assessment

      This valuable study uses dynamic metabolic models to compare perturbation responses in a bacterial system, analyzing whether they return to their steady state or amplify beyond the initial perturbation. The evidence supporting the emergent properties of perturbed metabolic systems to network topology and sensitivity to specific metabolites is solid.

    2. Reviewer #1 (Public review):

      Summary:

      The author studied metabolic networks for central metabolism, focusing on how system trajectories returned to their steady state. To quantify the response, systematic perturbation was performed in simulation and the maximal destabilization away from steady state (compared with initial perturbation distance) was characterized. The author analyzed the perturbation response and found that sparse network and networks with more cofactors are more "stable", in the sense that the perturbed trajectories have smaller deviation along the path back to the steady state.

      Strengths and major contributions:

      The author compared three metabolic models and performed systematic perturbation analysis in simulation. This is the first work characterized how perturbed trajectories deviate from equilibrium in large biochemical systems and illustrated interesting findings about the difference between sparse biological systems and randomly simulated reaction networks.

      Discussion and impact for the field:

      Metabolic perturbation is an important topic in cell biology and has important clinical implication in pharmacodynamics. The computational analysis in this study provides an initiative for future quantitative analysis on metabolism and homeostasis.

      Comments on latest version:

      In the latest version of this work, the author included NADH, NADPH into the analysis, and perform some comparison about sensitivity analysis. I think this paper is ready to be finalized, and many open questions inspired from this work can be studied in future.

    3. Reviewer #2 (Public review):

      The authors have conducted a valuable comparative analysis of perturbation responses in three nonlinear kinetic models of E. coli central carbon metabolism found in the literature. They aimed to uncover commonalities and emergent properties in the perturbation responses of bacterial metabolism. They discovered that perturbations in the initial concentrations of specific metabolites, such as adenylate cofactors and pyruvate, significantly affect the maximal deviation of the responses from steady-state values. Furthermore, they explored whether the network connectivity (sparse versus dense connections) influences these perturbation responses. The manuscript is reasonably well written.

      Comments on latest version:

      The authors have adequately addressed my concerns.

    4. Author response:

      The following is the authors’ response to the previous reviews

      Reply to the comments of the second referee

      We sincerely appreciate the positive evaluation and the useful suggestions on our manuscript.

      (1) The authors identified key metabolites affecting responses to perturbations in two ways: (i) by fixing a metabolite's value and (ii) by performing a sensitivity analysis. It would be helpful for the modeling community to understand better the differences and similarities in the obtained results. Do both methods identify substrate-level regulators? Is freezing a metabolite's dynamics dramatically changing the metabolic response (and if yes, which ones are so different in the two cases)? Does the scope of the network affect these differences and similarities? 

      Thank you for these suggestions. We compared the Sobolʼ total sensitivity index with the absolute values of the change in the response coefficient (Figure S6 in the revised manuscript). There is no clear relationship between the two quantities. The Sobolʼ sensitivity analysis quantifies how a perturbation on the concentration of a metabolite X contributes to the overall dynamics. On the other hand, the analysis in which metabolitesʼ concentrations are fixed measures how strongly metabolite X helps propagate the perturbations on the other metabolites throughout the metabolic network. In other words, in the Sobolʼ analysis, we evaluate the outcome when the perturbation is applied directly to metabolite X, whereas in the fixing-metabolites analysis, we consider perturbations applied to other metabolites and assess how X influences those perturbations. We believe this conceptual difference explains why the two quantities do not correlate. We suspect that this lack of correlation is independent of the networkʼs scope, because each method evaluates a different aspect of the system.  We would say that both methods identify the effect of the metabolite dynamics on the overall dynamics whatever the form is, i.e. the methods do not distinguish the perturbation on the metabolite affecting the overall dynamics by whether the stoichiometric (reactant) way or, the substrate-level regulations. Thus, identifying the substrate-level regulation by utilizing the methods would be challenging. 

      (2) Regarding the issues the authors encountered when performing the sensitivity analysis, they can be approached in two ways. First, the authors can check the methods for computing conserved moieties nicely explained by Sauro's group (doi:10.1093/bioinformatics/bti800) and compute them for large-scale networks (but beware of metabolites that belong to several conserved pools). Otherwise, the conserved pools of metabolites can be considered as variables in the sensitivity analysis-grouping multiple parameters is a common approach in sensitivity analysis. 

      Thank you for this helpful suggestion. Following the method described in the reference, we have computed the Sobolʼ sensitivity index of NADH, NADPH, and Q8H2 (with their counterparts algebraically solved and treated as dependent variables). We have updated Figure S5 accordingly.

    1. eLife Assessment

      This important study includes convincing evidence to show that behavioral measures and hippocampal representations when animals use task-relevant information and ignore irrelevant information do not depend on the medial prefrontal cortex. The results are expected to be of interest to those studying neural mechanisms of cognitive control and functions of associational brain regions.

    2. Reviewer #1 (Public review):

      Summary:

      The authors examine the role of the medial prefrontal cortex (mPFC) in cognitive control, i.e. the ability to use task-relevant information and ignore irrelevant information, in the rat. According to the central-computation hypothesis, cognitive control in the brain is centralized in the mPFC and according to the local hypothesis, cognitive control is performed in task-related local neural circuits. Using the place avoidance task which involves cognitive control, it is predicted that if mPFC lesions affect learning, this would support the central computation hypothesis whereas no effect of lesions would rather support the local hypothesis. The authors thus examine the effect of mPFC lesions in learning and retention of the place avoidance task. They also look at functional interconnectivity within a large network of areas that could be activated during the task by using cytochrome oxydase, a metabolic marker. In addition, electrophysiological unit recordings of CA1 hippocampal cells are made in a subset of (mPFC-lesioned or intact) animals to evaluate overdispersion, a firing property that reflects cognitive control in the hippocampus. The results indicate that mPFC lesions disrupted correlations of activity between functionally-related regions. Behaviorally, lesions did not impair place avoidance learning and retention (though flexibility was altered during conflict training). In addition, hippocampal place cell overdispersion was decreased in lesioned rats only in the absence of cognitive control challenge (pretraining). Cognitive control seen in hippocampal place cell activity (alternation of frame-specific firing) was not affected by the lesion. Overall, the absence of effects of mPFC lesions on cognitive control in the task or in hippocampal place cells firing support the local hypothesis.

      Strengths:

      Straightforward hypothesis: clarification of the involvement of the mPFC in the brain is expected and achieved. Appropriate use of fully mastered methods (active place avoidance task, electrophysiological unit recordings, measure of metabolic marker cytochrome oxidase) and rigorous analysis of the data. The conclusion is strongly supported by the data.

      Weaknesses:

      No notable weaknesses in the conception, making of the study and data analysis.

      Comments on revisions:

      The authors have satisfactorily addressed all my comments in the revised version.

