eLife assessment

This study presents a valuable computational model for elaborating on the interpretation of chromosomal mosaicism in preimplantation embryos. The evidence supporting the claims of the authors is incomplete due to the assumption that is possible to quantify the cells in the embryo, oversimplification of mitotic errors, and the inclusion of the self-correction premise. The work will be of interest to embryologists, and geneticists working on reproductive medicine.

Reviewer #1 (Public Review):

Summary:

The manuscript presents a compelling model to explain the impact of mosaicism in preimplantation genetic testing for aneuploidies.

Strengths:

A new view of mosaicism is presented with a computational model, that brings new insights into an "old" debate in our field. It is a very well-written manuscript.

Weaknesses:

Although the manuscript is very well written, this is in a way that assumes that the reader has existing knowledge about specific terms and topics. This was apparent through a lack of definitions and minimal background/context to the aims and conclusions for some of the author's findings.

There is a need for some examples to connect real evidence and scenarios from clinical reports with the model.

Reviewer #2 (Public Review):

Summary:

Although an oversimplification of the biological complexities, this modeling work does add, in a limited way, to the current knowledge on the theoretical difficulties of detecting mosaicism in human blastocysts from a single trophectoderm biopsy in PGT. However, many of the premises that the modeling was built on are theoretical and based on unproven biological and clinical assumptions that could yet lead to be untrue. Therefore, the work should be considered only as a simplified model that could assist in further understanding of the complexities of preimplantation embryo mosaicism, but assumptions of real-world application are, at this stage, premature and should not be considered as evidence in favour of any clinical strategies.

Strengths:

The work has presented an intriguing theoretical model for elaborating on the interpretation of complex and still unclear biological phenomena such as chromosomal mosaicism in preimplantation embryos.

Weaknesses:

Lines 134-138: The spatial modeling of mitotic errors in the embryo was oversimplified in this manuscript. There is only limited (and non-comprehensive) evidence that meiotic errors leading to chromosome mosaicism arise from chromosome loss or gain only (e.g. anaphase lag). This work did not take into account the (more recognised) possibility of mitotic nondisjunction where following the event there would be clones of cells with either one more or one less of the same chromosome. Although addressed in the discussion (lines 572-574), not including this in the most basic of modeling is a significant oversight that, based on the simple likelihood, could significantly affect results.

General comment: the premise of the manuscript is that an embryologist (embryology laboratory) is aware of and can accurately quantify the number of cells in a blastocyst or TE biopsy. The reality is that it is not possible to accurately do this without the destruction of the sample which is obviously not clinically applicable. Based on many assumptions the findings show that taking small biopsies poorly classifies mosaic embryos, which is not disputed. However, extrapolating this to the clinic and making suggestions to biopsy a certain amount of cells (lines 539-540) is careless and potentially harmful by suggesting the introduction of potential change in clinical practice without validation. Additionally, no embryologist in the field can tell how many cells are present in a clinical TE biopsy, making this suggestion even more impractical.

On a more general clinical consideration, the authors should acknowledge that when reporting findings of unproven clinical utility and unknown predictive values this inevitably results in negative consequences for infertile couples undergoing IVF. It is proven and established that when couples face the decision on how to manage a putative mosaicism finding, the vast majority decide on embryo disposal. It was recently reported in an ESHRE survey that about 75% of practitioners in the field consider discarding or donating to research embryos with reported mosaicism. A prospective clinical trial showed that about 30% live birth rate reduction can be expected if mosaic embryos are not considered (Capalbo et al., AJHG 2021). The real-world experience is that when mosaicism is reported, embryos with almost normal reproductive potential are discarded. The authors should be more careful with the clinical interpretation and translation of these theoretical findings.

There is a robust consensus within the field of clinical genetics and genomics regarding the necessity to exclusively report findings that possess well-established clinical validity and utility. This consensus is grounded in the imperative to mitigate misinterpretation and ineffective actions in patient care. However, the clinical framework delineated in this manuscript diverges from the prevailing consensus in clinical genetics. Clinical genetics and genomics prioritize the dissemination of findings that have undergone rigorous validation processes and have demonstrated clear clinical relevance and utility. This emphasis is crucial for ensuring accurate diagnosis, prognosis, and therapeutic decision-making in patient care. By adhering to established standards of evidence and clinical utility, healthcare providers can minimize the potential for misinterpretation and inappropriate interventions. The framework proposed in this manuscript appears to deviate from the established principles guiding clinical genetics practice. It is imperative for clinical frameworks to align closely with the consensus guidelines and recommendations set forth by professional organizations and regulatory bodies in the field. This alignment not only upholds the integrity and reliability of genetic testing and interpretation but also safeguards patient well-being and clinical outcomes.

References:<br /> ACMG Board of Directors. (2015). Clinical utility of genetic and genomic services: a position statement of the American College of Medical Genetics and Genomics. Genetics in Medicine, 17(6), 505-507. https://doi.org/10.1038/gim.2014.194.<br /> Richards, S., Aziz, N., Bale, S., Bick, D., Das, S., Gastier-Foster, J., ... ACMG Laboratory Quality Assurance Committee. (2015). Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 17(5), 405-424. https://doi.org/10.1038/gim.2015.30

Line 61: "Self correction" - This terminology is unfortunately indiscriminately used in the field for PGT when referring to mosaicism and implies that the embryo can actively correct itself from a state of inherent abnormality. Apart from there being no evidence to suggest that there is an active process by which the embryo itself can correct chromosomal errors, most presumed euploid/aneuploid mosaic embryos will have been euploid zygotes and therefore "self-harm" may be a better explanation. True self-correction in the form of meiotic trisomy/monosomy rescue is of course theoretically possible but not at all clinically significant. The concept being conveyed in this part of the manuscript is not disputed but it is strongly suggested that the term "self correction" is not used in this context, nor in the rest of the manuscript, to prevent the perpetuation of misinformation in the field and instead use a better description.

Lines 69-73: The ability to quantify aneuploidy in known admixtures of aneuploid cells is indeed well established. However, the authors claim that the translation of this to embryo biopsy samples is inferred with some confidence and that if a biopsy shows an intermediate chromosome copy number (ICN), that the biopsy and the embryo are mosaic. There are no references provided here and indeed the only evidence in the literature relating to this is to the contrary. Multifocal biopsy studies have shown that an ICN result in a single biopsy is often not seen in other biopsies from the same embryo (Capalbo et al 2021; Kim et al., 2022; Girardi et al., 2023; Marin, Xu, and Treff 2021). Multifocal biopsies showing reciprocal gain and loss which would provide stronger validation for the presence of true mosaicism are also rare. In this work, the entire manuscript is based on the accuracy of ICN in a biopsy being reflective of mosaicism in the embryo. The evidence however points to a large proportion of ICN detected in embryo biopsy potentially being technical artifacts (misdiagnosing both constitutionally normal and abnormal (meiotic aneuploid) embryos as mosaic. Therefore, although results from the modelling provide insight into theoretical results, these can not be used to inform clinical decision-making at all.

Lines 87-89: The authors make the claim that emerging evidence is suggestive that the majority of embryos are mosaic to some degree. If in fact, mosaicism is the norm, the clinical importance may be limited.

Line 102-103: The statement that data shows that the live birth rate per ET is generally lower in mosaic embryos than euploid embryos is from retrospective cohort studies that suffer from significant selection bias. The authors have ignored non-selection study results (Capalbo et al, ajhg 2021) that suggest that putative mosaicism has limited predictive value when assessed prospectively and blinded.

Lines 94-98: The authors have misrepresented the works they have presented as evidence for biopsy result accuracy (Kim et al., 2023; Victor et al 2019; Capalbo et al., 2021; Girardi et al., 2023, and any others). These studies show that a mosaic biopsy is not representative of the whole embryo and can actually be from embryos where the remainder of the embryo shows no evidence of mosaicism. There is also a missing key reference of Capalbo et al, AJHG 2021, and Girardi et al., HR 2023 where multifocal biopsies were taken.

Lines 371-372: "Selecting the embryo with the lowest number of aneuploid cells in the biopsy for transfer is still the most sensible decision". Where is the evidence for this other than the modeling which is affected by oversimplification and unproven assumptions? Although the statement seems logical at face value, there is no concrete evidence that the proportion of aneuploid cells within a biopsy is valuable for clinical outcomes, especially when co-evaluated with other more relevant clinical information.

Lines 431-463: In this section, the authors discuss clinical outcome data from the transfer of putative mosaic embryos and make conclusions about the relationship between ICN level in biopsy and successful pregnancy outcomes. The retrospective and selective nature of the data used in forming the results has the potential to lead to incorrect conclusions when applied to prospective unselected data.

Reviewer #3 (Public Review):

Unfortunately, this study fails to incorporate the most important variable impacting the ability to predict mosaicism, the accuracy of the test. The fact is that most embryos diagnosed as mosaic are not mosaic. There may be 4 cases out of thousands and thousands of transfers where a confirmation was made. Mosaicism has become a category of diagnosis in which embryos with noisy NGS profiles are placed. With VeriSeq NGS it is not possible to routinely distinguish true mosaicism from noise. An analysis of NGS noise levels (MAPD) versus the rate of mosaics by clinic using the registry will likely demonstrate this is the case. Without accounting for the considerable inaccuracy of the method of testing the proposed modeling is meaningless.

Recent data using more accurate methods of identifying mosaicism indicate that the prevalence of true preimplantation embryonic mosaicism is only 2%, which is also consistent with findings made post-implantation. This model fails to account for the possibility that, because so few embryos are actually mosaic, there is actually no relevance to clinical care whatsoever. In fact, differences in clinical outcomes of embryos designated as mosaic could be entirely attributed to poor embryo quality resulting in noise levels that make NGS results fall into the "mosaic" category.

Additional comments:

Indeed, as more data emerges, it appears that the majority of embryos from both healthy and infertile couples are mosaic to some degree (Coticchio et al., 2021; Griffin et al., 2022).

This statement should be softened as all embryos will be considered mosaic when a method with a 10% false positive rate is applied to 10 more parts of the same embryo. The distinction between artifact and true mosaicism cannot be made with nearly all current methods of testing. When virtually no embryos display uniform aneuploidy in a rebiopsy study, there should be great concern over the accuracy of the testing used. The vast majority of aneuploidy is meiotic in origin.

Experimental data provides strong evidence that, for the most part, the biopsy result obtained accurately represents the chromosome constitution of the rest of the embryo (Kim 96 et al., 2022; Navratil et al., 2020; Victor et al., 2019).

This statement is incorrect given published systematic review of the literature indicates a 10% false positive rate based on rebiopsy results.

This shows that accurately classifying a mosaic embryo based on a single biopsy is not robust.

This is exactly why the practice of designating embryo mosaics with intermediate copy numbers should not exist.

eLife assessment

This useful work by Park attempts to use machine learning algorithms to identify correlates of different observed shedding patterns in two COVID-19 cohorts. The evidence supporting the study conclusions is incomplete due to the lack of uniformity in assays between the 2 cohorts, relevant metadata (previous infection/vaccination status, viral variant), early viral load data in the cohorts, and incomplete statistical analyses. With a strengthened analysis, the work may be of interest to virologists, clinicians, and public health scientists.

Reviewer #1 (Public Review):

Summary:

This study by Park and colleagues uses longitudinal saliva viral load data from two cohorts (one in the US and one in Japan from a clinical trial) in the pre-vaccine era to subset viral shedding kinetics and then use machine learning to attempt to identify clinical correlates of different shedding patterns. The stratification method identifies three separate shedding patterns discriminated by peak viral load, shedding duration, and clearance slope. The authors also assess micro-RNAs as potential biomarkers of severity but do not identify any clear relationships with viral kinetics.

Strengths:

The cohorts are well developed, the mathematical model appears to capture shedding kinetics fairly well, the clustering seems generally appropriate, and the machine learning analysis is a sensible, albeit exploratory approach. The micro-RNA analysis is interesting and novel.

Weaknesses:

The conclusions of the paper are somewhat supported by the data but there are certain limitations that are notable and make the study's findings of only limited relevance to current COVID-19 epidemiology and clinical conditions.

(1) The study only included previously uninfected, unvaccinated individuals without the omicron variant. It has been well documented that vaccination and prior infection both predict shorter duration shedding. Therefore, the study results are no longer relevant to current COVID-19 conditions. This is not at all the authors' fault but rather a difficult reality of much retrospective COVID research.

(2) The target cell model, which appears to fit the data fairly well, has clear mechanistic limitations. Specifically, if such a high proportion of cells were to get infected, then the disease would be extremely severe in all cases. The authors could specify that this model was selected for ease of use and to allow clustering, rather than to provide mechanistic insight. It would be useful to list the AIC scores of this model when compared to the model by Ke.

(3) Line 104: I don't follow why including both datasets would allow one model to work better than the other. This requires more explanation. I am also not convinced that non-linear mixed effects approaches can really be used to infer early model kinetics in individuals from one cohort by using late viral load kinetics in another (and vice versa). The approach seems better for making population-level estimates when there is such a high amount of missing data.

(4) Along these lines, the three clusters appear to show uniform expansion slopes whereas the NBA cohort, a much larger cohort that captured early and late viral loads in most individuals, shows substantial variability in viral expansion slopes. In Figure 2D: the upslope seems extraordinarily rapid relative to other cohorts. I calculate a viral doubling time of roughly 1.5 hours. It would be helpful to understand how reliable of an estimate this is and also how much variability was observed among individuals.

(5) A key issue is that a lack of heterogeneity in the cohort may be driving a lack of differences between the groups. Table 1 shows that Sp02 values and lab values that all look normal. All infections were mild. This may make identifying biomarkers quite challenging.

(6) Figure 3A: many of the clinical variables such as basophil count, Cl, and protein have very low pre-test probability of correlating with virologic outcome.

(7) A key omission appears to be micoRNA from pre and early-infection time points. It would be helpful to understand whether microRNA levels at least differed between the two collection timepoints and whether certain microRNAs are dynamic during infection.

(8) The discussion could use a more thorough description of how viral kinetics differ in saliva versus nasal swabs and how this work complements other modeling studies in the field.

(9) The most predictive potential variables of shedding heterogeneity which pertain to the innate and adaptive immune responses (virus-specific antibody and T cell levels) are not measured or modeled.

(10) I am curious whether the models infer different peak viral loads, duration, expansion, and clearance slopes between the 2 cohorts based on fitting to different infection stage data.

Reviewer #2 (Public Review):

Summary:

This study argues it has found that it has stratified viral kinetics for saliva specimens into three groups by the duration of "viral shedding"; the authors could not identify clinical data or microRNAs that correlate with these three groups.

Strengths:

The question of whether there is a stratification of viral kinetics is interesting.

Weaknesses:

The data underlying this work are not treated rigorously. The work in this manuscript is based on PCR data from two studies, with most of the data coming from a trial of nelfinavir (NFV) that showed no effect on the duration of SARS-CoV-2 PCR positivity. This study had no PCR data before symptom onset, and thus exclusively evaluated viral kinetics at or after peak viral loads. The second study is from the University of Illinois; this data set had sampling prior to infection, so has some ability to report the rate of "upswing." Problems in the analysis here include:

-- The PCR Ct data from each study is treated as equivalent and referred to as viral load, without any reports of calibration of platforms or across platforms. Can the authors provide calibration data and justify the direct comparison as well as the use of "viral load" rather than "Ct value"? Can the authors also explain on what basis they treat Ct values in the two studies as identical?

-- The limit of detection for the NFV PCR data was unclear, so the authors assumed it was the same as the University of Illinois study. This seems a big assumption, as PCR platforms can differ substantially. Could the authors do sensitivity analyses around this assumption?

-- The authors refer to PCR positivity as viral shedding, but it is viral RNA detection (very different from shedding live/culturable virus, as shown in the Ke et al. paper). I suggest updating the language throughout the manuscript to be precise on this point.

-- Eyeballing extended data in Figure 1, a number of the putative long-duration infections appear to be likely cases of viral RNA rebound (for examples, see S01-16 and S01-27). What happens if all the samples that look like rebound are reanalyzed to exclude the late PCR detectable time points that appear after negative PCRs?

-- There's no report of uncertainty in the model fits. Given the paucity of data for the upslope, there must be large uncertainty in the up-slope and likely in the peak, too, for the NFV data. This uncertainty is ignored in the subsequent analyses. This calls into question the efforts to stratify by the components of the viral kinetics. Could the authors please include analyses of uncertainty in their model fits and propagate this uncertainty through their analyses?

-- The clinical data are reported as a mean across the course of an infection; presumably vital signs and blood test results vary substantially, too, over this duration, so taking a mean without considering the timing of the tests or the dynamics of their results is perplexing. I'm not sure what to recommend here, as the timing and variation in the acquisition of these clinical data are not clear, and I do not have a strong understanding of the basis for the hypothesis the authors are testing.

It's unclear why microRNAs matter. It would be helpful if the authors could provide more support for their claims that (1) microRNAs play such a substantial role in determining the kinetics of other viruses and (2) they play such an important role in modulating COVID-19 that it's worth exploring the impact of microRNAs on SARS-CoV-2 kinetics. A link to a single review paper seems insufficient justification. What strong experimental evidence is there to support this line of research?

Reviewer #3 (Public Review):

The article presents a comprehensive study on the stratification of viral shedding patterns in saliva among COVID-19 patients. The authors analyze longitudinal viral load data from 144 mildly symptomatic patients using a mathematical model, identifying three distinct groups based on the duration of viral shedding. Despite analyzing a wide range of clinical data and micro-RNA expression levels, the study could not find significant predictors for the stratified shedding patterns, highlighting the complexity of SARS-CoV-2 dynamics in saliva. The research underscores the need for identifying biomarkers to improve public health interventions and acknowledges several limitations, including the lack of consideration of recent variants, the sparsity of information before symptom onset, and the focus on symptomatic infections.

The manuscript is well-written, with the potential for enhanced clarity in explaining statistical methodologies. This work could inform public health strategies and diagnostic testing approaches. However, there is a thorough development of new statistical analysis needed, with major revisions to address the following points:

(1) Patient characterization & selection: Patient immunological status at inclusion (and if it was accessible at the time of infection) may be the strongest predictor for viral shedding in saliva. The authors state that the patients were not previously infected by SARS-COV-2. Was Anti-N antibody testing performed? Were other humoral measurements performed or did everything rely on declaration? From Figure 1A, I do not understand the rationale for excluding asymptomatic patients. Moreover, the mechanistic model can handle patients with only three observations, why are they not included? Finally, the 54 patients without clinical data can be used for the viral dynamics fitting and then discarded for the descriptive analysis. Excluding them can create a bias. All the discarded patients can help the virus dynamics analysis as it is a population approach. Please clarify. In Table 1 the absence of sex covariate is surprising.

(2) Exact study timReviewer #3 (Public Review):eline for explanatory covariates: I understand the idea of finding « early predictors » of long-lasting viral shedding. I believe it is key and a great question. However, some samples (Figure 4A) seem to be taken at the end of the viral shedding. I am not sure it is really easier to micro-RNA saliva samples than a PCR. So I need to be better convinced of the impact of the possible findings. Generally, the timeline of explanatory covariate is not described in a satisfactory manner in the actual manuscript. Also, the evaluation and inclusion of the daily symptoms in the analysis are unclear to me.

(3) Early Trajectory Differentiation: The model struggles to differentiate between patients' viral load trajectories in the early phase, with overlapping slopes and indistinguishable viral load peaks observed in Figures 2B, 2C, and 2D. The question arises whether this issue stems from the data, the nature of Covid-19, or the model itself. The authors discuss the scarcity of pre-symptom data, primarily relying on Illinois patients who underwent testing before symptom onset. This contrasts earlier statements on pages 5-6 & 23, where they claim the data captures the full infection dynamics, suggesting sufficient early data for pre-symptom kinetics estimation. The authors need to provide detailed information on the number or timing of patient sample collections during each period.

(4) Conditioning on the future: Conditioning on the future in statistics refers to the problematic situation where an analysis inadvertently relies on information that would not have been available at the time decisions were made or data were collected. This seems to be the case when the authors create micro-RNA data (Figure 4A). First, when the sampling times are is something that needs to be clarified by the authors (for clinical outcomes as well). Second, proper causal inference relies on the assumption that the cause precedes the effect. This conditioning on the future may result in overestimating the model's accuracy. This happens because the model has been exposed to the outcome it's supposed to predict. This could question the - already weak - relation with mir-1846 level.

(5) Mathematical Model Choice Justification and Performance: The paper lacks mention of the practical identifiability of the model (especially for tau regarding the lack of early data information). Moreover, it is expected that the immune effector model will be more useful at the beginning of the infection (for which data are the more parsimonious). Please provide AIC for comparison, saying that they have "equal performance" is not enough. Can you provide at least in a point-by-point response the VPC & convergence assessments?

(6) Selected features of viral shedding: I wonder to what extent the viral shedding area under the curve (AUC) and normalized AUC should be added as selected features.

(7) Two-step nature of the analysis: First you fit a mechanistic model, then you use the predictions of this model to perform clustering and prediction of groups (unsupervised then supervised). Thus you do not propagate the uncertainty intrinsic to your first estimation through the second step, ie. all the viral load selected features actually have a confidence bound which is ignored. Did you consider a one-step analysis in which your covariates of interest play a direct role in the parameters of the mechanistic model as covariates? To pursue this type of analysis SCM (Johnson et al. Pharm. Res. 1998), COSSAC (Ayral et al. 2021 CPT PsP), or SAMBA ( Prague et al. CPT PsP 2021) methods can be used. Did you consider sampling on the posterior distribution rather than using EBE to avoid shrinkage?

(8) Need for advanced statistical methods: The analysis is characterized by a lack of power. This can indeed come from the sample size that is characterized by the number of data available in the study. However, I believe the power could be increased using more advanced statistical methods. At least it is worth a try. First considering the unsupervised clustering, summarizing the viral shedding trajectories with features collapses longitudinal information. I wonder if the R package « LongituRF » (and associated method) could help, see Capitaine et al. 2020 SMMR. Another interesting tool to investigate could be latent class models R package « lcmm » (and associated method), see Proust-Lima et al. 2017 J. Stat. Softwares. But the latter may be more far-reached.

(9) Study intrinsic limitation: All the results cannot be extended to asymptomatic patients and patients infected with recent VOCs. It definitively limits the impact of results and their applicability to public health. However, for me, the novelty of the data analysis techniques used should also be taken into consideration.

Strengths are:<br /> - Unique data and comprehensive analysis.<br /> - Novel results on viral shedding.

Weaknesses are:<br /> - Limitation of study design.<br /> - The need for advanced statistical methodology.

Author response:

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

Public Reviews

Reviewer #1 (Public Review):

Summary:

The findings in this manuscript are important in the gene editing in human-derived hematopoietic stem and progenitor cells. By optimizing the delivery tool, adding DNA-PK inhibitor and including spacer-breaking silent mutations, the editing efficiency is significantly increased, and the heterozygosity could be tuned. The editing is even across the hematopoietic hierarchy.

Strengths:

The precise gene editing is important in gene therapy in vitro and in vivo. The manuscript provides solid evidence showing the efficacy and uniqueness of their gene editing approach.

Weaknesses:

There are several extended and unique points shown in this paper but in a specific cell population.

The findings are indeed in a specific cell lineage, though it should be noted the editing crossed multiple cell types within that lineage. More importantly though, HSPC have substantial relevance to understanding adult stem cell biology, blood formation, and leukemia. Critically, they are also the target cells for a plethora of gene therapies for anemias, immunodeficiencies, metabolic disorders, and are also being explored for use with CAR technologies. Indeed, CRISPR-based gene therapy was recently approved for clinical use. As such, the findings here are of substantial relevance for multiple areas of research including hematology, stem cell biology, cancer, immunology and more.

Reviewer #2 (Public Review):

Summary:

This work by Cloarec-Ung et al. sets out to uncover strategies that would allow for the efficient and precision editing of primitive human hematopoietic stem and progenitor cells (HSPCs). Such effective editing of HSPCs via homology directed repair has implications for the development of tractable gene therapy approaches for monogenic hematopoietic disorders as well as precise engineering of these cells for clinical regenerative and/or cell therapy strategies. In the setting of experimental hematology, precision introduction of disease relevant mutations would also open the door to more robust disease modeling approaches. It has been recognized that to encourage HDR, NHEJ as the dominant mode of repair in quiescent HSPCs must be inhibited. Testing editing of human cord blood HSPCs the authors first incorporate a prestimulation phase then identify optimal RNP amounts and donor types/amounts using standard editing culture conditions identifying optimal concentrations of AAV and short single-stranded oligonucleode donors (ssODNs) that yield minimal impacts to cell viability while still enabling heightened integration efficiency. They then demonstrate the superiority of AZD7648, an inhibitor of NHEJ-promoting DNA-PK, in allowing for much increased HDR with toxicities imparted by this compound reduced substantially by siRNAs against p53 (mean targeting efficiencies at 57 and 80% for two different loci). Although AAV offered the highest HDR frequencies, differing from ssODN by a factor by ~2-fold, the authors show that spacer breaking sequence mutations introduced into the ssODN to better mimic the disruption of the spacer sequence provided by the synthetic intron in the AAV backbone yielded ssODN HDR frequencies equal to that attained by AAV. By examining editing efficiency across specific immunophenotypically identified subpopulations they further suggest that editing efficiency with their improved strategy is consistent across stem and early progenitors and use colony assays to quantify an approximate 4-fold drop in total colony numbers but no skewing in the potentiality of progenitors in the edited HSPC pool. Finally, the authors provide a strategy using mutation-introducing AAV mixed with different ratios of silent ssODN repair templates to enable tuning of zygosity in edited CD34+ cells.

Strengths:

The methods are clearly described and the experiments for the most part also appropriately powered. In addition to using state of the art approaches the authors also provided useful insights into optimizing the practicalities of the experimental procedures that will aid bench scientists in effectively carrying out these editing approaches, for example avoiding longer handling times inherent when scaling up to editing over multiple conditions.

The sum of the adjustments to the editing procedure have yielded important advances towards minimizing editing toxicity while maximizing editing efficiency in HSPCs. In particular, the significant increase in HDR facilitated by the authors' described application of AZD7648 and the preservation of a pool of targeted progenitors is encouraging that functionally valuable cell types can be effectively edited.

The discovery of the effectiveness of spacer breaking changes in ssODNs allowing for substantially increased targeting efficiency is a promising advance towards democratizing these editing strategies given the ease of designing and synthesizing ssODNs relative to the production of viral donors.

The ability to zygosity tune was convincingly presented and provides a valuable strategy to modify this HDR procedure towards more accurate disease modelling.

Weaknesses:

Despite providing convincing evidence that functional progenitors can be successfully edited by their procedure, as the authors acknowledge it remains to be verified to what degree the self-renewal capacity and in vivo regenerative potential of the more primitive fractions is maintained with their strategy.

As other the 53BP1-based editing strategy that also disrupt DNA-PK have demonstrated maintained allele frequencies over engraftment time (De Ravin et al. Blood 2021), this suggests that a transient disruption of DNA-PK shouldn’t compromise regenerative potential. Of course, we strongly agree that maintained regenerative potential is important in any editing strategy. As such, for the version of record we have added clonal LT-CIC assessment using conditions that we’ve previously demonstrated predict long-term repopulating potential (Knapp et al. Nat Cell Bio 2018). This data, which has been added to Figure 3, shows no significant reduction in the frequency of the most potent LT-CIC in edited cells compared to unedited controls.

Assessments of the potential for off-target effects via the authors' approach was somewhat cursory and would have benefited from a more thorough evaluation.

Once again in the 53BP1 strategy, the authors of that study already performed CHANGE-seq, long-range PCR, NGS, and SKY with inhibition of this same pathway without obvious increases in off-target editing (as long as HDR donor was present, though they did interestingly observe increased large deletions when HDR donors were absent, De Ravin et al. Blood 2021). Our tests here were designed to confirm that our molecule was similarly not affecting off-target editing rather than to launch a large-scale investigation. We agree, however, that off-targets and particularly structural re-arrangements that could be missed by other approaches remain a concern. We have added in nanopore sequencing of the predicted off-target sites and thus verified more deeply that there was no change (indeed no observable off-target activity) at any of these sites. This data has been added to Figure 2 and to a new supplementary Figure S5. Additionally, while it’s beyond the scope of the current manuscript, a focused follow-up dedicated to structural rearrangements downstream of both single and multiple edits is currently in progress and will be submitted separately later this year.

Viability was assessed by live cell counting however given the short-term nature of the editing assay, more sensitive readouts of potentially compromised cell health could have provided a more stringent assessment of how the editing methodology impacted cell fitness.

Of course, we agree that viable cell counting does not fully predict whether the cell is viable in terms of retained proliferative potential or other functional potentials. This point was addressed for myeloid progenitors at least by the CFC assays already in the manuscript, as to form a colony these cells were definitionally viable at input. Indeed, in these tests, we did see a reduction beyond that of the viable counts as already discussed in the text. Similarly, we already inadvertently answered this in the general CD34+CD45RA- population in Figure 4C where we measured clonal growth following editing with different mutant to silent donor ratios. In this instance we observed 30-40% clonogenic frequencies (Figure 4C), though in this case without a specific non-edited control (as this was not the intended question). None-the-less, this would indicate that any general viability loss was no more than observed in the CFC tests (even if we assume 100% cloning efficiency if the cells had been unedited). Finally, the clonal LTC-IC show that while there is perhaps some loss in more committed progenitors, those with the highest self-renewal potential are not compromised in the edited condition compared to control (Figure 3I).

Recommendations for the authors

Reviewer #2 (Recommendations For The Authors):

It will be important to include the author-provided new paragraph in the discussion to contextualize this work in the existing HSPC editing landscape and your unique findings.

A new paragraph detailing how our manuscript fits with other recently published works is now included in the discussion.

The legend for Figure 3 needs correction. Panel E is incorrectly labeled as panel D and panel F is incorrectly labeled as panel E.

Thank you for catching this typo. It has been fixed.

In Figure 4 axis headings in panel C and D require clarity beyond simply titles of "Mean Frequency".

These axis labels have been clarified.

eLife assessment

This study presents an important methodology to increase the efficiency and precision of gene editing in human hematopoietic stem and progenitor cells. The evidence supporting the claims is convincing in that primitive LTC-ICs were minimally affected as a result of the editing procedure and the lack of edits at predicted off-target sites. The work will be of interest to biologists studying hematopoietic stem and progenitor cells and genome editing for potential clinical applications.

Reviewer #2 (Public Review):

Summary:

This work by Cloarec-Ung et al. sets out to uncover strategies that would allow for the efficient and precision editing of primitive human hematopoietic stem and progenitor cells (HSPCs). Such effective editing of HSPCs via homology directed repair has implications for the development of tractable gene therapy approaches for monogenic hematopoietic disorders as well as precise engineering of these cells for clinical regenerative and/or cell therapy strategies. In the setting of experimental hematology, precision introduction of disease relevant mutations would also open the door to more robust disease modeling approaches. It has been recognized that to encourage HDR, NHEJ as the dominant mode of repair in quiescent HSPCs must be inhibited. Testing editing of human cord blood HSPCs the authors first incorporate a prestimulation phase then identify optimal RNP amounts and donor types/amounts using standard editing culture conditions identifying optimal concentrations of AAV and short single-stranded oligonucleotide donors (ssODNs) that yield minimal impacts to cell viability while still enabling heightened integration efficiency. They then demonstrate the superiority of AZD7648, an inhibitor of NHEJ-promoting DNA-PK, in allowing for much increased HDR with toxicities imparted by this compound reduced substantially by siRNAs against p53 (mean targeting efficiencies at 57 and 80% for two different loci). Although AAV offered the highest HDR frequencies, differing from ssODN by a factor by ~2-fold, the authors show that spacer breaking sequence mutations introduced into the ssODN to better mimic the disruption of the spacer sequence provided by the synthetic intron in the AAV backbone yielded ssODN HDR frequencies equal to that attained by AAV. By examining editing efficiency across specific immunophenotypically identified subpopulations they further suggest that editing efficiency with their improved strategy is consistent across stem and early progenitors and use colony assays to quantify an approximate 4-fold drop in total colony numbers but no skewing in the potentiality of progenitors in the edited HSPC pool. Finally, the authors provide a strategy using mutation-introducing AAV mixed with different ratios of silent ssODN repair templates to enable tuning of zygosity in edited CD34+ cells.

Strengths:

The methods are clearly described and the experiments for the most part also appropriately powered. In addition to using state-of-the-art approaches, the authors also provided useful insights into optimizing the practicalities of the experimental procedures that will aid bench scientists in effectively carrying out these editing approaches, for example avoiding longer handling times inherent when scaling up to editing over multiple conditions.

The sum of the adjustments to the editing procedure have yielded important advances towards minimizing editing toxicity while maximizing editing efficiency in HSPCs. In particular, the significant increase in HDR facilitated by the authors' described application of AZD7648 and the preservation of a pool of targeted progenitors is encouraging that functionally valuable cell types can be effectively edited.

The discovery of the effectiveness of spacer breaking changes in ssODNs allowing for substantially increased targeting efficiency is a promising advance towards democratizing these editing strategies given the ease of designing and synthesizing ssODNs relative to the production of viral donors.

The ability to zygosity tune was convincingly presented and provides a valuable strategy to modify this HDR procedure towards more accurate disease modelling.

Weaknesses:

Despite providing convincing evidence that functional progenitors can be successfully edited by their procedure, as the authors acknowledge it remains to be verified to what degree the survival/self-renewal capacity and in vivo regenerative potential of the more primitive fractions is maintained with their strategy. That said the inclusion of LTC-IC assays that verify the lack of effect on these quite primitive cells is encouraging that functionality of stem cells will be similarly spared.

Author response:

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

In this letter, we respond to each of the reviewers’ comments. We support responses by referring to the revised manuscript and, where necessary, by including additional descriptions and analyses that we consider extrinsic to the manuscript itself. In this letter, all changes to the manuscript are shown in blue. As noted, the displayed figures have been added to the manuscript or the SI. We believe that we have successfully addressed all comments and that the quality of our paper has improved significantly.

Comment 1: In addition to the technical comments by the reviewers, I would encourage the authors to discuss the dependency of their observations, e.g. emergence of microphase separation, not only on the sequence of the polypeptides, but also on the solution conditions. Similarly, the distributions of ions in the condensate bulk, interphase, and diluted phase, and hence the interfacial free energy, are significantly affected both by the chemical composition of the condensate and the salt concentration itself, see: https://pubs.acs.org/doi/10.1021/acs.nanolett.1c03138

We thank the editor for this suggestion. Here, we have focused on the effect of sequence on condensate organization. We agree that how changes in solution condition affect condensate, including microphase separation of ELPs, is potentially interesting as well. We note this as a possible future direction at multiple places in the revised Conclusions and Discussion:

“The simulations successfully reproduced condensate stability variation upon amino acid substitution. While our study is performed at set salt concentration and temperature to isolate the contributions of amino acid hydrophobicity to condensate organization, future studies may consider implementing temperature [cite] or salt [cite] dependent models to explore how solution conditions affect the organization of ELP condensates.”

“Such a microenvironment arises from the collective behavior of many proteins, can deviate from that of individual chains, and is likely sensitive to the solution conditions,[cite] which are held constant in our study. Future work on systems with double amino acid substitutions or changes to salt concentration or temperature could elucidate the generality of the mean field interpretation and the additivity of individual contributions.”

Response to referee 1

Comment 0: This is an interesting, informative, and well-designed study that combines theoretical and experimental methodologies to tackle the phenomenon of higher-resolution structures/substructures in model biomolecular condensates. The results should be published. However, there is significant room for improvement in the presentation and interpretation of the results. As it stands, the precise definition of “frustration,” which is a main theme of this manuscript (as emphasized in the title), is not sufficiently well articulated. This situation should be rectified to avoid ””rustration” becoming a ”catch-all” term without a clear perimeter of applicability rather than a precise, informative description of the physical state of affairs. There are also a few other concerns, e.g., regarding interpretation of correlation of phase-separation critical temperature and transfer free energy of amino acid residues as well as the difference between critical temperature and onset temperature, and the way the simulated configurations are similar to that of gyroids.

We want to thank the reviewers for their insightful comments. We revised the manuscript extensively to improve its clarity and to address the reviewers’ concerns. In the following, we provide point-to-point responses to all the comments.

Comment 1: It is accurately pointed out on p.4 that elastin-like polypeptides (ELPs) undergo heat-induced phase separation and therefore exhibit lower critical solution temperatures (LCSTs). But it is not entirely clear how this feature is reproduced by the authors’ simulation. A relationship between simulated surface tension and “transition temperature” is provided in Fig.1C; but is the ”transition temperature” (authors cited ref.41 by Urry) the same as critical temperature? Apparently, Urry’s Tt is””critical onset temperature”, the temperature when phase separation happens at a given polymer concentration. This is different from the (global) critical temperature LCST - though the two may be correlated-or not-depending on the shape of the phase boundary. Moreover, is the MOFF coarse-grained forcefield (first step in the multi-scale simulation), by itself, capable of reproducing heat-induced phase separation in a way similar to the forcefield of Dignon et al., ACS Cent Sci 5, 821-230 (2019)? Or is this temperature-dependent effect appearing only subsequently, after the implementation of the MARTINI and/or all-atom steps? Clarification is needed. To afford a more informative context for the authors’ introductory discussion, the aforementioned Dignon et al. work and the review by Cinar et al. [Chem Eur J 25, 13049-13069 (2019)], both touching upon the physical underpinning of the LCST feature of elastin, should also be cited along with refs.41-43.

We thank the reviewer for their comment. First, we apologize for the lack of clarity between the global lower critical solution temperature, Tc, and the transition temperature, Tt. We have modified the manuscript to be more explicit that the transition temperature we utilize is dependent on the solution conditions, instead of the global lower critical solution temperature.

Author response image 1.

Tt as a function of concentration for ELP[V5A2G3] constructs of different chain lengths. Logarithmic fits to the data for each construct using Eq. 1 are also shown. It is evident that the different curves converge to the critical temperature Tc at the critical concentration Cc. Figure reproduced from ref.[2] CC BY 4.0.