    3. Reviewer #2 (Public review):

      Park et al. set out to test two competing hypotheses about the role of the medial prefrontal cortex (PFC) in cognitive control, the ability to use task-relevant cues and ignore task-irrelevant cues to guide behavior. The "central computation" hypothesis assumes that cognitive control relies on computations performed by the PFC, which then interacts with other brain regions to accomplish the task. Alternatively, the "local computation" hypothesis suggests that computations necessary for cognitive control are carried out by other brain regions that have been shown to be essential for cognitive control tasks, such as the dorsal hippocampus and the thalamus. If the central computation hypothesis is correct, PFC lesions should disrupt cognitive control. Alternatively, if the local computation hypothesis is correct, cognitive control would be spared after PFC lesions. The task used to assess cognitive control is the active place avoidance task in which rats must avoid a sector of a rotating arena using the stationary room cues and ignoring the local olfactory cues on the rotating platform. Performance on this task has previously been shown to be disrupted by hippocampal lesions and hippocampal ensembles dynamically represent the room and arena depending on the animal's proximity to the shock zone. They found no group (lesion vs. sham) differences in the three behavioral parameters tested: distance traveled, latency to enter the shock zone, and number of shock zone entries for both the standard task and the "conflict" task in which the shock zone was rotated by 180 degrees. The only significant difference was the savings index; the lesion group entered the new shock zone more often than the sham group during the first 5 minutes of the second conflict session. This deficit was interpreted as a cognitive flexibility deficit rather than a cognitive control failure. Next, the authors compared cytochrome oxidase activity between sham and lesion groups in 14 brain regions and found that only the amygdala shows significant elevation in the lesion vs. sham group. Pairwise correlation analysis revealed a striking difference between groups, with many correlations between regions lost in the lesion group (between reuniens and hippocampus, reuniens and amygdala and a correlation between dorsal CA1 and central amygdala that appeared in the lesion group and were absent in the sham group. Finally, the authors assessed dorsal hippocampal representations of the spatial frame (arena vs. room) and found no differences between lesion and sham groups. The only difference in hippocampal activity was reduced overdispersion in the lesion group compared to the sham group on the pretraining session only and this difference disappeared after the task began. Collectively, the authors interpret their findings as supporting the local computation hypothesis; computations necessary for cognitive control occur in brain regions other than the PFC.

      Strengths:

      The data were collected in a rigorous way with experimental blinding and appropriate statistical analyses.<br /> Multiple approaches were used to assess differences between lesion and sham groups, including behavior, metabolic activity in multiple brain regions, and hippocampal single unit recording.

      Weaknesses:

      Only male rats were used with no justification provided for excluding females from the sample.

      The conceptual framework used to interpret the findings was to present two competing hypotheses with mutually exclusive predictions about the impact of PFC lesions on cognitive control. The authors then use mainly null findings as evidence in support of the local computation hypothesis. They acknowledge that some people may question the notion that the active place avoidance task indeed requires cognitive control, but then call the argument "circular" because PFC has to be involved in cognitive control. This assertion does not address the possibility that the active place avoidance task simply does not require cognitive control.

      The authors did not link the CO activity with the behavioral parameters even though the CO imaging was done on a subset of the animals that ran the behavioral task nor do they make any attempt to interpret these findings in light of the two competing hypotheses posed in the introduction. Moreover, the discussion is lacking any mechanistic interpretations of the findings. For example, there are no attempts to explain why amygdala activity and its correlation with dCA1 activity might be higher in the PFC lesioned group.

      Publishing null results is important to avoid wasting animals, time, and money. This study's results will have a significant impact on how the field views the role of the PFC in cognitive control. Whether or not some people reject the notion that the active place avoidance task measures cognitive control, the findings are solid and can serve as a starting point for generating hypotheses about how brain networks change when deprived of PFC input.

    4. Reviewer #3 (Public review):

      Summary:

      This study by Park and colleagues investigated how the medial prefrontal cortex (mPFC) influences behavior and hippocampal place cell activity during a two-frame active place avoidance task in rats. Rats learned to avoid the location of mild shock within a rotating arena, with the shock zone being defined relative to distal cues in the room. Permanent chemical lesions of the mPFC did not impair the ability to avoid the shock zone by using the distal cues and ignoring proximal cues in the arena. In parallel, hippocampal place cells alternated between two spatial tuning patterns, one anchored to the distal cues and the other to the proximal cues, and this alteration was not affected by the mPFC lesion. Based on these findings, the authors argue that the mPFC is not essential for differentiating between task-relevant and irrelevant information.

      Strengths:

      This study was built on substantial work by the Fenton lab that validated their two-frame active place avoidance task and provided sound theoretical and analytical foundations. Additionally, the effectiveness of mPFC lesions was validated by several measures, enabling the authors to base their argument on the lack of lesion effects on behavior and place cell dynamics.

      Weaknesses:

      The authors define cognitive control as "the ability to judiciously use task-relevant information while ignoring salient concurrent information that is currently irrelevant for the task." (Lines 77-78). This definition is much simpler than the one by Miller and Cohen: "the ability to orchestrate thought and action in accordance with internal goals (Ref. 1)" and by Robbins: "processes necessary for optimal scheduling of complex sequence of behaviour." (Dalley et al., 2004, PMID: 15555683). Differentiating between task-relevant and irrelevant information is required in various behavioral tasks, such as differential learning, reversal learning, and set-shifting tasks. Previous rodent behavioral studies have shown that the integrity of the mPFC is necessary for set-shifting but not for differential or reversal learning (e.g., Enomoto et al., 2011, PMID: 21146155; Cho et al., 2015, PMID: 25754826). In the present task design, the initial training is a form of differential learning between proximal and distal cues, and the conflict training is akin to reversal learning. Therefore, the lack of lesion effects is somewhat expected. It would be interesting to test whether mPFC lesions impair set-shifting in their paradigm (e.g., the shock zone initially defined by distal cues and later by proximal cues). If the mPFC lesions do not impair this ability and associated hippocampal place dynamics, it will provide strong support for the authors' local-computation hypothesis.

      Comments on revisions:

      The authors fully addressed my comments. I do not have any additional suggestions.

    5. Author response:

      The following is the authors’ response to the original reviews

      Reviewer #1 (Public review): 

      Summary:

      The authors examine the role of the medial prefrontal cortex (mPFC) in cognitive control, i.e. the ability to use task-relevant information and ignore irrelevant information, in the rat. According to the central-computation hypothesis, cognitive control in the brain is centralized in the mPFC and according to the local hypothesis, cognitive control is performed in task-related local neural circuits. Using the place avoidance task which involves cognitive control, it is predicted that if mPFC lesions affect learning, this would support the central computation hypothesis whereas no effect of lesions would rather support the local hypothesis. The authors thus examine the effect of mPFC lesions in learning and retention of the place avoidance task. They also look at functional interconnectivity within a large network of areas that could be activated during the task by using cytochrome oxidase, a metabolic marker. In addition, electrophysiological unit recordings of CA1 hippocampal cells are made in a subset of (lesioned or intact) animals to evaluate overdispersion, a firing property that reflects cognitive control in the hippocampus. The results indicate that mPFC lesions do not impair place avoidance learning and retention (though flexibility is altered during conflict training), do not affect cognitive control seen in hippocampal place cell activity (alternation of frame-specific firing), a measure of location-specific firing variability, in pretraining. It nevertheless has some effect on functional interconnections. The results overall support the local hypothesis. 

      Strengths:

      Straightforward hypothesis: clarification of the involvement of the mPFC in the brain is expected and achieved. Appropriate use of fully mastered methods (behavioral task, electrophysiological recordings, measure of metabolic marker cytochrome oxidase) and rigorous analysis of the data. The conclusion is strongly supported by the data. 

      Weaknesses:

      No notable weaknesses in the conception, making of the study, and data analysis. The introduction does not mention important aspects of the work, i.e. cytochrome oxidase measure and electrophysiological recordings. The study is actually richer than expected from the introduction. 

      The revised Introduction now includes:

      “We used cytochrome oxidase, a metabolic marker of baseline neuronal activity, to confirm the mPFC lesions were effective and that there are non-local network consequences despite the local lesion. We first evaluated cytochrome oxidase activity in regions known to be associated with performance in the active place avoidance task, or regions with known connectivity to the mPFC. We then evaluated covariance of activity amongst the regions in an effort to detect network consequences of the lesion.”

      Reviewer #2 (Public review): 

      Park et al. set out to test two competing hypotheses about the role of the medial prefrontal cortex (PFC) in cognitive control, the ability to use task-relevant cues and ignore taskirrelevant cues to guide behavior. The "central computation" hypothesis assumes that cognitive control relies on computations performed by the PFC, which then interacts with other brain regions to accomplish the task. Alternatively, the "local computation" hypothesis suggests that computations necessary for cognitive control are carried out by other brain regions that have been shown to be essential for cognitive control tasks, such as the dorsal hippocampus and the thalamus. If the central computation hypothesis is correct, PFC lesions should disrupt cognitive control. Alternatively, if the local computation hypothesis is correct, cognitive control would be spared after PFC lesions. The task used to assess cognitive control is the active place avoidance task in which rats must avoid a section of a rotating arena using the stationary room cues and ignoring the local olfactory cues on the rotating platform. Performance on this task has previously been shown to be disrupted by hippocampal lesions and hippocampal ensembles dynamically represent the room and arena depending on the animal's proximity to the shock zone. They found no group (lesion vs. sham) differences in the three behavioral parameters tested: distance traveled, latency to enter the shock zone, and number of shock zone entries for both the standard task and the "conflict" task in which the shock zone was rotated by 180 degrees. The only significant difference was the savings index; the lesion group entered the new shock zone more often than the sham group during the first 5 minutes of the second conflict session. This deficit was interpreted as a cognitive flexibility deficit rather than a cognitive control failure. Next, the authors compared cytochrome oxidase activity between sham and lesion groups in 14 brain regions and found that only the amygdala showed significant elevation in the lesion vs. sham group. Pairwise correlation analysis revealed a striking difference between groups, with many correlations between regions lost in the lesion group (between reuniens and hippocampus, reuniens and amygdala and a correlation between dorsal CA1 and central amygdala that appeared in the lesion group and were absent in the sham group. Finally, the authors assessed dorsal hippocampal representations of the spatial frame (arena vs. room) and found no differences between lesion and sham groups. The only difference in hippocampal activity was reduced overdispersion in the lesion group compared to the sham group on the pretraining session only and this difference disappeared after the task began. Collectively, the authors interpret their findings as supporting the local computation hypothesis; computations necessary for cognitive control occur in brain regions other than the PFC. 

      Strengths:

      (1) The data were collected in a rigorous way with experimental blinding and appropriate statistical analyses. 

      (2) Multiple approaches were used to assess differences between lesion and sham groups, including behavior, metabolic activity in multiple brain regions, and hippocampal singleunit recording. 

      Weaknesses:

      (1) Only male rats were used with no justification provided for excluding females from the sample.

      This is a weakness we acknowledge. The experiments were performed at a time when we did not have female rats in the lab.

      (2) The conceptual framework used to interpret the findings was to present two competing hypotheses with mutually exclusive predictions about the impact of PFC lesions on cognitive control. The authors then use mainly null findings as evidence in support of the local computation hypothesis. They acknowledge that some people may question the notion that the active place avoidance task indeed requires cognitive control, but then call the argument "circular" because PFC has to be involved in cognitive control. This assertion does not address the possibility that the active place avoidance task simply does not require cognitive control. 

      We beg to differ that the possibility was not addressed. Prior to making the assertion, the manuscript describes the evidence that the active place avoidance task requires cognitive control. The evidence is multifold, and includes task design, behavior, and electrophysiology; we argue that this is more evidence than has been provided for other tasks that are asserted to require cognitive control. Specifically line 417 states:

      “We have previously demonstrated cognitive control in the active place avoidance task variant we used (Fig. 1) because the rats must ignore local rotating place cues to avoid the stationary shock zone. Even when the arena does not rotate, rats distinctly learn to avoid the location of shock according to distal visual room cues and local olfactory arena cues, such that the distinct place memories can be independently manipulated using probe trials [49, 50]. When the arena rotates as in the present studies, neural manipulations that impair the place avoidance are no longer impairing when the irrelevant arena cues are hidden by shallow water [14, 15, 51, 52]. Furthermore, persistent hippocampal neural circuit changes caused by active place avoidance training are not detected when shallow water hides the irrelevant arena cues to reduce the cognitive control demand [10, 31, 33]. While these findings unequivocally demonstrate the salience of relevant stationary room cues to use for avoiding shock and irrelevant arena cues to ignore during active place avoidance, the most compelling evidence of cognitive control comes from recording hippocampal ensemble discharge. Hippocampal ensemble discharge purposefully represents current position using stationary room information when the subject is close to the stationary shock zone and alternatively represents rotating arena information when the mouse is far from the stationary shock zone [Fig. 4; 10].”

      Line 436, however, acknowledges a fact that will always be true: no matter what anyone opines - until there are universally agreed upon objective criteria, it is logically possible that active place avoidance does not require cognitive control. The revision states: Despite this evidence from task design, behavioral observations, and direct electrophysiological representational switching as required to directly demonstrate cognitive control, one might still argue that it is logically possible that the active place avoidance task does not require cognitive control and this is why the mPFC lesion did not impair place avoidance of the initial shock zone. We consider such reasoning to be unproductive because it presumes that only tasks that require an intact mPFC can be cognitive control tasks. We nonetheless acknowledge that for some, we have not provided sufficient evidence that the active place avoidance requires cognitive control.

      “We assert the evidence is compelling, and together these findings require rejecting the central-computation hypothesis that the mPFC is essential for the neural computations that are necessary for all cognitive control tasks.”

      (3) The authors did not link the CO activity with the behavioral parameters even though the CO imaging was done on a subset of the animals that ran the behavioral task nor did they make any attempt to interpret these findings in light of the two competing hypotheses posed in the introduction. Moreover, the discussion lacks any mechanistic interpretations of the findings. For example, there are no attempts to explain why amygdala activity and its correlation with dCA1 activity might be higher in the PFC lesioned group. 

      The CO study was performed to assess the effects of the lesion, as stated on line 262 “Cytochrome oxidase (CO), a sensitive metabolic marker for neuronal function [27], was used to evaluate whether lesion effects were restricted to the mPFC.” Furthermore, as a matter of fact, line 411 states “Thus, CO imaging and electrophysiological evidence identify changes in the brain beyond the directly damaged mPFC area. In particular, the dorsal hippocampus loses the inhibitory input from mPFC [45, 46] and loses the metabolic correlation with the nucleus reuniens, which is thought to be a relay between the mPFC and the dorsal hippocampus [47, 48].”

      These CO measures assess baseline metabolic function and so it would be inappropriate to correlate them with the measures of behavior. Because the lesion and control groups do not differ on most measures of behavior, a relationship to CO measures is not expected. Importantly, even if there were differences in correlations between CO activity and behavioral measures, what could they mean? The study was designed to distinguish between two hypotheses, not to determine what CO differences could mean for behavior. As such, it is not at all clear how metabolic consequences of the lesion relate to the two hypotheses being evaluated, and so we consider it inappropriate to speculate. We did examine, and now include, the correlation between lesion size and conflict behavior. The Fig. 1 legend states “Savings was not related to lesion size r = 0.009, p = 0.98. *p < 0.05.”

      (4) Publishing null results is important to avoid wasting animals, time, and money. This study's results will have a significant impact on how the field views the role of the PFC in cognitive control. Whether or not some people reject the notion that the active place avoidance task measures cognitive control, the findings are solid and can serve as a starting point for generating hypotheses about how brain networks change when deprived of PFC input. 

      We thank the reviewer for the acknowledgement.

      Reviewer #3 (Public review): 

      Summary:

      This study by Park and colleagues investigated how the medial prefrontal cortex (mPFC) influences behavior and hippocampal place cell activity during a two-frame active place avoidance task in rats. Rats learned to avoid the location of mild shock within a rotating arena, with the shock zone being defined relative to distal cues in the room. Permanent chemical lesions of the mPFC did not impair the ability to avoid the shock zone by using distal cues and ignoring proximal cues in the arena. In parallel, hippocampal place cells alternated between two spatial tuning patterns, one anchored to the distal cues and the other to the proximal cues, and this alteration was not affected by the mPFC lesion. Based on these findings, the authors argue that the mPFC is not essential for differentiating between task-relevant and irrelevant information. 

      Strengths:

      This study was built on substantial work by the Fenton lab that validated their two-frame active place avoidance task and provided sound theoretical and analytical foundations. Additionally, the effectiveness of mPFC lesions was validated by several measures, enabling the authors to base their argument on the lack of lesion effects on behavior and place cell dynamics. 