However, as shown by Chilkoti and coworkers [1, 2] and in Author response image 1, the critical temperature of ELPs Tc is indeed linearly related to Tt with the following relationship

The above equation highlights the dependence of Tt on the chain length (length) and polymer concentration (conc). The parameter Cc is the corresponding theoretical polypeptide concentration that would be required to achieve Tc, and k is the proportionality constant. Instead of making computationally expensive predictions of condensate critical temperatures, we focused on the surface tension, which can be more readily determined from single constant temperature simulations as detailed in the Methods section. This decision was made so to make it computationally feasible to systematically probe the properties of all 20 amino acids in diblock ELPs in our multiscale model. Furthermore, an expected relationship between the critical temperature and the surface tension can be inferred based on the Flory Huggins theory. In particular, relationships between the Flory Huggins parameter, χ, and interfacial tension (τ) have been investigated, and the relationship can be approximated as

where α is a positive constant, whose exact value depends on the proximity of χ to the critical value of χ necessary for phase separation (χC).[3, 4] As detailed in new Supplemental Theory of the Supporting Information, for systems undergoing LCST,

with Therefore, we have

Several conclusions can be drawn from Eq. 4. First, for α = 1, τ is linearly proportional to Tc. Secondly, τ decreases at larger values for Tc since trend that is consistent with results presented in Figure 1 of the main text. Finally, as detailed in the Supplemental Theory, the inverse relationship between τ and Tc is only expected for systems exhibiting LCSTs. For systems with UCST, τ increases at larger Tc. Therefore, reproducing the correct trend supports the model’s ability to capture the temperature-dependent effect specific to the ELP system.

We modified the text to define the physical meaning of Tt more explicitly. Furthermore, we added a new section in the Supporting Information titled Supplemental Theory to detail the relationship between Tt, Tc, the Flory-Huggins parameter χ, and the surface tension τ. The updated text now reads:

“Utilizing the simulated condensate conformations, we computed various quantities to benchmark against experimental measurements. While the critical temperature has been widely used as a measure for condensate stability, determining it computationally is expensive. As an alternative, we computed the surface tension, τ, using 100-µs-long MARTINI simulations performed with the NPNAT ensemble.[cite] As detailed in the Supplemental Theory in the Supporting information, an inverse relationship is expected between τ and the critical temperature, Tc, for systems exhibiting LCSTs. We further approximate Tc with the transition temperatures (Tt) of ELP sequences,[cite] which are the temperatures at which ELPs undergo an LCST transition at a specified solution condition. Tt was shown to be linearly proportional to TC[cite]. As expected, a negative correlation can be readily seen between computed surface tension and experimental Tt (Fig. 1C). This observed negative correlation between Tt and τ supports the simulation approach’s accuracy in reproducing the sequence-dependent changes in ELP phase behavior.”

The reviewer is correct that MOFF does not explicitly account for temperature-dependent effects in its interaction parameters. But as mentioned above and indicated by the reviewer, the following steps with explicit solvent simulations in the multiscale strategy succeed in capturing sequence-dependent differences in ELP systems, which are evident in both transition temperature and surface tension.

We cited the two references suggested by the reviewer in the introduction. We further added the following text in the discussion section to suggest explicitly exploring temperature-dependent effects as an interesting future direction.

“While our study is performed at set salt concentration and temperature to isolate the contributions of amino acid hydrophobicity to condensate organization, future studies may consider implementing temperature[cite] or salt[cite] dependent models to explore how solution conditions effect the organization of ELP condensates.”

Comment 2: “Frustration” and ”frustrated” are used prominently in the manuscript to characterize certain observed molecular configurations (11 times total, in both the title and in the abstract). Apparently, it is the most significant conceptual pronouncement of this work, hence its precise meaning is of central importance to the authors’ thesis. Whereas one should recognize that the theoretical and experimental observations are striking without invocation of the “frustration” terminology, usage of the term can be useful if it offers a unifying conceptual framework. However, as it stands, a clear definition of the term “frustration” is lacking, leaving readers to wonder what molecular configurations are considered “frustrated” and what are not (i.e., is the claim of observation of frustration falsifiable?). For instance, “frustrated microphase separation” appears in both the title and abstract. A logical question one may ask is: “Are all microphase separations frustrated”? If the answer is in the affirmative, does invocation of the term “frustration” add anything to our physical insight? If the answer is not in the affirmative, then how does one distinguish between microphase separations that are frustrated from those that are not frustrated? Presumably all simulated and experimental molecular configurations in the present study are those of lowest free energy for the given temperature. In other words, they are what they are. In the discussion about frustrated phase separation on p.13, for example, the authors appear to refer to the fact that chain connectivity is preventing hydrophobic residues to come together in a way to achieve the most favorable interactions as if there were no chain connectivity (one may imagine in that case all the hydrophobic residues will form a large cluster without microphase separation). Is this what the authors mean by “frustration”? If that’s true, isn’t that merely stating the obvious, at least for the observed microphase separation? In general, does “frustration” always mean deviation of actual, physical molecular configurations from certain imagined/hypothetical/reference molecular configurations, and therefore dependent upon the choice of the imagined reference configuration? If this is how the authors apply the term “frustration” in the present work, what is the zero-frustration reference state/configuration for microphase separation? And, similarly, what is the zero-frustration reference state/configuration when frustrated EPS-water interactions are discussed (p.14-p.15, Fig.5)? How do non-frustrated water-protein interactions look like? Is the classic clathrate-like organization of water hydrogen bonds around small nonpolar solute “frustrated”?

We thank the reviewer for their insightful comment, and agree that the concept of “frustration” is both important to our conclusions and, upon review, is too vague in our previous draft of the manuscript.

For conceptual simplicity and to maximize transferability to real biological systems, we will focus our discussion of frustration on one specific type, which we term “chain frustration.” Chain frustration occurs in states where tertiary interactions between chemically distinct polymer blocks favor phase separation, while chain connectivity prevents macroscopic phase separation from occurring.[5] This frustration leads to microphase separation with microdomains of different monomers.

We agree with the reviewer that “all microphase separations” are frustrated, and have revised the title to

“Microphase Separation Produces Interfacial Environment within Diblock Biomolecular Condensates”

Furthermore, we also removed frustration from the abstract to read

“The interspersion of hydrophilic and hydrophobic residues and a lack of secondary structure formation result in an interfacial environment, which explains both the strong correlation between ELP condensate stability and interfacial hydrophobicity scales, as well as the prevalence of protein-water hydrogen bonds.”

We have limited our discussion of the frustration to the incomplete separation of hydrophobic and hydrophobic groups. As pointed out by the reviewer, in this case, frustration refers to the fact that chain connectivity is preventing hydrophobic residues from coming together in a way to achieve the most favorable interactions as if there were no chain connectivity. The reference would be a perfectly macroscopic phase separation that partitions hydrophobic from hydrophilic groups.

While the frustration from chain connectivity is well understood for block copolymers[5], its effect on producing the interfacial solvation environment, to the best of our knowledge, has not been emphasized before. We have revised the text at the point where we mention frustration to clearly define its meaning.

“Therefore, while microphase separation occurs in ELP condensates, frustration remains in the system. Hydrophilic residues cannot completely separate from hydrophobic ones due to constraints imposed by the acid sequence, creating unique microenvironments.”

When discussing the interactions between ELP and water, we used the hydrogen bond analysis to emphasize the interfacial environment. For example, the hydrophobic residues tend to “repel” water molecules, reducing the hydrogen bond density; on the other hand, hydrophilic residues and backbone retain water molecules. This difference resulted in the positive and negative correlation with Tt shown in Fig 5C. The behavior of water molecules is, therefore, inhomogeneous inside the condensate. We expect water molecules to become frustrated due to the simultaneous contact with both hydrophobic and hydrophilic chemical groups, and a perfect reference state would be the pure water environment. However, since this point is not central to our study, to avoid confusion, we have avoided mentioning frustration and revised the text to read amino acid sequence, creating unique microenvironments.”

“The water hydrogen bond density also highlights an interfacial environment of blended hydrophobic and hydrophilic regions.”

After revising the text, frustration only appears three times in the manuscript.

Comment 3: In the discussion about the correlation of various transfer free energy scales for amino acids and Urry’s critical onset temperature (ref.41) on p.11 and Fig.4, is there any theoretical relationship to be expected between the interactions among amino acids of ELPs and their critical onset temperatures? While a certain correlation may be intuitively expected if the free energy scale ”is working”, is there any theoretical insight into the mathematical form of this relationship? A clarifying discussion is needed because it bears logically on whether the observed correlation or lack thereof for different transfer energy scales is a good indication of the adequacy of the energy scales in describing the actual physical interactions at play. This question requires some prior knowledge of the expected mathematical relationship between interaction parameters and onset temperature.

We thank the reviewer for their comment. The exact relationship between the interactions between amino acids and their transition temperature can be understood in terms of the Flory-Huggins theory, which describes the thermodynamics of polymer mixtures using a lattice model. The chemical composition of the mixture is built into the polymer-solvent interaction parameter

Where is the coordination number, T is the temperature, kB is the Boltzmann constant, and {ϵpp, ϵss, ϵps} are the strength of polymer-polymer, solventsolvent, and polymer-solvent interactions respectively.[6]

From the original derivation of Flory-Huggins theory, it can be shown that phase separation occurs when χ is greater than its critical value, or χC, we can derive the critical temperature as

Δϵ can indeed be interpreted as the free energy cost of transferring a polymer bead from a solution phase to a polymer phase. It corresponds to the change of energy from a mixed state, with contacts between polymer and solvent (ϵps), to the demixed state with only polymer-polymer (ϵpp) and solvent-solvent (ϵss) contacts.

Therefore, the transfer free energy, and the interactions among amino acids of ELPs, are expected to correlate with the critical temperature. The above discussion has been incorporated into the new section Supplemental Theory in the Supporting Information. There, we also discuss the more general scenario where Δϵ is temperature dependent, which is essential for giving rise to LCST.

We have modified the main text in the discussions of Figure 4 to better explain these mathematical relationships and their necessary assumptions in order to help interpret our simulations. Here is an expert from where we discuss Figure 4:

“The strong dependence of molecular organization on amino acid hydrophobicity suggests that the solvation environment of individual residues might be a determining factor for condensate stability. Indeed, as shown in the Supplemental Theory of the Supporting Information, the critical temperature is closely related to the free energy cost of transferring polymer beads from a solution state to a polymer-only environment. This transfer free energy is often used to quantify the hydrophobicity of amino acids [cite]. To explore their relationship more quantitatively, we compared the transition temperature for ELP condensates measured by Urry [cite] to several hydrophobicity scales.”

Comment 4: To provide a more comprehensive context for the present study, it is useful to compare the microphase separation seen in the authors’ simulation with the micelle-like structures observed in recent simulated condensed/aggregated states of hydrophobic-polar (HP) model sequences in Statt et al., J Chem Phys 152, 075101 (2020) [see esp. Fig.6] and Wesse´n et al., J Phys Chem B 126, 9222-9245 (2022) [see, e.g., Fig.10].

We thank the reviewer for this suggestion. The results of Statt et al. and Wessen et al.´ indeed provide a nice comparison to our results. While we capture some of the same behavior they observe, the full array of chemical space in our model seems to give some additional morphologies as well.

First, as predicted by the self-consistent field theory, block copolymers are expected to form primarily lamellar like micelles that clearly seperate the dense and dilute phase when the volume fraction, f, is 0.5 (Response to Comment 5). This prediction is indeed consistent with results from simulations with the HP model, and is consistent with our simulations when the substituted amino acid, X, is sufficiently polar.

However, this observation is only one of several behaviors we observe. In particular, our simulations also produce gyroid-like structures, which are predicted to emerge at small volume differences, i.e. f ≈ 0.4 or f ≈ 0.6. These different configurations likely emerge due to the more realistic representation of amino acids in our model, which presents more frustration than the HP model. In particular, the backbone atoms are inherently hydrophilic and cannot separate from the hydrophobic side chains. Therefore, under microphase separation, it is inherently difficult to separate the different chemical groups to form lamellar or micelle-like structures. This produces a condensate interior with interfacial properties that may not be captured by the HP model.

We make note of the micelle-like topologies predicted by HP models in the revised text, citing both Statt et al. and Wessen et al.:´

“Surprisingly, microphase separation did not produce lamellar morphology as expected for block copolymers with equal volume fraction of the two blocks (Fig. S3 in the Supporting Information) [cite]. In particular, the condensates appear to form gyroid-like structures (Fig. S4 in the Supporting Information), in which the V and X blocks form two interpenetrating networks. This morphology also differs from micelle-like structures seen in simplified hydrophobicpolar (HP) polymers [cite]. It promotes interfacial contacts while maintaining substantial self-interactions as well. Weak interfacial tension between different ELP blocks has also been noted by Hassouneh et al.[cite]”

Comment 5: ”Gyroid-like morphology” is mentioned several times in the manuscript (p.4, p.8, p.17, Fig.S3). This is apparently an interesting observation, but a clear explanation is lacking. A more detailed and specific discussion, perhaps with additional graphical presentations, should be provided to demonstrate why the simulated condensed-phase ELP configurations are similar to the classical description of gyroid as in, e.g., Terrones & Mackay, Chem Phys Lett 207, 45-50 (1993) and Lambert et al., Phil Trans R Soc A 354, 2009-2023 (1996).

We thank the reviewer for their comment. Gyroids are canonical structures for diblock copolymers.[5, 7, 8, 9] Their stability is predicted using self-consistent field theory (SCFT), and occurs due to the balance of the volume fraction of polymer block A (fA), the length of the polymer (N), and the Flory-Huggins interaction parameter (χ).[8, 9] The prediction from SCFT suggests that gyroids occur at smaller values of χN and values fA near, but not equal to 0.5 (Author response image 2).[10] We hypothesize that these configurations emerge at equal molar fraction of V and X amino acids due to small differences in solvation volume between each half of the polymer chain.

Our support for gyroid-like structures is mainly from observations of two interpenetrating networks formed by the two ELP blocks. We have revised Figure S4 to clearly highlight the two networks as shown in Author response image 3.

We have revised the main text to clearly define the gyroid-like structures as interpenetrating networks, and added the theoretical phase diagram of diblock copolymers predicted by SCFT as Figure S3 in the Supporting Information.

“In particular, the condensates appear to form gyroid-like structures (Fig. S4 in the Supporting Information), in which the V and X blocks form two interpenetrating networks. This morphology also differs from micelle-like structures seen in simplified hydrophobic-polar (HP) polymers [cite]. It promotes interfacial contacts while maintaining substantial self-interactions as well. Weak interfacial tension between different ELP blocks has also been noted by Hassouneh et al.[cite]”

We note, however, that proving that our observations are indeed gyroid structures requires more sophisticated mathematical analysis that is beyond the scope of the study. It is also possible that these structures are metastable in our simulations. We emphasize these caveats in the updated Discussion Section.

“Further studies on the thermodynamic stability of these morphologies and comparing them with predictions from the self-consistent field theory shall provide more insights into the driving forces for their emergence [cite].”

Author response image 2.

Theoretical phase diagram[8] and corresponding morphologies for diblock copolymers. The phases are labeled as: body centered cubic (BCC), hexagonal cylinders (HEX), gyroid (GYR), and lamellar (LAM). fA is the volume fraction of a single polymer block, denoted A, χ is the Flory-Huggins interaction parameter, and N is the total degree of polymerisation. Figure reproduced from ref.[10] CC BY 4.0.

Author response image 3.

Representative configurations of (A) V5F5 and (B) V5L5 condensates from MARTINI simulations. The valine substituted half of the chain is colored blue (V5) and the X substituted half of the chain is colored red (X5). To highlight the interpenetrating networks formed by the two halves, only the X substituted half of the chain is shown on the left. Simulation interfaces are once repeated periodically in the positive x and positive y dimensions for clarity. High density regions formed by the multiple X substituted half of the chains are highlighted in yellow circles, with one of the chain shown in green.

Response to referee 2

Comment 1: The experimental characterization relies on BODIPY and SBD reporting, respectively, on viscosity and polarity. The fluorescent signal of these dyes can possibly depend on many other factors, including quenching. Additional controls are required, or a more extensive discussion with additional references, and a mention to potential limitations of this approach.

We agree with the reviewer that the fluorescence lifetime signal will be affected by many factors. Compared with the fluorescence intensity, the fluorescence lifetime mainly depends on the dyes’ self properties and environmental factors. BODIPY and SBD have been used in biological systems to detect the microviscosity and micropolarity of condensates. Our group published the same SBD and BODIPY fluorophores in previous work to quantify the microenvironment of protein aggregation and condensations. The extended data (ChemBioChem 20:1078–1087. doi: 10.1002/cbic.201800782; Aggregate 4:e301. doi:10.1002/agt2.301; Nat Chem Biol 1–9. doi:10.1038/s41589-023-01477-1) shows evidences that the BODIPY is only sensitive to the viscosity while SBD is only sensitive to the polarity, but nonsensitive to other environmental factors. As for the quenched issue, the fluorophores with extended pi-rich structure display aggregation-caused quenching (ACQ) effect in high probe concentration, which will lower the fluorescence lifetime and intensity. We usually labeled the 20% molar ratio of the ELPs using NHS-ester fluorophores to get stock solutions. Due to the labeling efficiency, the exact labeling ratio is much lower than 20%. The labeled ELP stock solution will be further mixed with unlabeled ELP to get ELP solutions with low labeling fractions. We measured the ELPs labeled with a different fraction of dyes. The result shows that only BODIPY performs slight ACQ phenomena at a high

Author response image 4.

FLIM images of ELP condensates labeled with different fractions of dyes. A) FLIM images of V30A30 condensates with 5%, 2.5%, and 1% BODIPY labels. B) FLIM images of V30A30 condensates with 5%, 2.5%, and 1% fraction of SBD. Droplets were formed with a final concentration of 70 µM ELP labeled with different fractions of BODIPY or SBD in 2 M NaCl solution. Scale bar:5 µm.

To mostly avoid the potential ACQ effect and achieve enough fluorescence signals, we finally use the ELP labeled with a lower fraction of dyes, 1% of BODIPY and 2.5 % of SBD, to perform the FLIM experiments. The data in Figure 3 will be corrected with the following data.

Author response image 5.

Structures of NHS-BODIPY and NHS-SBD, and representative FLIM images of V30A30, A30V30, V30G30 and G30V30 labeled with respective fluorophores. The fluorescence lifetime of each image is the average acquired from three independent experiments. Scale bar: 5 µm.

We revised the text in the section Microphase separation of ELP condensates as follows “To experimentally test the microphase separation behavior uncovered in simulations, we studied the micro-physicochemical properties of the V-end and X-end of the peptides. We constructed diblock peptides with the combination of 30 pentameric repeats of V block and X (A or G) block, namely V30A30 and V30G30 (Experimental Sequences Section in the Supporting Information). The amino-termini of V30A30 and V30G30 sequences were subsequently labeled with environmentally sensitive BODIPY or SBD fluorophores [cite], whose lifetime could be measured to quantify the viscosity or polarity of the V-end (Fig. 3A, left panel) [cite]. These probes have been reported to be only sensitive to single physicochemical properties.[cite] To avoid artifacts induced by fluorophore labeling, we usually used ELPs labeled with a low fraction of dyes. We also constructed A30V30 and G30V30 diblock peptides, wherein the viscosity or polarity of the A-end or the G-end could be measured by fluorophores that are attached at the amino-terminus (Fig. 3A, right panel). Using FLIM, we found that the lifetime of BODIPY for the V-end (5.43 ns) was longer than that for the A-end (4.35 ns), suggesting that the V-end indeed has a higher microviscosity than the A-end (ηV= 2233.54 cp vs ηA= 969.57 cp). Accordingly, the lifetime of SBD was longer for the V-end (8.75 ns) than the A-end (7.00 ns), indicating that the micropolarity of the V-end was lower than the A-end (ϵV= 13.25 vs ϵA = 18.97). These observations could be largely attributed to the greater extent of dehydration at the V-end due to its higher local peptide density. We further showed that the observed differences are not results of possible artifacts arising from any subtle distinctions between the two sequences V30A30 and A30V30 (Experimental Characterization of ELP Condensates Section in the Supporting Information, Fig. S8-S9 in the Supporting Information). Similar results were observed using the V-G sequences. FLIM experiments revealed that the V-end was more viscous than the G-end (ηV= 2972.72 cp vs ηG= 1958.60 cp) and the V-end was less polar than the G-end (ϵV= 9.14 vs ϵG = 27.50). These experimental observations provided the first line of evidence to support the microphase separation, as suggested by the simulation results.”