      Weaknesses:

      The authors define cognitive control as "the ability to judiciously use task-relevant information while ignoring salient concurrent information that is currently irrelevant for the task." (Lines 77-78). This definition is much simpler than the one by Miller and Cohen: "the ability to orchestrate thought and action in accordance with internal goals (Ref. 1)" and by Robbins: "processes necessary for optimal scheduling of complex sequence of behaviour." (Dalley et al., 2004, PMID: 15555683). Differentiating between task-relevant and irrelevant information is required in various behavioral tasks, such as differential learning, reversal learning, and set-shifting tasks. Previous rodent behavioral studies have shown that the integrity of the mPFC is necessary for set-shifting but not for differential or reversal learning (e.g., Enomoto et al., 2011, PMID: 21146155; Cho et al., 2015, PMID: 25754826). In the present task design, the initial training is a form of differential learning between proximal and distal cues, and the conflict training is akin to reversal learning. Therefore, the lack of lesion effects is somewhat expected. It would be interesting to test whether mPFC lesions impair set-shifting in their paradigm (e.g., the shock zone initially defined by distal cues and later by proximal cues). If the mPFC lesions do not impair this ability and associated hippocampal place dynamics, it will provide strong support for the authors' local computation hypothesis.

      Thank you for these comments. In addressing them we have provided a significant revision to the manuscript’s Introduction. While authors like those cited by the reviewer have defined cognitive control, those definitions are difficult to test rigorously, as it is almost a matter of opinion whether a subject is displaying “the ability to orchestrate thought and action in accordance with internal goals" or whether they are using "processes necessary for optimal scheduling of complex sequence of behaviour." What would such definitions of cognitive control predict about neuronal activity? We have deliberately used a simple, operational definition of cognitive control because it is physiologically testable. In the revision, starting at line 93, we have provided an excerpt from Miller and Cohen (2001) with discussion. The importance of that work is that it provides explicit neuronal criteria and a means to operationally define cognitive control. As stated on Line 118 “Accordingly, cognitive control would be at work when there is sustained neuronal network representations of task-relevant information that suppresses or gates representations of salient task-irrelevant information in accord with purposeful judicious behavior.”

      We used a R+A- task variant in which there is a stationary room-frame shock zone and task irrelevant arena-frame information. A strict correspondence to shift-shifting task design cannot be accomplished with active place avoidance because an A+R- task that requires avoiding an arena-frame shock zone in the absence of a room-frame shock zone can be accomplished trivially if the subject chooses to not move when it is in a place with no shock. However, the R+A+ task variant is readily learned, in which there is both a room-frame and an arena-frame shock zone (see cited work below). This task variant requires the subject to judiciously shift between avoiding the room-frame shock zone using stationary room information and avoiding the arena-frame shock zone using rotating arena information. This R+A+ task variant might meet the reviewer’s criteria for cognitive control. We have recorded hippocampal and entorhinal ensemble activity during the R+A+ task variant and it is very similar to the activity during the R+A- task we used. Nonetheless, future work will investigate the efect of mPFC lesion on the R+A+ task variant.

      Cited work:

      Fenton AA, Wesierska M, Kaminsky Y, Bures J (1998), Both here and there: simultaneous expression of autonomous spatial memories in rats. Proc Natl Acad Sci U S A 95:11493-11498. Kelemen E, Fenton AA (2010), Dynamic grouping of hippocampal neural activity during cognitive control of two spatial frames. PLoS Biol 8:e1000403.

      Burghardt NS, Park EH, Hen R, Fenton AA (2012), Adult-born hippocampal neurons promote cognitive flexibility in mice. Hippocampus 22:1795-1808.

      Park EH, Keeley S, Savin C, Ranck JB, Jr., Fenton AA (2019), How the Internally Organized Direction Sense Is Used to Navigate. Neuron 101:1-9.

      Recommendations for the authors:  

      Reviewer #1 (Recommendations for the authors): 

      (1) Incorporate the cytochrome oxidase and hippocampal recordings (rationale and hypothesis) in the introduction, explaining how these aspects are relevant to the general question. 

      We have done this as requested. See lines 159-173 of the revised introduction.

      (2) Figure 1C. On Day 4-5 (conflict training) in which the shock zone was relocated 180 deg from the initial location, the behavioral tracks did not show any presence of the rat in this sector (in particular for the lesion example). Figure 4 nevertheless indicates that entrances have been made (which was expected since rats have to know that the shock zone was relocated).

      Thanks for pointing this out. The tracks are from the end of the sessions. The labels have been changed to specify which trials the tracks are from.

      (3) Figure 1C. The caption is huge as it contains the statistical analyses details. I would prefer to have these details in the text and keep the caption at a "reasonable" length. At the end of the caption (l. 190-191), it would be less confusing the keep the numbering of the training days: replace D1T1 with D2T1 and D2T9 with D3T9).

      The statistical details have been relocated to the main text and the numbering updated, as suggested, thank you.

      (4) It was not inconsiderable to show that mPFC lesion had some effects in the present task if it were only to validate the effectiveness of the lesion. This brain area has been shown to be important for planning, cognitive flexibility, etc. Indeed the authors found that the saving index was greater in sham than in mPFC rats (overdispersion in hippocampal firing was also reduced in pretraining) and interpreted this result as impaired flexibility. Would an alternative explanation be a memory deficit? I nevertheless expected that impaired flexibility in mPFC rats would be expressed in conflict trials in the form of more entrances in the zone that was initially not associated with shock (at least in the first trials of Day 4). But it appears to not be the case.

      A memory deficit is unlikely to explain the difference between the groups on the first trial of Day 5. Memory in the lesion rats was tested multiple times, specifically at the start of each trial (time to first entrance), including on the 24-h retention test, and no deficits were observed. Performance on Day 9 trial 1 is worse in the lesion group than in the controls, but it is not parsimonious to attribute this to a simple memory deficit since 24-h memory was good and similar between lesion and control rats on days 3 and 4, and memory on Day 5 was equally poor in both the lesion and control rats, as measured by time to first entrance.  

      (5) Material and methods. The injected volume of ibotenic acid should be mentioned. 

      The volume 0.2 µl was added. See line 531.

      (6) The rationale for doing the conflict training session should be indicated somewhere. 

      The rationale was provided. See lines 204-208.

      Reviewer #2 (Recommendations for the authors): 

      (1) Line 132: The text states that all sham rats improved and only 6/10 lesion rats improved is followed by a t-test, which tests the difference between means; it does not compare proportions. Also, what criterion was used to determine if an improvement was seen or not? 

      The statistical comparison is provided (now lines 230: test of proportions z = 2.3, p = 0.03). Improvement was simply numerically fewer entrances.

      (2) Line 138: This is a very long and confusing sentence. Consider revising for clarity. 

      The sentence (now line 234) was revised.

      (3) Figure 1B only includes data from 3 animals. Most published studies show the whole dataset by presenting the largest and smallest lesions. 

      Supplemental Figure S2 was added with all the lesions depicted and quantified.

      (4) Figure 1C suggestion to make the schematic shock zone line up with the shock zone shown for the tracking data. 

      Graphically, it looks better as drawn as it uses to perspective to depict a three-dimensional structure.

      (5) Methods: Clarify if the shock zone location was the same across all rats. 

      Line 570 states that the shock zone was the same for all rats.

      (6) Line 158: "Behavioral tracks" is not clear. Suggest more precise wording.

      Reworded to “Tracked room-frame positions” (now line 249)

      (7) Line 166: "effect of trial" - should this be the main effect of trial?; "interaction" - should this be "group x trial" interaction? 

      Reworded (now line 181).

      (8) Line 167: "or their interaction" is awkward in the context of the sentence. 

      Reworded (now line 182).

      (9) Line 182: Avoid talking about "trends" as if they are almost significant unless the authors suspect that they did not have sufficient statistical power to detect differences. In that case, a power analysis should be provided. 

      Removed.