We revised the text in the section Experimental methods as follows

“The proteins of interest were labeled with NHS ester fluorophore. We used ELPs with 1% BODIPY labels or 2.5% SBD labels to form condensates, which avoid the artifacts induced by fluorophores. Droplets were formed with the final concentration of 70 µM ELP in 2 M NaCl for V-A and 1.5 M NH4SO4 for V-G diblock, respectively. A drop of droplets containing solution was placed on a 0.17 mm coverslip with a 500 µm spacer. Images were acquired by Leica Falcon Fluorescence Microscope equipped with Wil pulse laser and 63X/0.12 oil-immersion objective. The BODIPY was excited at 488 nm and the SBD was excited at 448 nm. The fluorescence lifetime fitting and image analysis were performed in LAS X and Image J.”

We also used a lower concentration of free dyes to remeasure the properties of the ELP condensates. The Figure S9 data are corrected as follows. The slight differences between the results are caused by experimental errors, which don’t affect the conclusion.

Author response image 6.

FLIM image of unlabeled ELP condensates. A) Chemical structure of free fluorophore, which can measure the physicochemical properties of condensates without labeling. B) Representative FLIM images of V30A30 and A30V30. The mix is the mixture of V30A30 (35 µM) and A30V30 (35 µM). Droplets were formed with a final concentration of 70 µM ELP in 2 M NaCl solution with 1 µM fluorophore. C) Representative FLIM images of V30G30 and G30V30. Droplets were formed with a final concentration of 70 µM ELP in 1.5 M (NH4)2SO4 solution with 1 µM fluorophore. The mix is the mixture of V30G30(35 µM) and G30V30 (35 µM). Scale bar, 5 µm. The fluorescence lifetime of each image is the average from three independent measurements.

We also revised the Sequence dependence of micro-viscosity and polarity section of the Supporting Information as follows

“Since we used V30X30 and X30V30 to quantify the V- and X-end of the V-X blocks, it is possible that the observed differences arose from the innate property of the V30X30 and X30V30 sequences. To rule out this artifact, we formed the ELP condensates with sequences of V30X30, X30V30, or the V30X30 and X30V30 mixture. The condensates were subsequently treated with the aldehydeBODIPY and methyl-ester SBD fluorophores without the NHS ester reactive warhead (Fig. S9A in the Supporting Information). After brief incubation, aldehyde-BODIPY and methyl-ester SBD fluorophores were recruited into and homogeneously distributed in the ELP condensates. The fluorescence lifetime of aldehyde-BODIPY was the same for V30A30 (4.96 ns), A30V30 (4.99 ns), and their mixture (4.98 ns) (Fig. S9B in the Supporting Information, upper panel). Interestingly, this value is around the average (4.89 ns) of the A-end (4.35 ns) and the V-end (5.43 ns) labeled NHS-BODIPY. For the SBD measurement, methyl-ester SBD resulted in almost identical lifetime values of V30A30 (8.25 ns), A30V30 (8.27 ns), and their mixture (8.28 ns) (Fig. S9B in the Supporting Information, lower panel), again around the average values (7.88 ns) of the A-end (7.00 ns) and the V-end (8.75 ns) labeled NHS-SBD. In addition to the V-A blocks, similar observations were made for the V-G blocks as V30G30 and G30V30 sequences (Fig. S9C in the Supporting Information). The slight difference between the results is attributed to the experiment errors. Because the fluorophores did not covalently label the amino-terminus of the ELP peptides, their lifetime reports closer to the averaged property of the condensates instead of the microscopic property of the V-end or the X-end when the number of molecules is sufficient and the molecular distribution has no preference.

Our results reveal that the V30X30 and X30V30 condensates exhibited similar macroscopic viscosity or polarity, suggesting that the previously observed different viscosity or polarity of V30X30 and X30V30 could be attributed to the microscopic property of the V-end or X-end.”

The FLIM technique combined with environment-sensitive fluorophores is a powerful tool for us to investigate the physicochemical properties of the microenvironment within the condensates. However, there are some limitations to this method. As the fluorophore is labeled in the protein, we can only detect the microenvironment surrounding the surface of the probe(the distance may be angstrom level). The fluorescence signal values we got are the statistical average of the fluorescence signals from the complex microenvironments. The signal from the probes is determined by the sampling position, orientation, and number of fluorescent probes. So the quantified values can be compared relatively, but these values can not accurately describe the physical or chemical states in different systems. In addition, the resolution in FLIM experiments is not enough to directly distinguish the microstructure in condensates.

Comment 2: It is unclear if, after the application of stretching, the micro-structure will eventually return to the original configuration or not. Overall, the point of this experiment remains somewhat unclear.

We thank the reviewer for this comment. The ELP condensates are actually viscous fluids and they could coalesce into larger droplets within seconds. Due to the high viscosity, ELP condensates show slow fluorescence recovery after photobleaching. As stretching the condensates, the micro-structure of condensates changes to show a response to the outer force. The fluorophores may be pulled out from the microenvironment. For such a dynamic system, we speculate that the microstructure will return to the original after the condensation system equilibrium, which may be a long process. However, it is hard to characterize whether these microstructures have completely returned to their original positions. The purpose of this experiment is to show the microenvironment properties of each terminal in another aspect. The experiment also shows evidence that the microenvironment around the V terminus is more dense than the A terminus.

Comment 3: The title is too generic and does not reflect the content of the work. There is no analysis of biological condensates. The results are specific to di-block polypetides with specific sequences. This should be clearly specified in text and title.

We have revised the title to ”Microphase Separation Produces Interfacial Environment within Diblock Biomolecular Condensates”

Comment 4: MD is out of the expertise of this reviewer. However, when looking at the density profiles (Figure S2), the simulation does not seem to be fully converged. The densities fluctuate inconsistently along the Z direction. The authors should comment on assessing simulation convergence. In many cases, the section used for the density values in the plot (i.e., below 0.06 box lengths away from the condensate center) does not seem representative of the dense phase. It should be justified, why these simulations can still be used for density/hydrogen bonding analysis.

We thank the reviewer for their comment, and agree that convergence of MD simulations is simultaneously important and difficult to control for. To demonstrate the convergence of our simulations, we have taken an example system (V5F5) and reproduced the density profile in 4 unique time windows of 50 ns each (Author response image 7A-D). We find that all distributions are nearly identical, indicating that further extending these simulations is unlikely to change our findings.

While we agree that the choice of 0.06 box lengths is arbitrary, it was chosen as an approximation for the interior of the condensate, where the more hydrophobic half of the protein chain tends to be at higher concentration. However, this choice is not important to our overall conclusion. Halving (Author response image 7E) or doubling (Author response image 7F) the cutoff maintains the inverse correlation between the protein density of the X5 half of the condensate and experimental transition temperature.

Finally, in our multiscale simulation approach, the all-atom portion of the simulation is mostly used to examine water structure and protein solvation. We can see that dividing the simulation into four independent time estimates does not substantially change these properties, resulting in low standard deviations in Figure 5 and Figure 6. Similarly, our previous work on the dielectric of ELP condensates has shown that choosing different starting structures from MARTINI simulations is unlikely to effect the estimate of similar quantities.[11]

Author response image 7.

Checking convergence of all-atom simulations of ELP condensates. (A-D) The relative mass density along the Z-distance from the condensate center is shown for the V-substituted and X-substituted halves of V5F5 in four independent time windows of 50 ns each. The Z−axis is defined as the direction perpendicular to the condensate-water interface. The dashed line represents a Z-distance of 0.06 box lengths away from the condensate center, which was the original cutoff for correlation analysis. E-F) Correlation between the mass fraction of the X5 half of the condensate and transition temperature (Tt) from Urry.[12] The condensate is defined as having a Z-distance of 0.03 box lengths (E) or 0.12 box lengths (F) away from the condensate center. ρ is the Pearson correlation coefficient between the two data sets, and the dashed diagonal line is the best fit line. Error bars represent standard deviations of the mean taken over box length intervals of 0.01.

References

(1) McDaniel JR, Radford DC, Chilkoti A (2013) A unified model for de novo design of elastin-like polypeptides with tunable inverse transition temperatures. Biomacromolecules 14:2866–2872.

](2) Meyer DE, Chilkoti A (2004) Quantification of the effects of chain length and concentration on the thermal behavior of elastin-like polypeptides. Biomacromolecules 5:846–851.

(3) Helfand E, Tagami Y (1972) Theory of the interface between immiscible polymers. J. Chem. Phys. 56:3592.

(4) Roe RJ (1975) Theory of the interface between polymers or polymer solutions. I. Two components system. J. Chem. Phys. 62:490–499.

(5) Shi AC (2021) Frustration in block copolymer assemblies. J. Phys. Condens. Matter 33.

(6) Flory PJ (1942) Thermodynamics of high polymer solutions. J. Chem. Phys. 10:51.

(7) Grason GM (2006) The packing of soft materials: Molecular asymmetry, geometric frustration and optimal lattices in block copolymer melts. Phys. Rep. 433:1–64.

(8) Matsen MW, Bates FS (1996) Unifying weak- and strong-segregation block copolymer theories. Macromolecules 29:1091–1098.

(9) Matsen MW, Schick M (1994) Stable and unstable phases of a diblock copolymer melt. Phys. Rev. Lett. 72:2660–2663.

(10) Swann JM, Topham PD (2010) Design and application of nanoscale actuators using block-copolymers. Polymers 2:454–469.

(11) Ye S et al. (2023) Micropolarity governs the structural organization of biomolecular condensates. Nat. Chem. Biol. pp 1–9.

(12) Urry DW (1997) Physical chemistry of biological free energy transduction as demonstrated by elastic protein-based polymers. J. Phys. Chem. B 101:11007–11028.

eLife assessment

This important study investigates the structural organization of a series of diblock elastin-like polypeptide condensates. The methodology is highly compelling, as it combines multiscale simulations and fluorescence lifetime imaging microscopy experiments. The results increase our understanding of model biomolecular condensates.

Reviewer #1 (Public Review):

This is an interesting, informative, and well-designed study that combines theoretical and experimental methodologies to tackle the phenomenon of higher-resolution structures/substructures in model biomolecular condensates. However, there is significant room for improvement in the presentation and interpretation of the results. As it stands, the precise definition of "frustration," which is a main theme of this manuscript (as emphasized in the title), is not sufficiently well articulated. This situation should be rectified to avoid "frustration" becoming a "catch-all" term without a clear perimeter of applicability rather than a precise, informative description of the physical state of affairs. There are also a few other concerns, e.g., regarding interpretation of correlation of phase-separation critical temperature and transfer free energy of amino acid residues as well as the difference between critical temperature and onset temperature, and the way the simulated configurations are similar to that of gyroids. Accordingly, the manuscript should be revised to address the following:

(1) It is accurately pointed out on p.4 that elastin-like polypeptides (ELPs) undergo heat-induced phase separation and therefore exhibit lower critical solution temperatures (LCSTs). But it is not entirely clear how this feature is reproduced by the authors' simulation. A relationship between simulated surface tension and "transition temperature" is provided in Fig.1C; but is the "transition temperature" (authors cited ref.41 by Urry) the same as critical temperature? Apparently, Urry's Tt is "critical onset temperature", the temperature when phase separation happens at a given polymer concentration. This is different from the (global) critical temperature LCST - though the two may be correlated-or not-depending on the shape of the phase boundary. Moreover, is the MOFF coarse-grained forcefield (first step in the multi-scale simulation), by itself, capable of reproducing heat-induced phase separation in a way similar to the forcefield of Dignon et al., ACS Cent Sci 5, 821-230 (2019)? Or, is this temperature-dependent effect appearing only subsequently, after the implementation of the MARTINI and/or all-atom steps? Clarification is needed. To afford a more informative context for the authors' introductory discussion, the aforementioned Dignon et al. work and the review by Cinar et al. [Chem Eur J 25, 13049-13069 (2019)], both touching upon the physical underpinning of the LCST feature of elastin, should also be cited along with refs.41-43.

(2) "Frustration" and "frustrated" are used prominently in the manuscript to characterize certain observed molecular configurations (11 times total, in both the title and in the abstract). Apparently, it is the most significant conceptual pronouncement of this work, hence its precise meaning is of central importance to the authors' thesis. Whereas one should recognize that the theoretical and experimental observations are striking without invocation of the "frustration" terminology, usage of the term can be useful if it offers a unifying conceptual framework. However, as it stands, a clear definition of the term "frustration" is lacking, leaving readers to wonder what molecular configurations are considered "frustrated" and what are not (i.e.,is the claim of observation of frustration falsifiable?). For instance, "frustrated microphase separation" appears in both the title and abstract. A logical question one may ask is: "Are all microphase separations frustrated"? If the answer is in the affirmative, does invocation of the term "frustration" add anything to our physical insight? If the answer is not in the affirmative, then how does one distinguish between microphase separations that are frustrated from those that are not frustrated? Presumably all simulated and experimental molecular configurations in the present study are those of lowest free energy for the given temperature. In other words, they are what they are. In the discussion about frustrated phase separation on p.13, for example, the authors appear to refer to the fact that chain connectivity is preventing hydrophobic residues to come together in a way to achieve the most favorable interactions as if there were no chain connectivity (one may imagine in that case all the hydrophobic residues will form a large cluster without microphase separation). Is this what the authors mean by "frustration"? If that's true, isn't that merely stating the obvious, at least for the observed microphase separation? In general, does "frustration" always mean deviation of actual, physical molecular configurations from certain imagined/hypothetical/reference molecular configurations, and therefore dependent upon the choice of the imagined reference configuration? If this is how the authors apply the term "frustration" in the present work, what is the zero-frustration reference state/configuration for microphase separation? And, similarly, what is the zero-frustration reference state/configuration when frustrated EPS-water interactions are discussed (~p.14-p.15, Fig.5)? How do non-frustrated water-protein interactions look like? Is the classic clathrate-like organization of water hydrogen bonds around small nonpolar solute "frustrated"?

(3) In the discussion about the correlation of various transfer free energy scales for amino acids and Urry's critical onset temperature (ref.41) on p.11 and Fig.4, is there any theoretical relationship to be expected between the interactions among amino acids of ELPs and their critical onset temperatures? While a certain correlation may be intuitively expected if the free energy scale "is working", is there any theoretical insight into the mathematical form of this relationship? A clarifying discussion is needed because it bears logically on whether the observed correlation or lack thereof for different transfer energy scales is a good indication of the adequacy of the energy scales in describing the actual physical interactions at play. This question requires some prior knowledge of the expected mathematical relationship between interaction parameters and onset temperature.

(4) To provide a more comprehensive context for the present study, it is useful to compare the microphase separation seen in the authors' simulation with the micelle-like structures observed in recent simulated condensed/aggregated states of hydrophobic-polar (HP) model sequences in Statt et al., J Chem Phys 152, 075101 (2020) [see esp. Fig.6] and Wessén et al., J Phys Chem B 126, 9222-9245 (2022) [see, e.g., Fig.10].

(5) "Gyroid-like morphology" is mentioned several times in the manuscript (p.4, p.8, p.17, Fig.S3). This is apparently an interesting observation but a clear explanation is lacking. A more detailed and specific discussion, perhaps with additional graphical presentations, should be provided to demonstrate why the simulated condensed-phase ELP configurations are similar to the classical description of gyroid as in, e.g., Terrones & Mackay, Chem Phys Lett 207, 45-50 (1993) and Lambert et al., Phil Trans R Soc A 354, 2009-2023 (1996).

Comments on the revised manuscript:

The authors have adequately addressed my previous concerns.

Reviewer #2 (Public Review):

Summary:

Latham A.P. et al. apply simulations and FLIM to analyse several di-block elastin-like polypetides and connect their sequence to the micro-structure of coacervates resulting from their phase-separation.

Strengths:

Understanding the molecular grammar of phase separating proteins and the connection with mesoscale properties of the coacervates is highly relevant. This work provides insights into micro-structures of coacervates resulting from di-block polypetides.

Weaknesses:

The results apply to a very specific architecture (di-block polypetides) with specific sequences.

eLife assessment

This study address a fundamental question: Do lipid rafts play a role in trafficking in the secretory pathway? By performing carefully controlled experiments with synthetic membrane proteins derived from the transmembrane region of LAT, the authors describe, model and quantify the importance of transmembrane domains in the kinetics of trafficking of a protein through the cell, from the ER to the cell surface via the Golgi. While their findings are solid, further experiments that relate to the existence and nature of domains at the TGN are necessary to provide a direct connection between the phase partitioning capability of the transmembrane regions of membrane proteins and the sorting potential of rafts.

Author Response

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

eLife assessment

This valuable study advances our understanding of the brain nuclei involved in rapid-eye movement (REM) sleep regulation. Using a combination of imaging, electrophysiology, and optogenetic tools, the study provides convincing evidence that inhibitory neurons in the preoptic area of the hypothalamus influence REM sleep. This work will be of interest to neurobiologists working on sleep and/or brain circuitry.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This paper identifies GABA cells in the preoptic hypothalamus which are involved in REM sleep rebound (the increase in REM sleep) after selective REM sleep deprivation. By calcium photometry, these cells are most active during REM, and show more claim signals during REM deprivation, suggesting they respond to "REM pressure". Inhibiting these cells ontogenetically diminishes REM sleep. The optogenetic and photometry work is carried out to a high standard, the paper is well-written, and the findings are interesting.

We thank the reviewer for the detailed feedback and thoughtful comments on how to improve our manuscript. To address the reviewer’s concerns, we revised our discussion and added new data. Below, we address the concerns point by point.

Points that could be addressed or discussed:

(1) The circuit mechanism for REM rebound is not defined. How do the authors see REM rebound as working from the POAGAD2 cells? Although the POAGAD2 does project to the TMN, the actual REM rebound could be mediated by a projection of these cells elsewhere. This could be discussed.