      (10) Line 190: "left:...right..." is hard to follow, especially with acronyms like D1T1. Consider revising for clarity. 

      Revised (now lines 246-248).

      (11) Line 195: "effectiveness of the PFC to impair" is unnecessarily verbose. 

      Reworded (now lines 255-257).

      (12) Savings results: There is a lot of variability in the lesion group. It would be interesting to know if the extent of the lesion correlates with savings.

      Savings was not related to lesion. See line 259.

      (13) Line 300: The thalamic recording results are not reported in the results section (other than appearing in the table). Moreover, there is no detail about which thalamic nucleus these recordings are from.

      Lines 411 and 614 provides these details.  

      (14) Line 312: "no longer impair" contains a grammatical error. 

      Corrected (now line 422)

      (15) Line 325: "was not impairing" contains a grammatical error. 

      Corrected (now line 437).

      (16) Line 327: The sentence ending with "...opinion of others" seems unnecessarily confrontational. 

      Previous reviewers at other journals have maintained this position, we therefore included such a strong statement in our initial submission. However, we now revised this statement to avoid appearing confrontational.

      (17) Line 329: Sentence is awkward. Consider revising. 

      Revised (now line 443).

      (18) Line 384: The authors should disclose if there was an objective metric for determining the adequacy of the lesion. 

      The lesion assessment and quantification is better explained in the Methods under “Cytochrome oxidase activity and Nissl staining,” (lines 708-714).

      (19) Line 385: The authors should clarify how they got from 15 rats (Line 376) to 10. 

      This information is provided in the methods.

      (20) Line 390: It is not clear why skin irritation in the cage mate would prevent the rat from being tested. 

      This has been explained in the Methods under “Behavioral analysis followed by cytochrome oxidase activity” (lines 515-518).

      (21) Methods section: The authors should describe how the tracking data were acquired. Overhead camera? Tracker based on luminance or body position? What software program was used? What was the sampling rate? 

      This is now better explained in the Methods under “Active place avoidance task) (lines 538551).

      (22) Methods section: Include how fast the arena was rotating and other details about the task such as where rats were placed during the ITI. 

      Better explained in the Methods under “Active place avoidance task”.

      (23) Line 439: The recording system used (hardware & software) should be stated. 

      This is now included in the Methods (line 538).

      (24) Line 435: Though overdispersion calculation is described thoroughly, there is nothing in the paper that tells me what overdispersion means. 

      What the measure means is now described in the Methods under “Electrophysiology data analysis” (lines 646-650).

      (25) Line 561: The test used to assess effect sizes should be stated. 

      Effect sizes corresponding to the statistical tests are provided.

      Reviewer #3 (Recommendations for the authors): 

      (1) At the end of the conflict training, rats with mPFC lesions learned to avoid the new shock zone (Figure 1F, Block 16), but their place cells did not show room-preferring activity near the shock zone (Figure 4B). This observation questions whether spatial frame-specific representation is relevant for active avoidance. Can the authors clarify this point?

      This is a dynamic behavior and the hippocampal dynamics match, changing with a dynamic that is a few seconds, as we have shown in several published papers. The lack of a preference averaged over 20 minutes when the rats are avoiding both the current and former shock zones during the conflict session is pretty much what would be expected from such a coarse measurement. The important measure is the spatially-resolved measure of room versus arena preference. Figure 4B shows that in the lesion rats there is less of a frame preference during conflict, generally (consistent with poorer flexibility). However, Figure 4D quantifies the frame preference near and far from the shock zone and accordingly, there is no difference between the groups.

      (2) Related to the point above, the author might consider including panels in Figures 4C and D to show the neural activity during the pretraining and conflict training retention period. I assume p(room) will be comparable between the Near and Far segment in both sessions, but the p(room) may be higher in the Conflict training session than the Pretraining session. This would show that the mPFC lesion impairs suppressing the place cell activity encoding the old shock location. 

      Thanks for the suggestion. While we don’t think we can draw any strong conclusions from this analysis we are fine to show it. The issue is that during conflict, the rats have two perfectly reasonable representations of where there was shock, the initial location that was turned off to make the conflict, and the most recent conflict location of shock. Importantly, these recordings are during conflict retention after we turned off the shock for the retention recording (for the second time in the rat’s experience). Turning off the shock allows us to exactly match the physical conditions of pretraining, initial retention and conflict retention, which was the experimental design’s goal. However, the experiential history of the rats prior to initial retention and conflict retention cannot match, because during initial retention the rats had never experienced a changed shock zone whereas, by conflict retention, they had experienced multiple changes. Importantly, we have previously shown that mouse hippocampal ensembles represent both initial and conflict shock locations, as the animals consider their options during conflict trials (see Dvorak et al 2018, PLoS Biol 16:e2003354). Consequently, we cannot make any strong predictions about whether or not hippocampal activity during conflict retention should be room-frame preferring selectively in the vicinity of the current shock zone. As I am sure the reviewer appreciates from their own introspection, mental representations are mercifully not obliged to dictate behavior. In fact, that is what is interesting and controversial about cognitive control – it is a dynamic internal process and the innovation of our work lies in demonstrating that one cannot only rely on behavior to assess this process. Nonetheless, we did this analysis and now present it in the revised Fig. 4. During pretraining both lesion and sham groups express no particular spatially-modulated preference for either the room or the arena frame, as expected. During initial training both groups express a room-frame preference in the vicinity of the shock zone, as we initially reported. By inspection, during conflict, the sham rats express a preference for room-frame activity in the vicinity of the most recent shock zone location; this preference is weaker than what is expressed during initial retention. The lesion rats do not show this preference. These impressions are quantified in revised Fig. 4D; the comparisons within the conflict retention sessions did not reach statistical significance. We leave it to the reader to interpret what that means. Thanks for the nudge.

      (3) The significant group difference in place cell overdispersion during the pretraining phase (Figure 3C) is interesting, but some readers would appreciate additional sentences on its functional implication. Does it mean the spatial tuning of place cells was disrupted by the mPFC lesion?

      Only the reliability of spatial firing was altered, not the spatial tuning.

      (4) Although the method section described how to calculate overdispersion and SFEP, some concise, intuitive descriptions of these measures in the result section would help readers understand these results.

      Overdispersion is better explained. See lines 646-650.

      (5) I recommend adding a figure of the task performance of the rats used in the electrophysiological recording experiment and a table summarizing the number of cells recorded per animal. 

      We have included Table S2 with the cell counts and a summary of the performance for each of the rat in the electrophysiological recording experiment.

      (6) Readers would appreciate additional information on task apparatus, such as the size, appearance, and rotating speed of the arena, as well as stationary cues available in the room. 

      This is now provided in the Methods under “Active place avoidance task”.

      (7) Lines 425-416: "On the fourth day of the behavioral training, the rats had a single trial with the shock on to test retention of the training." Shouldn't it be "shock off"? 

      No the shock was on to prevent extinction learning and to increase the challenge for conflict learning.

    1. eLife Assessment

      The authors provide a useful summary of ten years of Brain Initiative funding including the historical development, the specific funding mechanisms, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed. The evidence for impact is incomplete due to the omission of a comparison group of funded grants.

    2. Reviewer #1 (Public review):

      Summary:

      This is a convincing description of approximately ten years of funding from the NIH BRAIN initiative. It is of particular value at this moment in history, given the cataclysmic changes in the US government structure and function occurring in early 2025.

      Strengths:

      The paper contains a fair bit of documentation so that the curious reader can actually parse what this BRAIN program funded.

      Weaknesses:

      There are too many acronyms, and the manuscript reads as if it were an internal NIH document, where the audience knows all of the NIH nomenclature and program details. It is not particularly friendly to the outside, lay reader.

    3. Reviewer #2 (Public review):

      Summary:

      The authors provide an important summary of ten years of Brain Initiative funding including a description of the historical development of the initiative, the specific funding mechanisms utilized, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed.