We demonstrate thatPOA GAD2→TMN cells become more frequently activated as the pressure for REMs builds up, whereas inhibiting these neurons during high REMs pressure leads to a suppression of the REMs rebound. It is not known how POA GAD2→TMN cells encodeincreased REMs pressure and subsequently influence the REMs rebound. REMsdeprivation wasshown to changethe intrinsic excitabilityof hippocampal neurons and impact synaptic plasticity (McDermott et al., 2003; Mallick and Singh, 2011 ; Zhou et al., 2020) . We speculate that increasedREMs pressure leads to an increase in the excitabilityof POA->TMN neurons, reflected inthe increased number ofcalcium peaks. The increased excitability of POA GAD2→TMN neurons in turn likely leads to stronger inhibition of downstream REM-off neurons. Consequently, as soon as REMsdeprivation stops, there is an increased chance for enteringREMs. The time coursefor how long it takes till the POA excitability resettles toits baseline consequently sets a permissive time window for increasedamounts of REMs to recover its lostamount. For future studies, it would be interesting to map how quickly the excitability ofPOA neurons increases or decays as afunction of the lost or recovered amount of REMs andunravel the cellularmechanisms underlying the elevated activity of POAGAD2 →TMN neurons during highREMs pressure, e.g., whether changes in the expression of ion channels contribute to increasedexcitability of these neurons (Donlea et al., 2014) . As we mentioned in the Discussion, the POAalso projects to other REMs regulatorybrain regions such as the vlPAG and LH. Therefore, it remains to be tested whether POA GAD2 →TMN neurons also innervate these brain regions to potentially regulate REMs homeostasis. We explicitly state this now in the revised Discussion.

(2) The "POAGAD2 to TMN" name for these cells is somewhat confusing. The authors chose this name because they approach the POAGAD2 cells via retrograde AAV labelling (rAAV injected into the TMN). However, the name also seems to imply that neurons (perhaps histamine neurons) in the TMN are involved in the REM rebound, but there is no evidence in the paper that this is the case. Although it is nice to see from the photometry studies that the histamine cells are selectively more active (as expected) in NREM sleep (Fig. S2), I could not logically see how this was a relevant finding to REM rebound or the subject of the paper. There are many other types of cells in the TMN area, not just histamine cells, so are the authors suggesting that these non-histamine cells in the TMN could be involved?

We acknowledge that other types of neurons in the TMN may also be involved in the REMs rebound, and therefore inhibition of histamine neurons by POA GAD2 →TMN neurons may not be the sole source of the observed effect. To stress that other neurons within the TMN and/or brain regions may also contribute to the REMs rebound, we have revised the Results section.

We performed complementary optogenetic inhibition experiments of TMN HIS neurons to investigate if suppression of these neurons is sufficient to promote REMs. We foundthat SwiChR++ mediated inhibition of TMNHIS neurons increased theamount of REMs compared withrecordings without laser stimulation in the same mice and eYFPmice withlaser stimulation. Thus, while TMN HIS neurons may not bethe only downstream target of GABAergic POA neurons, these data suggest that they contribute to REMs regulation. We have incorporated these results in Fig. S4 .

We further investigated whether the activity of TMN HIS neurons changes between two REMs episodes. Assumingthat REMs pressure inhibits the activity ofREM-off histamine neurons,their firing rates should behighest right after REMs ends when REMs pressure is lowest, and progressivelydecay throughout the inter-REM interval, and reach their lowest activity right before the onset of REMs ( Park et al., 2021) , similarto the activity profile observed for vlPAG REM-off neurons (Weber et al., 2018).We indeed found that TMNHIS neurons displaya gradual decrease in their activity throughout theinter-REM interval and thus potentially reflect the build up of REM pressure ( Fig. S2F ).

(3) It is a puzzle why most of the neurons in the POA seem to have their highest activity in REM, as also found by Miracca et al 2022, yet presumably some of these cells are going to be involved in NREM sleep as well. Could the same POAGAD2-TMN cells identified by the authors also be involved in inducing NREM sleep-inhibiting histamine neurons (Chung et al). And some of these POA cells will also be involved in NREM sleep homeostasis (e.g. Ma et al Curr Biol)? Is NREM sleep rebound necessary before getting REM sleep rebound? Indeed, can these two things (NREM and REM sleep rebound) be separated?

Previous studies have demonstrated that POA GABAergic neurons, including those projecting to the TMN, are involved in NREMs homeostasis (Sherin et al., 1998; Gong et al., 2004; Ma et al., 2019) . Therefore, we predict that POA neurons that are involved in NREMs homeostasis are a subset of POA GAD2 → TMN neurons in our manuscript.

Using optrode recordings in the POA, we recently reported that 12.4% of neurons sampled have higher activity during NREMs compared with REMs; in contrast, 43.8% of neurons sampled have the highest activity during REMs compared with NREMs (Antila et al., 2022) indicating that the proportion of NREM max neurons is smaller compared with REM max neurons. These proportions of neurons are in agreement with previous results (Takahashi et al., 2009) . Considering fiber photometry monitors the average activity of a population of neurons as opposed to individual neurons, it is possible that we recorded neural activity across heterogeneous populations and therefore our findings may disguise the neural activity of the low proportion of NREMs neurons. We previously reported thespiking activity of POA GAD2 →TMN neurons at the singlecell level (Chung et al., 2017) . We have noted in themanuscript thatwhile the activity ofPOA GAD2→TMN neurons is highestduring REMs, theneural activity increases at NREMs → REMs transitions indicating these neurons also areactive during NREMs.

Using our REMs restriction protocol, we selectively restricted REMs leading to the subsequent rebound of REMs without affecting NREMs and consequently we did not find an increase in the amount of NREMs during the rebound or an increase in slow-wave activity, a key characteristic of sleep rebound that gradually dissipates during recovery sleep (Blake and Gerard, 1937; Williams et al., 1964; Rosa and Bonnet, 1985; Dijk et al., 1990; Neckelmann and Ursin, 1993; Ferrara et al., 1999) . However, during total sleep deprivation when subjects are deprived of both NREMs and REMs, isolating NREMs and REMs rebound may not be attainable.

(4) Is it possible to narrow down the POA area where the GAD2 cells are located more precisely?

POA can be subdivided into anatomically distinct regions such as medial preoptic area, median preoptic area, ventrolateral preoptic area, and lateral preoptic area (MPO, MPN, VLPO, and LPO respectively). To quantify where the virus expressing GAD2 cells and optic fibers are located within the POA, we overlaid the POA coronal reference images (with red boundaries denoting these anatomically distinct regions) over the virus heat maps and optic fiber tracts from datasets used in Figure 1A. We found that virus expression and optic fiber tracts were located in the ventrolateral POA, lateral POA, and the lateral part of medial POA, and included this description in the text.

Author response image 1.

Location of virus expression (A) and optic fiber placement (B) within subregions of POA.

(5) It would be ideal to further characterize these particular GAD2 cells by RT-PCR or RNA seq. Which other markers do they express?

Single-cell RNA-sequencing of POA neurons has revealed an enormous level of molecular diversity, consisting of nearly 70 subpopulations based on gene expression of which 43 can be clustered into inhibitory neurons (Moffitt et al., 2018) . One of the most studied subpopulation of POA sleep-active neurons contains the inhibitory neuropeptide galanin (Sherin et al., 1998; Gaus et al., 2002; Chung et al., 2017; Kroeger et al., 2018; Ma et al., 2019; Miracca et al., 2022) . Galanin neurons have been demonstrated to innervate the TMN (Sherin et al., 1998) yet, within the galanin neurons 7 distinct clusters exist based on unique gene expression (Moffitt et al., 2018) . In addition to galanin, we have previously performed single-cell RNA-seq on POA GAD2 → TMN neurons and identified additional neuropeptides such as cholecystokinin (CCK), corticotropin-releasing hormone (CRH), prodynorphin (PDYN), and tachykinin 1 (TAC1) as subpopulations of GABAergic POA sleep-active neurons (Chung et al., 2017; Smith et al., 2023) . Like galanin, these neuropeptides can also be divided into multiple subtypes as well (Chen et al., 2017; Moffitt et al., 2018) . Thus while these molecular markers for POA neurons are immensely diverse, we agree that characterizing the molecular identity of POA GAD2 → TMN neurons and investigating the functional relevance of these neuropeptides in the context of REMs homeostasis would enrich our understanding of a neural circuit involved in REMs homeostasis and can stand as a separate extension of this manuscript.

Reviewer #2 (Public Review):

Maurer et al investigated the contribution of GAD2+ neurons in the preoptic area (POA), projecting to the tuberomammillary nucleus (TMN), to REM sleep regulation. They applied an elegant design to monitor and manipulate the activity of this specific group of neurons: a GAD2-Cre mouse, injected with retrograde AAV constructs in the TMN, thereby presumably only targeting GAD2+ cells projecting to the TMN. Using this set-up in combination with technically challenging techniques including EEG with photometry and REM sleep deprivation, the authors found that this cell-type studied becomes active shortly (≈40sec) prior to entering REM sleep and remains active during REM sleep. Moreover, optogenetic inhibition of GAD2+ cells inhibits REM sleep by a third and also impairs the rebound in REM sleep in the following hour. Despite a few reservations or details that would benefit from further clarification (outlined below), the data makes a convincing case for the role of GAD2+ neurons in the POA projecting to the TMN in REM sleep regulation.

We thank the reviewer for the thorough assessment of our study and supportive comments. We have addressed your concerns in the revised manuscript, and our point by point response is provided below.

The authors found that optogenetic inhibition of GAD2+ cells suppressed REM sleep in the hour following the inhibition (e.g. Fig2 and Fig4). If the authors have the data available, it would be important to include the subsequent hours in the rebound time (e.g. from ZT8.5 to ZT24) to test whether REM sleep rebound remains impaired, or recovers, albeit with a delay.

We thank the reviewer for this comment and agree that it would be interesting to know how REMs changes for a longer period of time throughout the rebound phase. For Fig. 2, we did not record the subsequent hours. For Fig 4, we recorded the subsequent rebound between ZT7.5 and 10.5. When we compare the REMs amount during this 4 hr interval, the SwiChR mice have less REMs compared with eYFP mice with marginal significance (unpaired t-test, p=0.0641). We also plotted the cumulative REMs amount during restriction and rebound phases, and found that the cumulative amount of REMs was still lower in SwiChR mice than eYFP mice at ZT 10.5 (Author response image 2). Therefore, it will be interesting to record for a longer period of time to test when the SwiChR mice compensate for all the REMs that was lost during the restriction period.

Author response image 2.

Cumulative amount of REMs during REMs deprivation and rebound combined with optogenetic stimulation in eYFP and SwiChR groups. This data is shown as bar graphs in Figure 4.

REM sleep is under tight circadian control (e.g. Wurts et al., 2000 in rats; Dijk, Czeisler 1995 in humans). To contextualize the results, it would be important to mention that it is not clear if the role of the manipulated neurons in REM sleep regulation hold at other circadian times of the day.

Author response image 3.

Inhibiting POA GAD2→ TMN neurons at ZT5-8 reduces REMs. (A) Schematic of optogenetic inhibition experiments. (B) Percentage of time spent in REMs, NREMs and wakefulness with laser in SwiChR++ and eYFP mice. Unpaired t-tests, p = 0.0013, 0.0469 for REMs and wakeamount. (C) Duration of REMs, NREMs, and wake episodes. Unpaired t-tests, p = 0.0113 for NREMs duration. (D) Frequency of REMs, NREMs, and wake episodes. Unpaired t-tests, p = 0.0063, 0.0382 for REMs and NREMs frequency.

REMs propensity is largest towards the end of the light phase (Czeisler et al., 1980; Dijk and Czeisler, 1995; Wurts and Edgar, 2000). As a control, we therefore performed the optogenetic inhibition experiments of POA GAD2→TMN neurons during ZT5-8 (Author response image 3). Similar to our results in Figure 2, we found that SwiChR-mediated inhibition of POA GAD2 →TMN neurons attenuated REMs compared with eYFP laser sessions. These findings suggest our results are consistentat other circadian times of the day.

The effect size of the REM sleep deprivation using the vibrating motor method is unclear. In FigS4-D, the experimental mice reduce their REM sleep to 3% whereas the control mice spend 6% in REM sleep. In Fig4, mice are either subjected to REM sleep deprivation with the vibrating motor (controls), or REM sleep deprivations + optogenetics (experimental mice).

The control mice (vibrating motor) in Fig4 spend 6% of their time in REM sleep, which is double the amount of REM sleep compared to the mice receiving the same treatment in FigS4-D. Can the authors clarify the origin of this difference in the text?

The effect size for REM sleep deprivation is now added in the text.

It is important to note that these figures are analyzing two different intervals of the REMs restriction. In Fig. S4D, we analyzed the total amount of REMs over the entire 6 hr restriction interval (ZT1.5-7.5). In Fig. 4, we analyzed the amount of REMs only during the last 3 hr of restriction (ZT4.5-7.5) as optogenetic inhibition was performed only during the last 3 hrs when the REMs pressure is high. In Fig. S4D, we looked at the amount of REMs during ZT1.5-4.5 and 4.5-7.5 and found that the amount of REMs during ZT4.5-7.5 (4.46 ± 0.25 %; mean ± s.e.m.) is indeed higher than ZT 1.5-4.5 (1.66 ± 0.62 %), and is comparable to the amount of REMs during ZT4.5-7.5 in eYFP mice (5.95 ± 0.52 %) in Fig. 4. We now clearly state in the manuscript at which time points we analyzed the amount, duration and frequency of REMs.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) A few further citations suggested: Discussion "The TMN contains histamine producing neurons and antagonizing histamine neurons causes sleepiness..." It would be appropriate to cite Uygun DS et al 2016 J Neurosci (PMID: 27807161) here. Using the same HDC-Cre mice as used by Maurer et al., Uygun et al found that selectively increasing GABAergic inhibition onto histamine neurons produced NREM sleep.

We apologize for omitting this important paper. In the revised manuscript, we added this citation.

(2) Materials and Methods.

Although the JAX numbers are given for the mouse lines based on researchers generously donating to JAX for others to use, please cite the papers corresponding to the GAD2-ires-Cre and HDC-ires-Cre mouse lines deposited at JAX.

GAD2-ires-Cre was described in Taniguchi H et al., 2011, Neuron (PMID: 21943598).

The construction of the HDC-ires-CRE line is described in Zecharia AY et al J Neurosci et al 2012 (PMID: 22993424).

We have now added these important citations in the revised manuscript.

(3) Similarly, for the viruses, please provide the citations for the AAV constructs that were donated to Addgene.

We have now added these citations in the revised manuscript.

Reviewer #2 (Recommendations For The Authors):

The authors rely heavily on their conclusions by using an optogenetic tool that inhibits the activity of GAD2+ neurons, however, it is not shown that these neurons are indeed inhibited as expected. An alternative approach to tackle this could be the application of a different technique to achieve the same output (e.g. chemogenetics). However, both experiments (confirmation of inhibition, or using a different technique) would require a significant amount of work, and given the numerous studies out there showing that these optogenetic tools tend to work, may not be necessary. Hence the authors could also cite a similar study that used a likewise construct and where it was indeed shown that this technique works (i.e. similar retrograde optogenetic construct with Cre depedendent expression combined with electrophysiological recordings).

This laser stimulation protocol was designed based on previous reports of sustained inhibition using the same inhibitory opsin and our prior results that recapitulate similar findings as inhibitory chemogenetic techniques (Iyer et al., 2016; Kim et al., 2016; Wiegert et al., 2017; Stucynski et al., 2022). We have now added this description in the Result section.

Fig1A - Right: the virus expression graphs are great and give a helpful insight into the variability. The image on the left (GCAMP+ cells) is less clear, the GCAMP+ cells don't differentiate well from the background. Perhaps the whole brain image with inset in POA can show the GCAMP expression more convincingly.

We have added a histology picture showing the whole brain image with inset in the POA in the updated Fig. 1A .

Statistics: The table is very helpful. Based on the degrees of freedom, it seems that in some instances the stats are run on the recordings rather than on the individual mice (e.g. Fig1). It could be considered to use a mixed model where subjects as taken into account as a factor.

Author response image 4.

ΔF/Factivity of POA GAD2→TMN neurons during NREMs. The duration of NREMs episodes was normalized in time, ranging from 0 to 100%. Shading, ± s.e.m. Pairwise t-tests with Holm-Bonferroni correctionp = 5.34 e-4 between80 and100. Graybar, intervals where ΔF/F activity was significantly different from baseline (0 to 20%, the first time bin). n = 10 mice. In Fig. 1E , we ran stats based on the recordings. In this data set, we ran stats based on the individual mice, and found that the activity also gradually increased throughout NREMs episodes.

There is an effect of laser in Fig2 on REM sleep amount, as well as an interaction effect with virus injection (from the table). Therefore, it would be helpful for the reader to also show REM sleep data from the control group (laser stimulation but no active optogenetics construct) in Fig 2.

To properly control laser and virus effect, we performed the same laser stimulation experiments in eYFP control mice (expressing only eYFP without optogenetic construct, SwiChR++) and the data is provided in Fig 2C .

Fig3B: At the start of the rebound of REM sleep, there is a massive amount of wakefulness, also reflected in the change of spectral composition. Could you comment on the text about what is happening here?

We quantified the amount of wakefulness during the first hour of REMs rebound and found that indeed there is no significant difference in wakefulness between REM restriction and baseline control conditions ( Fig. S4H ). Therefore, while the representative image in Fig 3B shows increased wakefulness at the beginning of REMs rebound, we do not think the overall amount of wakefulness is increased.

Fig 4, supplementary data: it would be helpful for the reader to have mentioned in the text the effect size of the REM sleep restriction protocol (e.g. mean and standard deviation).

Thank you for this suggestion. We have now added the effect size for the REM sleep restriction experiments in the main text.

REM sleep restriction and photometry experiment: could be improved by adding within the main body of text that, in order to conduct the photometry experiment in the last hours of REM sleep deprivation, the manual REM sleep deprivation had to be applied, because the vibrating motor technique disturbed the photometry recordings.

Thank you for this suggestion. We have added the description in the main text.

Suggestion to build further on the already existing data (not for this paper): you have a powerful dataset to test whether REM sleep pressure builds up during wakefulness or NREM sleep, by correlating when your optogenetic treatment occurs (NREM or wakefulness), with the subsequent rebound in REM sleep (see also Endo et al., 1998; Benington and Heller, 1994; Franken 2001).

We thank the reviewer for this excellent suggestion. We plan to carry out this experiment in the future.

References

Antila, H., Kwak, I., Choi, A., Pisciotti, A., Covarrubias, I., Baik, J., et al. (2022). A noradrenergic-hypothalamic neural substrate for stress-induced sleep disturbances. Proc. Natl. Acad. Sci. 119, e2123528119. doi: 10.1073/pnas.2123528119.

Blake, H., and Gerard, R. W. (1937). Brain potentials during sleep. Am. J. Physiol.-Leg. Content 119, 692–703. doi: 10.1152/ajplegacy.1937.119.4.692.

Chen, R., Wu, X., Jiang, L., and Zhang, Y. (2017). Single-Cell RNA-Seq Reveals Hypothalamic Cell Diversity. Cell Rep. 18, 3227–3241. doi: 10.1016/j.celrep.2017.03.004.

Chung, S., Weber, F., Zhong, P., Tan, C. L., Nguyen, T., Beier, K. T., et al. (2017). Identification of Preoptic Sleep Neurons Using Retrograde Labeling and Gene Profiling. Nature 545, 477–481. doi: 10.1038/nature22350.

Czeisler, C. A., Zimmerman, J. C., Ronda, J. M., Moore-Ede, M. C., and Weitzman, E. D. (1980). Timing of REM sleep is coupled to the circadian rhythm of body temperature in man. Sleep 2, 329–346.