      Strengths:

      This is a useful perspective on an important funding initiative over a ten-year period. It is clearly written and the illustrations and analyses are mostly useful for understanding the impact of the initiative.

      Weaknesses:

      The major limitation is that the bibliographic analysis does not provide a comparison group of funded grants. Because work that successfully competes for funding is likely to be more impactful than all work in a given area, the normalization of citations to field medians may reflect this "grant review" effect, rather than anything special about the Brain Initiative. Hopefully, this speculation is incorrect (I would guess that it is), but it would be helpful to try to demonstrate this more directly by including a funded comparison group.

      There are also minor inconsistencies in the numbering of the figures that need to be cleared up.

    4. Author response:

      eLife Assessment

      The authors provide a useful summary of ten years of Brain Initiative funding including the historical development, the specific funding mechanisms, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed. The evidence for impact is incomplete due to the omission of a comparison group of funded grants.

      In this combined version, we include a comparison group of non-BRAIN Initiative R01s derived from the parent notice of funding opportunity from FY2014-2022. We performed a bibliometric analysis of the publications, citations, RCR and budget productivity measure of the non-BRAIN parent R01. 

      Public Reviews:

      Reviewer #1 (Public review):

      Summary:

      This is a convincing description of approximately ten years of funding from the NIH BRAIN initiative. It is of particular value at this moment in history, given the cataclysmic changes in the US government structure and function occurring in early 2025.

      Strengths:

      The paper contains a fair bit of documentation so that the curious reader can actually parse what this BRAIN program funded.

      Weaknesses:

      There are too many acronyms, and the manuscript reads as if it were an internal NIH document, where the audience knows all of the NIH nomenclature and program details. It is not particularly friendly to the outside, lay reader.

      In this version, we have attempted to minimize acronyms and explain NIH nomenclature and program details to make it more accessible to readers not familiar with NIH terminology.

      Reviewer #2 (Public review):

      Summary:

      The authors provide an important summary of ten years of Brain Initiative funding including a description of the historical development of the initiative, the specific funding mechanisms utilized, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed.

      Strengths:

      This is a useful perspective on an important funding initiative over a ten-year period. It is clearly written and the illustrations and analyses are mostly useful for understanding the impact of the initiative.

      Weaknesses:

      The major limitation is that the bibliographic analysis does not provide a comparison group of funded grants. Because work that successfully competes for funding is likely to be more impactful than all work in a given area, the normalization of citations to field medians may reflect this "grant review" effect, rather than anything special about the Brain Initiative. Hopefully, this speculation is incorrect (I would guess that it is), but it would be helpful to try to demonstrate this more directly by including a funded comparison group.

      In this version, we have provided a comparison group of parent R01s that are not funded through the BRAIN Initiative from FY2014-2022 in Figure 3. We include publication metrics and budget efficiency measures for this comparison group.  

      There are also minor inconsistencies in the numbering of the figures that need to be cleared up.

      We have updated the figure numbers.

    1. eLife Assessment

      The manuscript presents some useful accounts of experiences funding team projects within the BRAIN Initiative. These would be more appropriate to add to the companion manuscript since the present manuscript contains some overlapping analyses and does not stand well on its own. Therefore the evidence supporting the conclusions is incomplete.

    2. Reviewer #1 (Public review):

      Summary:

      In this useful narrative, the authors attempt to capture their experience of the success of team projects for the scientific community.

      Strengths:

      The authors are able to draw on a wealth of real-life experience reviewing, funding, and administering large team projects, and assessing how well they achieve their goals.

      Weaknesses:

      The utility of the RCR as a measure is questionable. I am not sure if this really makes the case for the success of these projects. The conclusions do not depend on Figure 1.

    3. Reviewer #2 (Public review):

      Summary:

      The authors review the history of the team projects within the Brain initiative and analyze their success in progression to additional rounds of funding and their bibliographic impact.

      Strengths:

      The history of the team projects and the fact that many had renewed funding and produced impactful papers is well documented.

      Weaknesses:

      The core bibliographic and funding impact results have largely been reported in the companion manuscript and so represent "double dipping" I presume the slight disagreement in the number of grants (by one) represents a single grant that was not deemed to address systems/computational neuroscience. The single figure is relatively uninformative. The domains of study are sufficiently large and overlapping that there seems to be little information gained from the graphic and the Sankey plot could be simply summarized by rates of competing success.

    4. Author response:

      eLife Assessment 

      The manuscript presents some useful accounts of experiences funding team projects within the BRAIN Initiative. These would be more appropriate to add to the companion manuscript since the present manuscript contains some overlapping analyses and does not stand well on its own. Therefore the evidence supporting the conclusions is incomplete. 

      We appreciate the feedback on merging both manuscripts into one and have followed the advice in this version. 

      Public Reviews: 

      Reviewer #1 (Public review): 

      Summary: 

      In this useful narrative, the authors attempt to capture their experience of the success of team projects for the scientific community.  

      Strengths: 

      The authors are able to draw on a wealth of real-life experience reviewing, funding, and administering large team projects, and assessing how well they achieve their goals. 

      Weaknesses: 

      The utility of the RCR as a measure is questionable. I am not sure if this really makes the case for the success of these projects. The conclusions do not depend on Figure 1. 

      We respectfully disagree about the utility of the RCR, particularly because it is metric that is normalized by both year and topical area. We have added a more detailed description of how the RCR is calculated on page 6-7. Please note that figure 1 is aimed to highlight the funding opportunities, investments and number of awards associated with small lab (exploratory) versus team (elaborated, mature) research rather than a description of publication metrics.  

      Reviewer #2 (Public review): 

      Summary: 

      The authors review the history of the team projects within the Brain initiative and analyze their success in progression to additional rounds of funding and their bibliographic impact. 

      Strengths: 

      The history of the team projects and the fact that many had renewed funding and produced impactful papers is well documented. 

      Weaknesses: 

      The core bibliographic and funding impact results have largely been reported in the companion manuscript and so represent "double dipping" I presume the slight disagreement in the number of grants (by one) represents a single grant that was not deemed to address systems/computational neuroscience. The single figure is relatively uninformative. The domains of study are sufficiently large and overlapping that there seems to be little information gained from the graphic and the Sankey plot could be simply summarized by rates of competing success. 

      While we sincerely appreciate the feedback, we chose to retain these plots on domains and models to provide a sense of the broad spectrum of research topics contained in our TeamBCP awards. Further details on the awards can be derived from the award links provided in the text. Additionally, we retained the Sankey plots because these are a visual depiction of how awards transition from one mechanism to another, evolve in their funding sources, and advance in their research trajectories. The plot is an example of our continuity analysis which is only reported in the text and not visually shown for the remaining BCP programs.

    1. eLife Assessment

      This important computational study investigates homeostatic plasticity mechanisms that neurons may employ to achieve and maintain stable target activity patterns. The work extends previous analyses of calcium-dependent homeostatic mechanisms based on ion channel density by considering activity-dependent shifts in channel activation and inactivation properties that operate on faster and potentially variable timescales. The model simulations demonstrate the potential functional importance of these mechanisms, but the evidence is incomplete and would be strengthened by more in-depth analyses and explicit exposition.

    2. Reviewer #1 (Public review):

      This computational study builds on a previous study (Liu et al) from the Marder lab from 1998, where a model was proposed that demonstrated activity-dependent homeostatic recovery of activity in individual bursting neurons, based on three "sensors" of intrinsic calcium concentration. The original model modified levels of ion channel conductances. The current model builds on that and adds activity-dependent modifications of the voltage-dependence of these ionic currents, implemented to happen concurrently with maximum conductance levels, but at a different timescale. The faster timescale change in voltage dependence is justified by the assumption that such changes can occur by neuromodulatory chemicals or similar second messenger-based mechanisms that presumably act at a faster rate than the regulation of channel densities. The main finding is that the difference in timescales between the two homeostatic mechanisms (channel density vs. voltage dependence) could result in distinct subsets of parameters, depending on how fast the second messenger mechanisms operate.