Dijk, D. J., Brunner, D. P., Beersma, D. G., and Borbély, A. A. (1990). Electroencephalogram power density and slow wave sleep as a function of prior waking and circadian phase. Sleep 13, 430–440. doi: 10.1093/sleep/13.5.430.

Dijk, D. J., and Czeisler, C. A. (1995). Contribution of the circadian pacemaker and the sleep homeostat to sleep propensity, sleep structure, electroencephalographic slow waves, and sleep spindle activity in humans. J. Neurosci. Off. J. Soc. Neurosci. 15, 3526–3538. doi: 10.1523/JNEUROSCI.15-05-03526.1995.

Donlea, J. M., Pimentel, D., and Miesenböck, G. (2014). Neuronal machinery of sleep homeostasis in Drosophila. Neuron 81, 860–872. doi: 10.1016/j.neuron.2013.12.013.

Ferrara, M., De Gennaro, L., Casagrande, M., and Bertini, M. (1999). Auditory arousal thresholds after selective slow-wave sleep deprivation. Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol. 110, 2148–2152. doi: 10.1016/s1388-2457(99)00171-6.

Gaus, S. E., Strecker, R. E., Tate, B. A., Parker, R. A., and Saper, C. B. (2002). Ventrolateral preoptic nucleus contains sleep-active, galaninergic neurons in multiple mammalian species. Neuroscience 115, 285–294. doi: 10.1016/S0306-4522(02)00308-1.

Gong, H., McGinty, D., Guzman-Marin, R., Chew, K.-T., Stewart, D., and Szymusiak, R. (2004). Activation of c-fos in GABAergic neurones in the preoptic area during sleep and in response to sleep deprivation. J. Physiol. 556, 935–946. doi: 10.1113/jphysiol.2003.056622.

Iyer, S. M., Vesuna, S., Ramakrishnan, C., Huynh, K., Young, S., Berndt, A., et al. (2016). Optogenetic and chemogenetic strategies for sustained inhibition of pain. Sci. Rep. 6, 30570. doi: 10.1038/srep30570.

Kim, H., Ährlund-Richter, S., Wang, X., Deisseroth, K., and Carlén, M. (2016). Prefrontal Parvalbumin Neurons in Control of Attention. Cell 164, 208–218. doi: 10.1016/j.cell.2015.11.038.

Kroeger, D., Absi, G., Gagliardi, C., Bandaru, S. S., Madara, J. C., Ferrari, L. L., et al. (2018). Galanin neurons in the ventrolateral preoptic area promote sleep and heat loss in mice. Nat. Commun. 9, 4129. doi: 10.1038/s41467-018-06590-7.

Ma, Y., Miracca, G., Yu, X., Harding, E. C., Miao, A., Yustos, R., et al. (2019). Galanin Neurons Unite Sleep Homeostasis and α2-Adrenergic Sedation. Curr. Biol. CB 29, 3315-3322.e3. doi: 10.1016/j.cub.2019.07.087.

Mallick, B. N., and Singh, A. (2011). REM sleep loss increases brain excitability: role of noradrenaline and its mechanism of action. Sleep Med. Rev. 15, 165–178. doi: 10.1016/j.smrv.2010.11.001.

McDermott, C. M., LaHoste, G. J., Chen, C., Musto, A., Bazan, N. G., and Magee, J. C. (2003). Sleep deprivation causes behavioral, synaptic, and membrane excitability alterations in hippocampal neurons. J. Neurosci. Off. J. Soc. Neurosci. 23, 9687–9695. doi: 10.1523/JNEUROSCI.23-29-09687.2003.

Miracca, G., Anuncibay-Soto, B., Tossell, K., Yustos, R., Vyssotski, A. L., Franks, N. P., et al. (2022). NMDA Receptors in the Lateral Preoptic Hypothalamus Are Essential for Sustaining NREM and REM Sleep. J. Neurosci. 42, 5389–5409. doi: 10.1523/JNEUROSCI.0350-21.2022.

Moffitt, J. R., Bambah-Mukku, D., Eichhorn, S. W., Vaughn, E., Shekhar, K., Perez, J. D., et al. (2018). Molecular, spatial, and functional single-cell profiling of the hypothalamic preoptic region. Science 362. doi: 10.1126/science.aau5324.

Neckelmann, D., and Ursin, R. (1993). Sleep stages and EEG power spectrum in relation to acoustical stimulus arousal threshold in the rat. Sleep 16, 467–477.

Park, S.-H., Baik, J., Hong, J., Antila, H., Kurland, B., Chung, S., et al. (2021). A probabilistic model for the ultradian timing of REM sleep in mice. PLOS Comput. Biol. 17, e1009316. doi: 10.1371/journal.pcbi.1009316.

Rosa, R. R., and Bonnet, M. H. (1985). Sleep stages, auditory arousal threshold, and body temperature as predictors of behavior upon awakening. Int. J. Neurosci. 27, 73–83. doi: 10.3109/00207458509149136.

Sherin, J. E., Elmquist, J. K., Torrealba, F., and Saper, C. B. (1998). Innervation of histaminergic tuberomammillary neurons by GABAergic and galaninergic neurons in the ventrolateral preoptic nucleus of the rat. J. Neurosci. Off. J. Soc. Neurosci. 18, 4705–4721.

Smith, J., Honig-Frand, A., Antila, H., Choi, A., Kim, H., Beier, K. T., et al. (2023). Regulation of stress-induced sleep fragmentation by preoptic glutamatergic neurons. Curr. Biol. CB , S0960-9822(23)01585–3. doi: 10.1016/j.cub.2023.11.035.

Stucynski, J. A., Schott, A. L., Baik, J., Chung, S., and Weber, F. (2022). Regulation of REM sleep by inhibitory neurons in the dorsomedial medulla. Curr. Biol. CB 32, 37-50.e6. doi: 10.1016/j.cub.2021.10.030.

Takahashi, K., Lin, J.-S., and Sakai, K. (2009). Characterization and mapping of sleep-waking specific neurons in the basal forebrain and preoptic hypothalamus in mice. Neuroscience 161, 269–292. doi: 10.1016/j.neuroscience.2009.02.075.

Weber, F., Hoang Do, J. P., Chung, S., Beier, K. T., Bikov, M., Saffari Doost, M., et al. (2018). Regulation of REM and Non-REM sleep by periaqueductal GABAergic neurons. Nat. Commun. 9, 1–13. doi: 10.1038/s41467-017-02765-w.

Wiegert, J. S., Mahn, M., Prigge, M., Printz, Y., and Yizhar, O. (2017). Silencing Neurons: Tools, Applications, and Experimental Constraints. Neuron 95, 504–529. doi: 10.1016/j.neuron.2017.06.050.

Williams, H. L., Hammack, J. T., Daly, R. L., Dement, W. C., and Lubin, A. (1964). RESPONSES TO AUDITORY STIMULATION, SLEEP LOSS AND THE EEG STAGES OF SLEEP. Electroencephalogr. Clin. Neurophysiol. 16, 269–279. doi: 10.1016/0013-4694(64)90109-9.

Wurts, S. W., and Edgar, D. M. (2000). Circadian and homeostatic control of rapid eye movement (REM) sleep: promotion of REM tendency by the suprachiasmatic nucleus. J. Neurosci. Off. J. Soc. Neurosci. 20, 4300–4310. doi: 10.1523/JNEUROSCI.20-11-04300.2000.

Zhou, Y., Lai, C. S. W., Bai, Y., Li, W., Zhao, R., Yang, G., et al. (2020). REM sleep promotes experience-dependent dendritic spine elimination in the mouse cortex. Nat. Commun. 11, 4819. doi: 10.1038/s41467-020-18592-5.

Author response:

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

Public Reviews

Reviewer #1 (Public Review):

Summary:

Dormancy/diapause/hibernation (depending on how the terms are defined) is a key life history strategy that allows the temporal escape from unfavorable conditions. Although environmental conditions do play a major role in inducing and terminating dormancy (authors call this energy limitation hypothesis), the authors test a mutually non-exclusive hypothesis (life-history hypothesis) that sex-specific selection pressures, at least to some extent, would further shape the timing of these life-history events. Authors use a metanalytic approach to collect data (mainly on rodents) on various life-history traits to test trade-offs among these traits between sexes and how they affect entry and termination of dormancy.

Strengths:

I found the theoretical background in the Introduction quite interesting, to the point and the arguments were well-placed. How sex-specific selection pressures would drive entry and termination of diapause in insects (e.g. protandry), especially in temperate butterflies, is very well investigated. Authors attempt to extend these ideas to endotherms and trying to find general patterns across ectotherms and endotherms is particularly exciting. This work and similar evidence could make a great contribution to the life-history theory, specifically understanding factors that drive the regulation of life cycle timing.

Weaknesses:

(1) I felt that including 'ectotherms' in the title is a bit misleading as there is hardly (in fact any?) any data presented on ectotherms. Also, most of the focus of the discussion is heavily mammal (rodent) focussed. I believe saying endotherms in the title as well is a bit misleading as the data is mammalfocused.

We change the title to : "Evolutionnary trade-offs in dormancy phenology". This is a hybrid article comprising both a meta-analysis and a literature review. Each of these parts brings new elements to the hypotheses presented. The statistical analyses only concern mammals and especially rodent species. But the literature review highlighted links between the evolution of dormancy in ectotherms and endotherms that have not been linked in previous studies. We feel it is important for readers to know that much of the discussion will focus on the comparison of these two groups. But we understand that placing the term ectotherms in the title might suggest a meta-analysis including these two groups.

In addition, we indicated more specifically in the abstract and at the end of the introduction that the article includes two approaches associated with different groups of animals.

We also specified in the section « review criteria » that:

Only one bird species is considered to be a hibernator, and no information is available on sex differences in hibernation phenology (Woods and Brigham 2004, Woods et al. 2019).

We have also added a "study limitations" section, which explains that although the meta-analysis is limited by the data available in the literature, the information available for the species groups not studied seems to support our results.

(2) I think more information needs to be provided early on to make readers aware of the diversity of animals included in the study and their geographic distribution. Are they mostly temperate or tropical? What is the span of the latitude as day length can have a major influence on dormancy timings? I think it is important to point out that data is more rodent-centric. Along the line of this point, is there a reason why the extensively studied species like the Red Deer or Soay Sheep and other well-studied temperate mammals did not make it into the list?

We specified in the abstract and at the end of the introduction that the species studied in the metaanalysis are mainly Holarctic species. We have also added a map showing all the study sites used in the meta-analysis. Finally, we've noted in the methods and added a "study limitation" section at the end of the discussion an explanation for those species that were not studied in the meta-analysis and the consequences for the interpretation of results

The hypotheses developed in this article are based on the survival benefits of seasonal dormancy thanks to a period of complete inactivity lasting several months. The Red Deer or Soay Sheep remain active above ground throughout the year.

The effect of photoperiod on phenology is one of the mechanisms that has evolved to match an activity with the favorable condition. In this study, we are not interested in the mechanisms but in the evolutionary pressures that explain the observed phenology. Interspecific variation in the effect of photoperiod results from different evolutionary pressures, which we are trying to highlight. It is therefore not necessary to review mechanisms and effects of photoperiod, themselves requiring a lengthy review.

We also tested the “physiological constraint hypothesis” on several variables. Temperature and precipitation are factors correlated with sex differences in phenology of hibernation. These factors allow consideration of the geographical differences that influence hibernation phenology.

(3) Isn't the term 'energy limitation hypothesis' which is used throughout the manuscript a bit endotherm-centric? Especially if the goal is to draw generalities across ectotherms and endotherms. Moreover, climate (e.g. interaction of photoperiod and temperature in temperatures) most often induces or terminates diapause/dormancy in ectotherms so I am not sure if saying 'energy limitation hypothesis' is general enough.

We renamed this hypothesis the "physiological constraint hypothesis" and we have made appropriate changes in the text so as not to focus physiological constraints solely on energy aspects.

(4) Since for some species, the data is averaged across studies to get species-level trait estimates, is there a scope to examine within population differences (e.g. across latitudes)? This may further strengthen the evidence and rule out the possibility of the environment, especially the length of the breeding season, affecting the timing of emergence and immergence.

For a given species, data on hibernation phenology are averaged for different populations, but also for the same population when measurements are taken over several years. To test these hypotheses on a population scale, precise data on reproductive effort would be needed for each population tested, but this concerns very few species (less than 5).

Testing the effects of temperature and precipitation allows us to take into account the effects of climate on phenology.

(5) Although the authors are looking at the broader patterns, I felt like the overall ecology of the species (habitat, tropical or temperate, number of broods, etc.) is overlooked and could act as confounding factors.

Yes, that's why we also tested the physiological constraints hypothesis, including the effect of temperature and precipitation. For the life-history hypothesis, we also tested reproductive effort, which takes into account the number of offspring per year.

(6) I strongly think the data analysis part needs more clarity. As of now, it is difficult for me to visualize all the fitted models (despite Table 1), and the large number of life-history traits adds to this complexity. I would recommend explicitly writing down all the models in the text. Also, the Table doesn't make it clear whether interaction was allowed between the predictors or not. More information on how PGLS were fitted needs to be provided in the main text which is in the supplementary right now. I kept wondering if the authors have fit multiple models, for example, with different correlation structures or by choosing different values of lambda parameter. And, in addition to PGLS, authors are also fitting linear regressions. Can you explain clearly in the text why was this done?

To simplify the results, we reduced the number of models to just three: one for emergence and two for immergence. In place of Table 1, we have written the structure of the models used. We have added a sentence to the statistics section: “each PGLS model produces a λ parameter representing the effect of phylogeny ranging between 0 (no phylogeny effect) and 1 (covariance entirely explained by co-ancestry)”. We have tested only three PGLS models and the estimated lambda value for these models is 0.

(7) Figure 2 is unclear, and I do not understand how these three regression lines were computed. Please provide more details.

We tested new models and modified existing figures.

Reviewer #2 (Public Review):

Summary:

An article with lots of interesting ideas and questions regarding the evolution of timing of dormancy, emphasizing mammalian hibernation but also including ectotherms. The authors compare selective forces of constraints due to energy availability versus predator avoidance and requirements and consequences of reproduction in a review of between and within species (sex) differences in the seasonal timing of entry and exit from dormancy.

Strengths:

The multispecies approach including endotherms and ectotherms is ambitious. This review is rich with ideas if not in convincing conclusions.

Weaknesses:

The differences between physiological requirements for gameatogenesis between sexes that affect the timing of heterothermy and the need for euthermy during mammalian hibernator are significant issues that underlie but are under-discussed, in this contrast of selective pressures that determine seasonal timing of dormancy. Some additional discussion of the effects of rapid climate change on between and within species phenologies of dormancy would have been interesting.

Reviewer #2 (Recommendations For The Authors):

This review provides a very interesting and ambitious among and within-species comparison of the seasonal timing of entry and exit from dormancy, emphasizing literature from hibernating mammals (sans bats and bears) and with attention to ectotherms. The authors test hypotheses related to the timing of food availability (energy) versus life history considerations (requirements for reproduction, avoiding predation) while acknowledging that these are not mutually exclusive. I offer advice for clarifications and description of the limitations of the data (accuracy of emergence and immergence times), but mainly seek more emphasis for small mammalian hibernators on the contrast for requirements for significant periods of euthermy prior to the emergence in males versus females, a contrast that has energetic and timing consequences in both the active and hibernation seasons.

A consideration alluded to but not fully explained or discussed is the differences in mammals between species and sexes in the timing of what can be called ecological hibernation, which is the seasonal duration that an animal remains sequestered in its burrow or den, and heterothermic hibernation, between the beginning and end of the use of torpor. The two are not synonymous. When "emergence" is the first appearance above ground, there is a significant missing observation key to the energetic contrasts discussed in this review, that of this costly pre-emergence behavior.

To explain the difference between heterothermic hibernation and ecological hibernation, we've added a section in review Criteria from materials and methods :

“In this study, we addressed what can be called ecological hibernation, i.e. the seasonal duration that an animal remains sequestered in its burrow or den, which is assumed to be directly linked to the reduced risk of predation. In contrast, we did not consider heterothermic hibernation, which corresponds to the time between the beginning and end of the use of torpor. So when we mention hibernation, emergence or immergence, the specific reference is to ecological hibernation.”

In arctic and other ground squirrel species, males remain at high body temperatures after immerging and remaining in their burrows in the fall for several days to a week, and more consistently and importantly, males that will attempt to breed in the spring end torpor but remain constantly in their burrows for as much as one month at great expense whilst undergoing testicular growth, spermatogenesis, spemiation, and sperm capacitation, processes that require continuous euthermy. Female arctic ground squirrels and non-breeding males do not and typically enter their first torpor bout 1-2 days after immergence and first appear above ground 1-3 days after their last arousal in spring.

The weeks spent euthermic in a cold burrow in spring by males while undergoing reproductive maturation require a significant energetic investment (can equate to the cost of the previous heterothermic period) that contrasts profoundly with the pre-mating energetic investment by females.

Males cache food in their hibernacula and extend their active season in late summer/fall in order to do so and feed from these caches in spring after resuming euthermy, often emerging at body weights similar to that at immergence. Similar between-sex differences in the timing of hibernation and heterothermy occur in golden-mantled and Columbian ground squirrels and likely most other Urocitellus spp., though less well described in other species. These differences are related to life histories and requirements for male vs. female gameatogenesis and, at the same time, energetic considerations in the costs to males for remaining euthermic while undergoing spermatogenesis and the cost related to whether males undergo gonadal development being dependent on individual body mass and cache size. These issues should be better discussed in this review.

It is the time required to complete spermatogenesis, spermiation, and maturation of sperm not the time for growth of different sizes of testes that drives the preparation time for males. This is relatively constant among rodents. I challenge the assumption that larger testes take longer to grow than smaller ones.

We took this comment into account. As we found little evidence of an increase in testicular maturation time with relative testicular size (apart from table 4 in Kenagy and Trombulak, 1986), we no longer tested the effect of relative testicular size on protandry.

We examined whether the ability to store food before hibernation might reduce protandry. Although food storage in the burrow may be favored for overcoming harsh environments or predation, model selection did not retain the food-storing factor. Thus, the ability to accumulate food in the burrow was not by itself likely to keep males of some species from emerging earlier (e.g. Cricetus cricetus, protandry : 20 day, Siutz et al., 2016). Early emerging males may benefit from consuming higher quality food or in competition with other males (e.g., dominance assertion or territory establishment, Manno and Dobson 2008).

We developed these aspects in the discussion

While it is admirable to include ectotherms in such a broad review and modelling, I can't tell what data from how many ectothermic species contributed to the models and summary data included in the figures.

Too few data on ectotherms were available to include ectotherms in the meta-analysis

Some consideration should be made to the limitations of the data extracted from the literature of the accuracy of emergence and immergence dates when derived from only observations or trapping data. The most accurate results come from the use of telemetry for location and data logging reporting below vs. above ground positioning and body temperature.

We added a "study limits" section to the discussion to address all the limits in this commentary.

L64 "favor reproduction", better to say "allow reproduction", since there is strong evolutionary pressure to initiate reproduction early, often anticipating favorable conditions for reproduction, to maximize the time available for young to grow and prepare for overwintering themselves.

Also, generally, it is not how "harsh" an environment is but rather how short the growing season is.

We took this comment into account.

L80 More simply, individuals that have amassed sufficient energy reserves as fat and caches to survive through winter may opt to initiate dormancy. This may decrease but not obviate predation, since hibernating animals are dug from their burrows and eaten by predators such as bears and ermine.

In this sentence, we indicated a gap between dormancy phenology and the growing season, which suggests survival benefits of dormancy other than from a physiological point of view. We've changed the sentence to make it clearer : “However, some animals immerge in dormancy while environnemental conditions would allow them (from a physiological point of view) to continue their activity, suggesting other survival benefits than coping with a short growing season”

L88 other physiological or ecological factors.... (gameatogenesis).

In this study, we examine possible evolutionary pressures and therefore the environmental factors that may influence hibernation phenology. We focus on reproductive effort because, assuming predation pressure, we would expect a trade-off between survival and reproduction.

L113 beginning early to afford long active seasons to offspring while not compromising the survival of parents.

We added to the sentence:

“For females, emergence phenology may promote breeding and/or care of offspring during the most favorable annual period (e.g., a match of the peak in lactational energy demand and maximum food availability, Fig. 1) or beginning early to afford long active seasons to offspring while not compromising the survival of parents.”

L117 based on adequate preparation for overwintering and enter dormancy....

We modified the sentence as follows :

recovering from reproduction, and after acquiring adequate energy stores for overwintering”

L123 given that males outwardly invest the least time in reproduction yet generally have shorter hibernation seasons would seem to reject this hypothesis. This changes if you overtly include the time and energy that males expend while remaining euthermic preparing for hibernation, a cost that can be similar to energy expended during heterothermy.

Males invest a lot of time in reproduction before females emerge (whether for competition or physiological maturation) and some males seem to be subject to long-term negative effects linked to reproductive stress (see Millesi, E., Huber, S., Dittami, J., Hoffmann, I., & Daan, S. (1998). Parameters of mating effort and success in male European ground squirrels, Spermophilus citellus. Ethology, 104(4), 298-313). Both processes may contribute to reducing the duration of male hibernation.

L125 again, costs to support euthermy in males undergoing reproductive development is an investment in reproduction.

You're right, but it's difficult to quantify. We tested a model that takes into account the reproductive effort during reproduction and prior to reproduction. We also considered the hypothesis that species living in a cold climate might have a low protandry while having a high reproductive effort due to their ability to feed in the burrow (interaction effect between reproductive effort and temperature). We think these changes answer your comment.

L134 It isn't growing large testes that takes time, but instead completing spermatogenesis and maturation of sperm in the epdidymides.

We removed this part.

L140 Later immergence in male ground squirrels is related to accumulation and defense of cached food, activities that are related to reproduction the next spring. An experimental analysis that would be revealing is to compare immergence times in females that completed lactation to the independence of their litters vs. females that did not breed or lost their litters. Who immerges first?

Body mass variation from emergence to the end of mating in males seems to explain the delayed immergence of males in species that don't hide food in their burrows for hibernation. For example, in spermophilus citellus, males immege on average more than 3 weeks after females, yet they do not hide food in their burrows for the winter.

Such a study already exists and shows that non-breeding females immerge earlier than breeding females. We refer to it

L386: “In mammals, males and females that invest little or not at all in reproduction exhibit advances in energy reserve accumulation and earlier immergence for up to several weeks, while reproductive congeners continue activity (Neuhaus 2000, Millesi et al. 2008a).”

L164 So you examined literature from 152 species but included data from only 29 species? Did you include data from social hibernators (marmots) that mate before emergence?

With current models, we have 28 different species. We have few species because very few have data on both sex difference data and information on reproductive effort data (especially for males).

Data on sex differences in hibernation were not available for social hibernating species.

L169 Were these data from trapping or observation results? How reliable are these versus the use of information from implanted data loggers or collars that definitively document when euthermy is resumed and/or when immergence and first emergence occurs (through light loggers)?

We did not focus heterothermic hibernation, but in ecological hibernation. We have no idea of the margin of error for these types of data, but we have discussed these limitations in the "Study limitations" section.

L180, again, it is the time required to complete spermatogenesis and spermiation not the time for the growth of different sizes of testes that drives the preparation time for males. This is relatively constant among rodents. I challenge the assumption that larger testes take longer to grow than smaller ones.

We removed this part.

L200 Males that accumulate caches in fall and then feed from those during the spring pre-emergence euthermic interval and after will often be at their seasonal maximum in body mass. Declining from that peak may not be stressful.

It has been suggested that reproductive effort in Spermophilus citellus might induce long-term negative effects that delay male immergence.

Millesi, E., Huber, S., Dittami, J., Hoffmann, I., & Daan, S. (1998). Parameters of mating effort and success in male European ground squirrels, Spermophilus citellus. Ethology, 104(4), 298-313.

L210 How about altitude, which affects the length of the growing season at similar latitudes?

We extracted the location of each study site to determine the temperature and precipitation at that precise location (based on interpolated climate surface). We therefore take into account differences in growing season (based on temperature) in altitude between sites.

L267 How did whether males cache food or not figure into these comparisons? Refeeding before mating occurs during the pre-emergence euthermic interval.

We removed this part.

L332, 344 not a "proxy" but functionally related to advantages in mating systems with multiple mating males.

We removed this part.

L353 The need for a pre-emergence euthermic interval in male ground squirrels requires costs in the previous active season in accumulating and defending a cache and the proximal costs in spring while remaining at high body temperatures prior to emergence with resulting loss in body mass or devouring of the cache.

You're right, but in this section, we quickly explain the benefits of food catching compared with other species that don't do so.

L385 This review should discuss why females are not known to cache and contrast as "income breeders" from "capital breeder" males. What advantages of caches are females indifferent to (no need for a prolonged pre-emergence period) and what costs of accumulating caches do they avoid (prolonged activity period and defense of caches).

We clarified the case of female emergence.

L321 : “Thus, an early emergence of males may have evolved in response to sexual selection to accumulate energy reserve in anticipation of reproductive effort. Females, on the contrary, are not subject to intraspecific competition for reproduction and may have sufficient time before (generally one week after emergence) and during the breeding period to improve their body condition.”

L388 I don't understand the logic of the conclusion that "did not ...adequately explain the late male immergence" in this section. The greater mass loss in males over the mating period is afforded by the presence of a cache that requires later immergence.

We removed this part.

L412 Not just congeners that invest less in reproduction, but within species individuals that do not attempt to breed in one or more years and thus have no reproductive costs should be an interesting comparison for differences in phenology from individuals that do breed. Non-breeders are often yearlings but can be a significant overall proportion of males that fail to fatten or cache enough to afford a pre-emergence euthermic period.

L385: “In mammals, males and females that invest little or not at all in reproduction exhibit advances in energy reserve accumulation and earlier immergence for up to several weeks, while reproductive congeners continue activity (Neuhaus 2000, Millesi et al. 2008a).”

The sentence refers to individuals who reproduce little or not at all.

L445 Males that gain weight between emergence and mating may do so by feeding from a cache regardless of how "harsh" an environment is.

We observe this phenomenon even in species that are not known to hoard food

“Gains in body mass observed for some individuals, even in species not known to hoard food, may indicate that the environment allows a positive energy balance for other individuals with comparable energy demands.”

L492 Some insects retreat to refugia in mid-summer to avoid parasitism (Gynaephora).

Escape from parasites is also a benefit of dormancy.

Fig 1 - It is difficult to see the differences in black and green colors, esp if color blind.<br /> Maternal effort is front-loaded within the active season (line for "optimal period" shown in midseason).

Add "energy" underneath c) Prediction (H1) and "reproduction" underneath d) "Prediction (H2). Explain the orange vs black, green colors of triangles.

We made the necessary changes

Fig 2 - I don't buy the regression lines as significant in this figure. The red line, cannot have a regression with two sample points and without the left-hand most dot, nothing is significant.

We deleted this graph.

Fig 3 - females only?

We deleted this graph.

Reviewer #2 (Public Review):

Summary:

An article with lots of interesting ideas and questions regarding the evolution of timing of dormancy, emphasizing mammalian hibernation but also including ectotherms. The authors compare selective forces of constraints due to energy availability versus predator avoidance and requirements and consequences of reproduction in a review of between and within species (sex) differences in the seasonal timing of entry and exit from dormancy.

Strengths:

The multispecies approach including endotherms and ectotherms is ambitious. This review is rich with ideas if not in convincing conclusions. Limitations are discussed yet are impactful, namely that differences among and within species are contrast only for ecological hibernation (the duration of remaining sequestered) and not for "heterothermic hibernation" the period between first and last torpor. Differences between the two can have significant energetic consequences, especially for mammals returning to euthermic levels of body temperature whilst remaining in their cold burrows before emerging, eg. reproductively developing males in spring.

Weaknesses:

The differences between physiological requirements for gameatogenesis between sexes that affect the timing of heterothermy and need for euthermy during mammalian hibernator are significant issues that underlie, but are under discussed, in this contrast of selective pressures that determine seasonal timing of dormancy. Some additional discussion of the effects of rapid rapid climate change on between and within species phenologies of dormancy would have been interesting.

eLife assessment

This valuable and ambitious review examines seasonal dormancy in various species, including hibernating mammals (excluding bats and bears) and ectotherms. It provides a solid test of hypotheses on dormancy timing, considering energetic constraints and life history as alternative drivers. The review will be of interest to evolutionary biologists.

eLife assessment

This paper addresses a fundamental issue in the field of autophagy: how is a protein responsible for autophagosome-lysosome fusion recruited to mature autophagosomes but not immature ones? The work succeeds in its ambition to provide a new conceptual advance. The evidence supporting the conclusions is convincing, with fluorescence microscopy, biochemical assays, and molecular dynamics simulations. This work will be of broad interest to cell biologists and biochemists studying autophagy, and also those focusing on lipid/membrane biology.

Author Response

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

Reviewer #1 (Public Review):

In this manuscript, the authors report a molecular mechanism for recruiting syntaixn 17 (Syn17) to the closed autophagosomes through the charge interaction between enriched PI4P and the C-terminal region of Syn17. How to precisely control the location and conformation of proteins is critical for maintaining autophagic flux. Particularly, the recruitment of Syn17 to autophagosomes remains unclear. In this paper, the author describes a simple lipid-protein interaction model beyond previous studies focusing on protein-protein interactions. This represents conceptual advances.

We would like to thank Reviewer #1 for the positive evaluation of our study.

Reviewer #2 (Public Review):

Summary:

Syntaxin17 (STX17) is a SNARE protein that is recruited to mature (i.e., closed) autophagosomes, but not to immature (i.e., unclosed) ones, and mediates the autophagosome-lysosome fusion. How STX17 recognizes the mature autophagosome is an unresolved interesting question in the autophagy field. Shinoda and colleagues set out to answer this question by focusing on the C-terminal domain of STX17 and found that PI4P is a strong candidate that causes the STX17 recruitment to the autophasome.

Strengths:

The main findings are: 1) Rich positive charges in the C-terminal domain of STX17 are sufficient for the recruitment to the mature autophagosome; 2) Fluorescence charge sensors of different strengths suggest that autophagic membranes have negative charges and the charge increases as they mature; 3) Among a battery of fluorescence biosensors, only PI4P-binding biosensors distribute to the mature autophagosome; 4) STX17 bound to isolated autophagosomes is released by treatment with Sac1 phosphatase; 5) By dynamic molecular simulation, STX17 TM is shown to be inserted to a membrane containing PI4P but not to a membrane without it. These results indicate that PI4P is a strong candidate that STX17 binds to in the autophagosome.

We would like to thank Reviewer #2 for pointing out these strengths.

Weaknesses:

• It was not answered whether PI4P is crucial for the STX17 recruitment in cells because manipulation of the PI4P content in autophagic membranes was not successful for unknown reasons.

As we explained in the initial submission, we tried to deplete PI4P in autophagosomes by multiple methods but did not succeed. In this revised manuscript, we added the result of an experiment using the PI 4-kinase inhibitor NC03 (Figure 4―figure supplement 1), which shows no significant effect on the autophagosomal PI4P level and STX17 recruitment.

Author response image 1.

The PI 4-kinase inhibitor NC03 failed to suppress autophagosomal PI4P accumulation and STX17 recruitment. HEK293T cells stably expressing mRuby3–STX17TM (A) or mRuby3–CERT(PHD) (B) and Halotag-LC3 were cultured in starvation medium for 1 h and then treated with and without 10 μM NC03 for 10 min. Representative confocal images are shown. STX17TM- or CERT(PHD)-positive rates of LC3 structures per cell (n > 30 cells) are shown in the graphs. Solid horizontal lines indicate medians, boxes indicate the interquartile ranges (25th to 75th percentiles), and whiskers indicate the 5th to 95th percentiles. Differences were statistically analyzed by Welch’s t-test. Scale bars, 10 μm (main), 1 μm (inset).

• The molecular simulation study did not show whether PI4P is necessary for the STX17 TM insertion or whether other negatively charged lipids can play a similar role.

As the reviewer suggested, we performed the molecular dynamics simulation using membranes with phosphatidylinositol, a negatively charged lipid. STX17 TM approached the PI-containing membrane but was not inserted into the membrane within a time scale of 100 ns in simulations of all five structures. This data suggests that PI4P, which is more negatively charged than PI, is required for STX17 insertion. Thus, we have included these data in Figure 5E and F and added the following text to Lines 242–244. “Moreover, if the membrane contained phosphatidylinositol (PI) instead of PI4P, STX17 approached the PI-containing membrane but was not inserted into the membrane (Figure 5E, F, Video 3)."

Author response image 2.

(E) An example of a time series of simulated results of STX17TM insertion into a membrane consisting of 70% phosphatidylcholine (PC), 20% phosphatidylethanolamine (PE), and 10% phosphatidylinositol (PI). STX17TM is shown in blue. Phosphorus in PC, PE and PI are indicated by yellow, cyan, and orange, respectively. Short-tailed lipids are represented as green sticks. The time evolution series are shown in Video 3. (F) Time evolution of the z-coordinate of the center of mass (z_cm) of the transmembrane helices of STX17TM in the case of membranes with PI. Five independent simulation results are represented by solid lines of different colors. The gray dashed lines indicate the locations of the lipid heads. A scale bar indicates 5 nm.

• The question that the authors posed in the beginning, i.e., why is STX17 recruited to the mature (closed) autophagosome but not to immature autophagic membranes, was not answered. The authors speculate that the seemingly gradual increase of negative charges in autophagic membranes is caused by an increase in PI4P. However, this was not supported by the PI4P fluorescence biosensor experiment that showed their distribution to the mature autophagosome only. Here, there are at least two possibilities: 1) The increase of negative charges in immature autophagic membranes is derived from PI4P. However the fluorescence biosensors do not bind there for some reason; for example, they are not sensitive enough to recognize PI4P until it reaches a certain level, or simply, their binding does not occur in a quantitative manner. 2) The negative charge in immature membranes is not derived from PI4P, and PI4P is generated abundantly only after autophagosomes are closed. In either case, it is not easy to explain why STX17 is recruited to the mature autophagosome only. For the first scenario, it is not clear how the PI4P synthesis is regulated so that it reaches a sufficient level only after the membrane closure. In the second case, the mechanism that produces PI4P only after the autophagosome closure needs to be elucidated (so, in this case, the question of the temporal regulation issue remains the same).

We thank the reviewers for pointing this out. While the probe for weakly negative charges (1K8Q) labeled both immature and mature autophagosomes, the probes for intermediate charges (5K4Q and 3K6Q) and PI4P labeled only mature autophagosomes (Figure 2F, Figure 2–figure supplement 1B). Thus, we think that the autophagosomal membrane rapidly and drastically becomes negatively charged, and at the same time, PI4P is enriched. Although immature membranes may have weak negative charges, we did not examine which lipids contribute to the negative charges. Thus, we have added the following sentences to the Discussion part.

“Our data of the 1K8Q probe suggest that immature autophagosomal membranes may also have slight negative charges (Figure 2E). Although the source of the negative charge of immature autophagosomes is currently unknown, it may be derived from low levels of PI4P, which is undetectable by the PI4P probes and/or other negatively charged lipids such as PI and PS (Schmitt et al., EMBO Rep, 2022).” (Lines 279–283) “In any case, it would be important to elucidate how PI 4-kinase activity or PI4P synthesis is upregulated during autophagosome maturation.” (Lines 302–303)

Reviewer #3 (Public Review):

Summary:

In this study, the authors set out to address the question of how the SNARE protein Syntaxin 17 senses autophagosome maturation by being recruited to autophagosomal membranes only once autophagosome formation and sealing is complete. The authors discover that the C-terminal region of Syntaxin 17 is essential for its sensing mechanism that involves two transmembrane domains and a positively charged region. The authors discover that the lipid PI4P is highly enriched in mature autophagosomes and that electrostatic interaction with Syntaxin 17's positively charged region with PI4P drives recruitment specifically to mature autophagosomes. The temporal basis for PI4P enrichment and Syntaxin 17 recruitment to ensure that unsealed autophagosomes do not fuse with lysosomes is a very interesting and important discovery. Overall, the data are clear and convincing, with the study providing important mechanistic insights that will be of broad interest to the autophagy field, and also to cell biologists interested in phosphoinositide lipid biology. The author's discovery also provides an opportunity for future research in which Syntaxin 17's c-terminal region could be used to target factors of interest to mature autophagosomes.

Strengths:

The study combines clear and convincing cell biology data with in vitro approaches to show how Syntaxin 17 is recruited to mature autophagosomes. The authors take a methodical approach to narrow down the critical regions within Syntaxin 17 required for recruitment and use a variety of biosensors to show that PI4P is enriched on mature autophagosomes.

We would like to thank Reviewer #3 for the positive comments.

Weaknesses:

There are no major weaknesses, overall the work is highly convincing. It would have been beneficial if the authors could have shown whether altering PI4P levels would affect Syntaxin 17 recruitment. However, this is understandably a challenging experiment to undertake and the authors outlined their various attempts to tackle this question.

We thank Reviewer #3 for pointing this out. Please see our above response to Reviewer #2 (Public Review).

In addition, clear statements within the figure legends on the number of independent experimental repeats that were conducted for experiments that were quantitated are not currently present in the manuscript.

As pointed out by Reviewer #3, we have added the number of independent experimental repeats in the figure legends.

Reviewer #1 (Recommendations For The Authors):

This paper is well written and all experiments were conducted with a high standard. Several minor issues should be addressed before final publication.

(1) To further confirm the charge interaction, a charge screening experiment should be performed for Fig. 2A.

We have asked Reviewer #1 through the editor what this experiment meant and understood that it was to see the effects of high salt concentrations. We monitored the association of GFP-STX17TM with liposomes in the presence or absence of 1 M NaCl and found that it was blocked in a high ionic buffer. This data supports the electrostatic interaction of STX17 with membranes. We have included this data in Figure 2B and added the following sentences to Lines 124–126.

“The association of STX17TM with PI4P-containing membranes was abolished in the presence of 1 M NaCl (Figure 2B). These data suggest that STX17 can be recruited to negatively charged membranes via electrostatic interaction independent of the specific lipid species.”

Author response image 3.

GFP–STX17TM translated in vitro was incubated with rhodamine-labeled liposomes containing 70% PC, 20% PE and 10% PI4P in the presence of 1 M NaCl or 1.2 M sucrose. GFP intensities of liposomes were quantified and shown as in Figure 1C (n > 30).