      This study is an interesting and noteworthy extension of the theoretical ideas proposed by the classic study of Liu et al, 1998. It addresses a very important question: How do two known mechanisms of modifications of neuronal activity that occur at different timescales interact within an activity-dependent homeostatic framework? However, the study and its presentation have some major shortcomings that should be addressed to strengthen the claim.

      Major comments:

      (1) The main issue that I have with this study is the lack of exploration of "why" the model produces the results it does. Considering this is a model, it should be possible to find out why the three timescales of half-act/inact parameter modifications lead to different sets of results. Without this, it is simply an exploratory exercise. (The model does this, but we do not know the mechanism.) Perhaps this is enough as an interesting finding, but it remains unconvincing and (clearly) does not have the impact of describing a potential mechanism that could be potentially explored experimentally.

      (2) A related issue is the use of bootstrapping to do statistics for a family of models, especially when the question is in fact the width of the distribution of output attributes. I don't buy this. One can run enough models to find say N number of models within a tight range (say 2% cycle period) and the same N number within a loose range (say 20%) and compare the statistics within the two groups with the same N.

      (3) The third issue is that many of the results that are presented (but not the main one) are completely expected. If one starts with gmax values that would never work (say all of them 0), then it doesn't matter how much one moves the act/inact curves one probably won't get the desired activity. Alternately, if one starts with gmax values that are known to work and randomizes the act/inact midpoints, then the expectation would be that it converges to something that works. This is Figure 1 B and C, no surprise. But it should work the other way around too. If one starts with random act/inact curves that would never work and fixes those, then why would one expect any set of gmax values would produce the desired response? I can easily imagine setting the half-act/inact values to values that never produce any activity with any gmax.

      (4) A potential response to my previous criticism would be that you put reasonable constraints on gmax's or half-act/inact values or tie the half-act to half-inact. But that is simply arbitrary ad hoc decisions made to make the model work, much like the L8-norm used to amplify some errors. There is absolutely no reason to believe this is tied to the biology of the system.

      (5) The discussion of this manuscript is at once too long and not adequate. It goes into excruciating detail about things that are simply not explored in this study, such as phosphorylation mechanisms, justification of model assumptions of how these alterations occur, or even the biological relevance. (The whole model is an oversimplification - lack of anatomical structure, three calcium sensors, arbitrary assumptions, and how parameter bounds are implemented.) Lengthy justifications for why channel density & half-act/inact of all currents are obeying the same time constant are answering a question that no one asked. It is a simplified model to make an important point. The authors should make these parts concise and to the point. More importantly, the authors should discuss the mechanism through which these differences may arise. Even if it is not clear, they should speculate.

      (6) There should be some justification or discussion of the arbitrary assumptions made in the model/methods. I understand some of this is to resolve issues that had come up in previous iterations of this approach and in fact the Alonso et al, 2023 paper was mainly to deal with these issues. However, some level of explanation is needed, especially when assumptions are made simply because of the intuition of the modeler rather than the existence of a biological constraint or any other objective measure.

    3. Reviewer #2 (Public review):

      Summary:

      In this study, Mondal and co-authors present the development of a computational model of homeostatic plasticity incorporating activity-dependent regulation of gating properties (activation, inactivation) of ion channels. The authors show that, similar to what has been observed for activity-dependent regulation of ion channel conductances, implementing activity-dependent regulation of voltage sensitivity participates in the achievement of a target phenotype (bursting or spiking). The results however suggest that activity-dependent regulation of voltage sensitivity is not sufficient to allow this and needs to be associated with the regulation of ion channel conductances in order to reliably reach the target phenotype. Although the implementation of this biologically relevant phenomenon is undeniably relevant, the main conclusions of the paper and the insights brought by this computational work are difficult to grasp.

      Strengths:

      (1) Implementing activity-dependent regulation of gating properties of ion channels is biologically relevant.

      (2) The modeling work appears to be well performed and provides results that are consistent with previous work performed by the same group.

      Weaknesses:

      (1) The writing is rather confusing, and the state of the art explaining the need for the study is unclear.

      (2) The main outcomes and conclusions of the study are difficult to grasp. What is predicted or explained by this new version of homeostatic regulation of neuronal activity?

    4. Reviewer #3 (Public review):

      Mondal et al. use computational modeling to investigate how activity-dependent shifts in voltage-dependent (in)activation curves can complement activity-dependent changes in ion channel conductance to support homeostatic plasticity. While changes in the voltage-dependent properties of ion channels are known to modulate neuronal excitability, their role as a homeostatic plasticity mechanism interacting with channel conductance has been largely unexplored. The results presented here demonstrate that activity-dependent regulation of voltage-dependent properties can interact with plasticity in channel conductance to allow neurons to attain and maintain target activity patterns, in this case, intrinsic bursting. These results also show that the rate of channel voltage-dependent shifts can influence steady-state parameters reached as the model stabilizes into a stable intrinsic bursting state. That is, the rate of these modifications shapes the range of channel conductances and half-(in)activation parameters as well as activity characteristics such as burst period and duration. A major conclusion of the study is that altering the timescale of channel voltage dependence can seamlessly shift a neuron's activity characteristics, a mechanism that the authors argue may be employed by neurons to adapt to perturbations. While the study's conclusions are mostly well-supported, additional analyses, and simulations are needed.

      (1) A main conclusion of this study is that the speed at which (in)activation dynamics change determines the range of possible electrical patterns. The authors propose that neurons may dynamically regulate the timescale of these changes (a) to achieve alterations in electrical activity patterns, for example, to preserve the relative phase of neuronal firing in a rhythmic network, and (b) to adapt to perturbations. The results presented in Figure 4 clearly demonstrate that the timescale of (in)activation modifications impacts the range of activity patterns generated by the model as it transitions from an initial state of no activity to a final steady-state intrinsic burster. This may have important implications for neuronal development, as discussed by the authors.

      However, the authors also argue that the model neuron's dynamics - such as period, and burst duration, etc - could be dynamically modified by altering the timescale of (in)activation changes (Figure 6 and related text). The simulations presented here, however, do not test whether modifications in this timescale can shift the model's activity features once it reaches steady state. In fact, it is unlikely that this would be the case since, at steady-state, calcium targets are already satisfied. It is likely, however, as the authors suggest, that the rate at which (in)activation dynamics change may be important for neuronal adaptation to perturbations, such as changes in temperature or extracellular potassium. Yet, the results presented here do not examine how modifying this timescale influences the model's response to perturbations. Adding simulations to characterize how alterations in the rate of (in)activation dynamics affect the model's response to perturbations-such as transiently elevated extracellular potassium (Figure 5) - would strengthen this conclusion.

      (2) Another key argument in this study is that small, coordinated changes in channel (in)activation contribute to shaping neuronal activity patterns, but that, these subtle effects may be obscured when averaging across a population of neurons. This may be the case; however, the results presented don't clearly demonstrate this point. This point would be strengthened by identifying correlations, if they exist, between (in)activation curves, conductance, and the resulting bursting patterns of the models for the simulations presented in Figure 2 and Figure 4, for example. Alternatively, or additionally, relationships between (in)activation curves could be probed by perturbing individual (in)activation curves and quantifying how the other model parameters compensate, which could clearly illustrate this point.

    5. Author response:

      We thank the reviewers for their detailed and constructive comments on our manuscript entitled “Activity-Dependent Changes in Ion Channel Voltage-Dependence Influence the Activity Patterns Targeted by Neurons.” We appreciate the time and effort the reviewers invested in critiquing our work and are grateful for the opportunity to clarify and improve our manuscript.

      As noted by the reviewers, the main message of the manuscript is that the intrinsic properties and activity characteristics of targeted bursters depend on the timescale of half-(in)activation alterations in the homeostatic mechanism. However, the concerns of the reviewers reveal that the manuscript is organized in ways that detract from this message. Below we respond to the points the reviewers raise and close by outlining the changes that we will make to the manuscript as a result. Our goal will be to streamline the message of the paper while addressing the concerns of the reviewers.

      Response to Reviewer #1:

      Point 1: We interpret the reviewer’s question about “mechanism” to be: why do half-(in)activation alterations redirect degenerate bursters to different parameter regions? (A separate aspect of “mechanism,” namely how these alterations might be biologically implemented, is already addressed in the paper.)

      We speculate that Figure 3 illustrates this process. As conductance densities slowly evolve, rapid half-(in)activation changes cause the sensor variable (α) to jump abruptly as it searches for a voltage-dependence configuration that meets calcium targets (Figure 3A). The channel densities are slightly altered and this process continues again. Slowing the half-(in)activations alterations reduces these abrupt fluctuations (Figure 3B). Making the alterations infinitely slow effectively removes half-(in)activation changes altogether, leaving the system reliant solely on slower alterations in maximal conductances (Figure 3C). Because each timescale of half-(in)activation produces a different channel repertoire at each time step, the neuron follows distinct trajectories through the space of activity characteristics and intrinsic properties over the long term.

      Point 2: We appreciate the reviewer’s skepticism regarding our statistical approach with the “Group of 5” and “Group of 20.” These groups arose from historical aspects of our analysis and this analysis does not directly advance the main point—that changes in the timescale of channel voltage-dependence alterations impact the properties of bursters to which the homeostatic mechanism converges. Therefore, we plan to remove the references to the Group of 5 and focus on how the Group of 20 responds to variations in the timescale of voltage-dependent alterations.

      Point 3: Our paper claims that the half-(in)activation mechanism is subordinate to the maximal conductance mechanism. We agree with the reviewer that making this claim requires more care. The simulations we run are controls in the spirit described below.

      The reviewer notes that in our simulations, half-(in)activations are already near the range required for bursting, which forces maximal conductances to undergo larger changes and thus appear more critical. We however note that the opposite can also occur: if half-(in)activation values were already positioned in ranges required for bursting, an arrangement of small maximal conductances may potentially produce bursting. The latter might give the impression that maximal conductance alterations and half-(in)activation alterations are equally important. The simulations we ran are simply suggested this wasn’t true for these models.

      Points 4 - 6: In Point 4, the reviewer highlights model choices (e.g., constraints on maximal conductance and half-(in)activation, use of the L8 norm) are not clearly justified. In Point 5, the reviewer suggests that the paper provides excessive detail about other model choices. Point 6 appears to reiterate concerns about insufficient justification for some modeling decisions.

      Our intent was to acknowledge every caveat, which led us to include long section on Model Assumptions in the Discussion. However, as Point 5 notes, this makes the Discussion cumbersome. The Discussion should focus on remarks regarding the impact that timescale of half-(in)activation alterations have on the family of bursters targeted by the homeostatic mechanism. Consequently, we will relocate the extended discussion of model assumptions from the Discussion to the Methods section. This section already touches on how the constraints on half-(in)activation alterations compare to earlier versions of the model (noted in Point 6) and will be expanded to further explain our choice of the L8 norm (Point 4).

      Response to Reviewer #2:

      Weakness 1: The reviewer notes that the writing is “rather confusing.” This likely arises from the fact that we did not consistently emphasize the core message: the timescale of half-(in)activation alterations influences the intrinsic properties and activity characteristics of bursters targeted by the homeostatic mechanism. We will address this by reorganizing the manuscript to make that focus clearer, and we outline these planned revisions at the end of these responses.

      The reviewer specifically points out that the state-of-the-art is not clearly articulated. We will reorganize the Introduction to highlight this. Briefly, work on activity-dependent homeostasis has historically focused on changes in channel density. This is supported by experiment and has been modelled theoretically. In comparison, changes in channel voltage-dependence, while documented, are less explored due to the challenges of measuring them. In this work, we attempt to study the impact that alterations in channel voltage-dependence have on activity-dependent homeostasis. To do this, we extend existing computational models of activity-dependent homeostasis—models that have hitherto only altered channel density—by incorporating a mechanism that also adjusts channel voltage-dependence.

      Weakness 2: The Discussion highlights two potential implications of our findings—one for neuronal development and another for activity recovery following perturbations. However, they were outlined after the Model Assumptions section which, as Reviewer 1 points out, is quite detailed and cumbersome.

      Another aspect that may contribute to the challenge in interpreting our results may be our conceptual approach to neuronal excitability, which relies on a computational model of activity-dependent homeostasis that abstracts much of the underlying biochemistry. Our message is general: the timescale of half-(in)activation alterations influences the intrinsic properties and activity characteristics of bursters targeted by a homeostatic mechanism. As such, the implications are general. Their value lies in circumscribing a conceptual framework from which experimentalists may devise and test new hypotheses. We do not aim to predict or explain any specific phenomenon in this work. To address this concern however, we will expand our discussion of how these findings may guide experimental considerations, particularly regarding neuronal development and activity recovery during perturbations, to better illustrate the practical utility of our results.

      Response to Reviewer #3:

      Point 1: This reviewer suggests that our core message—namely, that the timescale of half-(in)activation alterations affects the intrinsic properties and activity patterns targeted by a homeostatic mechanism—should also apply during perturbations. We plan to address this by extending our analysis on the Group of 20 models. We will perturb activity by increasing extracellular potassium concentration and change the timescale of half-(in)activation alterations during the perturbation. This should underscore how the neuron’s stabilized activity pattern depends on this timescale, reinforcing our central message.

      Point 2: In this part of the Discussion, we noted that multiple half-activation shifts collectively shape the neuron’s global properties, and that averaging might obscure these effects. However, in light of the reviewers’ comments, we recognize that this observation alone does not directly advance the paper’s main message. To make it relevant, we would need to (1) identify correlations between intrinsic parameters (i.e., half-(in)activation and maximal conductance) and the resulting activity patterns, and (2) examine how these correlations shift under different timescales half-(in)activation alterations. Since we have not performed that analysis, we will revise this part of the Discussion to clarify its connection to the paper’s principal focus by noting that a deeper exploration of this notion using correlations will be the topic of future work.

      Conclusion: We outline updates we will make to the paper here.

      Introduction: In response to Reviewer 2, we will provide a clearer explanation of the state-of-the-art in activity-dependent homeostasis and highlight our specific contribution. We will emphasize that our conclusions, while generic, are relevant in experimental contexts.

      Results: We will reorganize this section to underscore the main point: the timescale of half-(in)activation alterations affects the intrinsic properties and activity characteristics of bursters in the homeostatic mechanism. Figures 1 will remain as it is. It shows assembly from random initial conditions and explain that for these simulations we must always consider the half-(in)activation mechanism with a mechanism that alters maximal conductances as the half-(in)activation alterations alone cannot form bursters. Figure 2 will remain as is, but we will remove any discussion of the “Group of 5,” addressing Reviewer 1’s feedback. What is presently Figure 4 will then follow, illustrating how timescale differences shape the properties of 20 degenerate solutions. We then present Figure 3 to address Reviewer 1’s critique on mechanism. Here we will explain how different timescales of half-(in)activation alteration cause the homeostatic mechanism to update channel properties differently, leading to distinct trajectories through the space of intrinsic properties and activity characteristics (as described in the response of Point 1 of Reviewer 1’s feedback). Finally, following Point 1 of Reviewer 3, we will add a new figure highlighting the role of half-(in)activation timescale during perturbation.

      Discussion: To streamline the Discussion, the “Model Assumptions” section will be moved to Methods. In line with Point 2 of Reviewer 3, we will clarify how the concept of "small half-(in)activation shifts lead to global changes in neuronal properties" aligns with our core message. Additionally, following Reviewer 2’s comments, we will expand our discussion of implications by including how experimentalists might use our findings to inform studies on perturbations and development.

      Methods: We will expand “Model Assumptions” to explain in more detail why we chose the L8 norm.