(2) The authors claim that "Autophagosomes become negatively charged during maturation", based on experiments using membrane charge probes. Since it's mainly about the membrane, it's better to refine the claim to "The membrane of autophasosomes becomes...", which would be more precise and close to the topic of this paper.

We would like to thank the reviewer for pointing this out. This point is valid. As recommended, we have collected the phrases “Autophagosomes become negatively charged during maturation” to “The membrane of autophagosomes becomes negatively charged during maturation” (Line 72, 118, 262, 969 (title of Figure2), 1068 (title of Figure2–figure supplyment1)).

(3) The authors should add more discussion regarding the "specificity" for recruiting Syn17 through the charge interaction. Particularly, how Syn17 could be maintained before the closure of autophagosomes? For the MD simulations in Fig. 5, the current results don't add much to the manuscript. The cell biology experiments have demonstrated the conclusion. The authors could try to find more details about the insertion by analyzing the simulation movies. Do membrane packing defects play a role during the insertion process? A similar analysis was conducted for alpha-synuclein (https://pubmed.ncbi.nlm.nih.gov/33437978/).

Regarding the mechanism of STX17 maintenance in the cytosol, we do not think that other molecules, such as chaperones, are essential because purified recombinant mGFP-STX17TM used in this study is soluble. However, it does not rule out such a mechanism, which would be a future study.

In the paper by Liu et al. (PMID: 33437978), small liposomes with diameters of 25–50 nm are used. Therefore, there are packing defects in the highly curved membranes, to which alpha-synuclein helices are inserted in a curvature-dependent manner. On the other hand, autophagosomes are much larger (~1 um in diameter) and almost flat for STX17 molecules, so we think it is unlikely that STX17 recognizes the packing defect.

Reviewer #2 (Recommendations For The Authors):

• The two (and other) possibilities with regards to the interpretation of the negative charge/PI4P result in autophagic membranes are hoped to be discussed.

As mentioned above, we have added the following sentences to the Discussion section. “Our data of the 1K8Q probe suggest that immature autophagosomal membranes may also have slight negative charges (Figure 2E). Although the source of the negative charge of immature autophagosomes is currently unknown, it may be derived from low levels of PI4P, which is undetectable by the PI4P probes and/or other negatively charged lipids such as PI and PS (Schmitt et al., EMBO Rep, 2022).” (Lines 279–283)

“In any case, it would be important to elucidate how PI 4-kinase activity or PI4P synthesis is upregulated during autophagosome maturation.” (Lines 302–303)

• Fluorescence biosensors are convenient to give an overview of the intracellular distribution of various lipids, but some of them show false-negative results. For example, evectin-2-PH for PS binds to endosomes but not to the plasma membrane, even though the latter contains abundant PS. With regards to PI4P, some biosensors illuminate both the Golgi and autophagosome, while others do not appear to bind the Golgi. Moreover, fluorescence biosensors for PI(3,5)P2 and PI(3,4)P2, which are also candidates for the STX17 insertion issue, are less reliable than others (e.g., those for PI3P and PI(4,5)P2). These problems need to be considered.

We agree with Reviewer #2 that fluorescence biosensors are not perfect for detecting specific lipids. Based on the Reviewer’s suggestion, we have included a comment on this in the Discussion section as follows (Lines 265–268).

“Given the possibility that fluorescence lipid probes may give false-negative results, a more comprehensive biochemical analysis, such as lipidomics analysis of mature autophagosomes, would be imperative to elucidate the potential involvement of other negatively charged lipids.”

• A negative control for the PI4P biosensor, i.e., a mutant lacking the PI4P binding ability, is better to be tested to confirm the presence of PI4P in autophagosomes.

We would like to thank the Reviewer for this comment. We conducted the suggested experiment and confirmed that the CERT(PHD)(W33A) mutant, which is deficient for PI4P binding (Sugiki et al., JBC. 2012), was diffusely present in the cytosol and did not localize to STX17-positive autophagosomes. This data supports our conclusion that PI4P is indeed present in autophagosomes. We have included this data in Figure 3–figure supplement 2A and explained it in the text (Lines 164–166).

Author response image 4.

Mouse embryonic fibroblasts (MEFs) stably expressing GFP–CERT(PHD)(W33A) and mRuby3–STX17TM were cultured in starvation medium for 1 h. Bars indicate 10 μm (main images) and 1 μm (insets).

• As a control to the molecular dynamic simulation study, STX17 TM insertion into a membrane containing other negative charge lipids, especially PI, needs to be tested. PI is a negative charge lipid that is likely to exist in autophagic membranes (as suggested by the authors' past study).

We thank the reviewers for this suggestion. As mentioned above (Reviewer #2, Public Review), we performed the molecular dynamics simulation using membranes containing PI and added the results in Figure 5E and F and Video 3.

• If the putative role of PI4P could be shown in the cellular context, the authors' conclusion would be much strengthened. I wonder if overexpression of PI4P fluorescence biosensors, especially those that appear to bind to the autophagosome almost exclusively, may suppress the recruitment of STX17 there.

We would like to thank the Reviewer for asking this question. In MEFs stably overexpressing PI4P probes driven by the CMV promoter, STX17 recruitment was not affected. Thus, simple overexpression of PI4P probes does not appear to be effective in masking PI4P in autophagosomes.

Another idea is to use an appropriate molecule (e.g., WIPI2, ATG5) and to recruit Sac1 to autophagic membranes by using the FRB-FKBP system or the like. I hope these and other possibilities will be tested to confirm the importance of PI4P in the temporal regulation of STX17 recruitment.

We tried the FRB-FKBP system using the phosphatase domain of yeast Sac1 fused to FKBP and LC3 fused to FRB, but unfortunately, this system failed to deplete PI4P from the autophagosomal membrane.

Reviewer #3 (Recommendations For The Authors):

A few areas for suggested improvement are:

(1) It would be helpful if the authors could clarify for all figures how many independent experiments were conducted for all experiments, particularly those that have quantitation and statistical analyses.

As pointed out by Reviewer #3, we have added the number of independent experimental repeats in the figure legends.

The authors made several attempts to modulate PI4P levels on autophagosomes although understandably this proved to be challenging. A couple of suggestions are provided to address this area:

(2) Given the reported role of GABARAPs in PI4K2a recruitment and PI4P production on autophagosomes, as well as autophagosome-lysosome fusion (Nguyen et al (2016) J Cell Biol) it would be worthwhile to assess whether GABARAP TKO cells have reduced PI4P and reduced Stx17 recruitment

According to the Reviewer’s suggestion, we examined the localization of STX17 TM and the PI4P probe CERT(PHD) in ATG8 family (LC3/GABARAP) hexa KO HeLa cells that were established by the Lazarou lab (Nguyen et al., JCB 2016). As in WT cells, STX17 TM and CERT(PHD) were still colocalized with each other in hexa KO cells, suggesting that neither STX17 recruitment nor PI4P enrichment depends on ATG8 family proteins (note: the size of autophagosomes in HeLa cells is smaller than in MEFs, making it difficult to observe autophagosomes as ring-shaped structures). We have included this result in Figure 3–figure supplement 2(F) and explained it in the text (Lines 194–196, 198).

Author response image 5.

(F) WT and ATG8 hexa KO HeLa cells stably expressing GFP–STX17TM and transiently expressing mRuby3–CERT(PHD) were cultured in starvation medium. Bars indicate 10 μm (main images) and 1 μm (insets).

(3) Can the authors try fusing Sac1 to one of the PI4P probes (CERT(PHD)) that were used, or alternatively to the c-terminus of Syntaxin 17? This approach would help to recruit Sac1 only to mature autophagosomes and could therefore prevent the autophagosome formation defect observed when fused to LC3B that targeted Sac1 to autophagosomes as they were forming. Understandably, this approach might seem a bit counterintuitive since the phosphatase is removing PI4P which is what is recruiting it but it could be a viable approach to keep PI4P levels low enough on mature autophagosomes so that Syntaxin 17 is no longer recruited. A Sac1 phosphatase mutant might be needed as a control.

We would like to thank the Reviewer for these suggestions. We tried the phosphatase domain of yeast Sac1 or human SAC1 fused with STX17TM, but unfortunately, these fusion proteins did not deplete PI4P from autophagosomes.

Reviewer #1 (Public Review):

In this manuscript, the authors report a molecular mechanism for recruiting syntaxin 17 (Syn17) to the closed autophagosomes through the charge interaction between enriched PI4P and the C-terminal region of Syn17. How to precisely control the location and conformation of proteins is critical for maintaining autophagic flux. Particularly, the recruitment of Syn17 to autophagosomes remains unclear. In this paper, the author describes a simple lipid-protein interaction model beyond previous studies focusing on protein-protein interactions. This represents conceptual advances.

Reviewer #2 (Public Review):

Summary:

Syntaxin17 (STX17) is a SNARE protein that is recruited to mature (i.e., closed) autophagosomes, but not to immature (i.e., unclosed) ones, and mediates the autophagosome-lysosome fusion. How STX17 recognizes the mature autophagosome is an unresolved interesting question in the autophagy field. Shinoda and colleagues set out to answer this question by focusing on the C-terminal domain of STX17 and found that PI4P is a strong candidate that causes the STX17 recruitment to the autophagosome.

Strengths:

The main findings are: 1) Rich positive charges in the C-terminal domain of STX17 are sufficient for the recruitment to the mature autophagosome; 2) Fluorescence charge sensors of different strengths suggest that autophagic membranes have negative charges and the charge increases as they mature; 3) Among a battery of fluorescence biosensors, only PI4P-binding biosensors distribute to the mature autophagosome; 4) STX17 bound to isolated autophagosomes is released by treatment with Sac1 phosphatase; 5) By dynamic molecular simulation, STX17 TM is shown to be inserted to a membrane containing PI4P but not to a membrane without it. These results indicate that PI4P is a strong candidate that STX17 binds to in the autophagosome.

Weaknesses:

• It was not answered whether PI4P is crucial for the STX17 recruitment in cells because manipulation of the PI4P content in autophagic membranes was not successful for unknown reasons.<br /> • The question that the authors posed in the beginning, i.e., why is STX17 recruited to the mature (closed) autophagosome but not to immature autophagic membranes, was not answered. The authors speculate that the seemingly gradual increase of negative charges in autophagic membranes is caused by an increase in PI4P. However, this was not supported by the PI4P fluorescence biosensor experiment that showed their distribution to the mature autophagosome only. Here, there are at least two possibilities: 1) The increase of negative charges in immature autophagic membranes is derived from PI4P. However the fluorescence biosensors do not bind there for some reason; for example, they are not sensitive enough to recognize PI4P until it reaches a certain level, or simply, their binding does not occur in a quantitative manner. 2) The negative charge in immature membranes is not derived from PI4P, and PI4P is generated abundantly only after autophagosomes are closed. In either case, it is not easy to explain why STX17 is recruited to the mature autophagosome only. For the first scenario, it is not clear how the PI4P synthesis is regulated so that it reaches a sufficient level only after the membrane closure. In the second case, the mechanism that produces PI4P only after the autophagosome closure needs to be elucidated (so, in this case, the question of the temporal regulation issue remains the same).

Reviewer #3 (Public Review):

Summary:

In this study, the authors set out to address the question of how the SNARE protein Syntaxin 17 senses autophagosome maturation by being recruited to autophagosomal membranes only once autophagosome formation and sealing is complete. The authors discover that the C-terminal region of Syntaxin 17 is essential for its sensing mechanism that involves two transmembrane domains and a positively charged region. The authors discover that the lipid PI4P is highly enriched in mature autophagosomes and that electrostatic interaction with Syntaxin 17's positively charged region with PI4P drives recruitment specifically to mature autophagosomes. The temporal basis for PI4P enrichment and Syntaxin 17 recruitment to ensure that unsealed autophagosomes do not fuse with lysosomes is a very interesting and important discovery. Overall, the data are clear and convincing, with the study providing important mechanistic insights that will be of broad interest to the autophagy field, and also to cell biologists interested in phosphoinositide lipid biology. The author's discovery also provides an opportunity for future research in which Syntaxin 17's c-terminal region could be used to target factors of interest to mature autophagosomes.

Strengths:

The study combines clear and convincing cell biology data with in vitro approaches to show how Syntaxin 17 is recruited to mature autophagosomes. The authors take a methodical approach to narrow down the critical regions within Syntaxin 17 required for recruitment and use a variety of biosensors to show that PI4P is enriched on mature autophagosomes.

Weaknesses:

There are no major weaknesses, overall the work is highly convincing. It would have been beneficial if the authors could have shown whether altering PI4P levels would affect Syntaxin 17 recruitment. However, this is understandably a challenging experiment to undertake and the authors outlined their various attempts to tackle this question.

Reviewer #3 (Public Review):

Summary:

This manuscript by Liu et al. presents a case that CAPSL mutations are a cause of familial exudative vitreoretinopathy (FEVR). Attention was initially focused on the CAPSL gene from whole exome sequence analysis of two small families. The follow-up analyses included studies in which CAPSL was manipulated in endothelial cells of mice and multiple iterations of molecular and cellular analyses. Together, the data show that CAPSL influences endothelial cell proliferation and migration. Molecularly, transcriptomic and proteomic analyses suggest that CAPSL influences many genes/proteins that are also downstream targets of MYC and may be important to the mechanisms.

Strengths:

This multi-pronged approach found a previously unknown function for CAPSLs in endothelial cells and pointed at MYC pathways as high-quality candidates in the mechanism.

Weaknesses:

Two issues shape the overall impact for me. First, the unreported population frequency of the variants in the manuscript makes it unclear if CAPSL should be considered an interesting candidate possibly contributing to FEVR, or possibly a cause. Second, it is unclear if the identified variants act dominantly, as indicated in the pedigrees. The studies in mice utilized homozygotes for an endothelial cell-specific knockout, leaving uncertainty about what phenotypes might be observed if mice heterozygous for a ubiquitous knockout had instead been studied.

In my opinion, the following scientific issues are specific weaknesses that should be addressed:

(1) Please state in the manuscript the number of FEVR families that were studied by WES. Please also describe if the families had been selected for the absence of known mutations, and/or what percentage lack known pathogenic variants.

(2) A better clinical description of family 3104 would enhance the manuscript, especially the father. It is unclear what "manifested with FEVR symptoms, according to the medical records" means. Was the father diagnosed with FEVR? If the father has some iteration of a mild case, please describe it in more detail. If the lack of clinical images in the figure is indicative of a lack of medical documentation, please note this in the manuscript.

(3) The TGA stop codon can in some instances also influence splicing (PMID: 38012313). Please add a bioinformatic assessment of splicing prediction to the assays and report its output in the manuscript.

(4) More details regarding utilizing a "loxp-flanked allele of CAPSL" are needed. Is this an existing allele, if so, what is the allele and citation? If new (as suggested by S1), the newly generated CAPSL mutant mouse strain needs to be entered into the MGI database and assigned an official allele name - which should then be utilized in the manuscript and who generated the strain (presumably a core or company?) must be described.

(5) The statement in the methods "All mice used in the study were on a C57BL/6J genetic background," should be better defined. Was the new allele generated on a pure C57BL/6J genetic background, or bred to be some level of congenic? If congenic, to what generation? If unknown, please either test and report the homogeneity of the background, or consult with nomenclature experts (such as available through MGI) to adopt the appropriate F?+NX type designation. This also pertains to the Pdgfb-iCreER mice, which reference 43 describes as having been generated in an F2 population of C57BL/6 X CBA and did not designate the sub-strain of C57BL/6 mice. It is important because one of the explanations for missing heritability in FEVR may be a high level of dependence on genetic background. From the information in the current description, it is also not inherently obvious that the mice studied did not harbor confounding mutations such as rd1 or rd8.

(6) In my opinion, more experimental detail is needed regarding Figures 2 and 3. How many fields, of how many retinas and mice were analyzed in Figure 2? How many mice were assessed in Figure 3?

(7) I suggest adding into the methods whether P-values were corrected for multiple tests.

eLife assessment

This study explores the role of calcyphosine-like (CAPSL) in Familial Exudative Vitreoretinopathy (FEVR) via the MYC pathway, offering valuable insights into disease mechanisms that are supported by a solid, multi-pronged approach. The overall significance of the study might, however, be limited due to weak support from human genetic studies.

Reviewer #1 (Public Review):

Summary:

The author presents the discovery and characterization of CAPSL as a potential gene linked to Familial Exudative Vitreoretinopathy (FEVR), identifying one nonsense and one missense mutation within CAPSL in two distinct patient families afflicted by FEVR. Cell transfection assays suggest that the missense mutation adversely affects protein levels when overexpressed in cell cultures. Furthermore, conditionally knocking out CAPSL in vascular endothelial cells leads to compromised vascular development. The suppression of CAPSL in human retinal microvascular endothelial cells results in hindered tube formation, a decrease in cell proliferation, and disrupted cell polarity. Additionally, transcriptomic and proteomic profiling of these cells indicates alterations in the MYC pathway.

Strengths:

The study is nicely designed with a combination of in vivo and in vitro approaches, and the experimental results are good quality.

Weaknesses:

My reservations lie with the main assertion that CAPSL is associated with FEVR, as the genetic evidence from human studies appears relatively weak. Further careful examination of human genetics evidence in both patient cohorts and the general population will help to clarify. In light of human genetics, more caution needs to be exercised when interpreting results from mice and cell models and how is it related to the human patient phenotype.

Reviewer #2 (Public Review):

Summary:

This work identifies two variants in CAPSL in two-generation familial exudative vitreoretinopathy (FEVR) pedigrees, and using a knockout mouse model, they link CAPSL to retinal vascular development and endothelial proliferation. Together, these findings suggest that the identified variants may be causative and that CAPSL is a new FEVR-associated gene.

Strengths:

The authors' data provides compelling evidence that loss of the poorly understood protein CAPSL can lead to reduced endothelial proliferation in mouse retina and suppression of MYC signaling in vitro, consistent with the disease seen in FEVR patients. The study is important, providing new potential targets and mechanisms for this poorly understood disease. The paper is clearly written, and the data generally support the author's hypotheses.

Weaknesses:

(1) Both pedigrees described appear to suggest that heterozygosity is sufficient to cause disease, but authors have not explored the phenotype of Capsl heterozygous mice. Do these animals have reduced angiogenesis similar to KOs? Furthermore, while the p.R30X variant protein does not appear to be expressed in vitro, a substantial amount of p.L83F was detectable by western blot and appeared to be at the normal molecular weight. Given that the full knockout mouse phenotype is comparatively mild, it is unclear whether this modest reduction in protein expression would be sufficient to cause FEVR - especially as the affected individuals still have one healthy copy of the gene. Additional studies are needed to determine if these variants alter protein trafficking or localization in addition to expression, and if they can act in a dominant negative fashion.

(2) The manuscript nicely shows that loss of CAPSL leads to suppressed MYC signaling in vitro. However, given that endothelial MYC is regulated by numerous pathways and proteins, including FOXO1, VEGFR2, ERK, and Notch, and reduced MYC signaling is generally associated with reduced endothelial proliferation, this finding provides little insight into the mechanism of CAPSL in regulating endothelial proliferation. It would be helpful to explore the status of these other pathways in knockdown cells but as the authors provide only GSEA results and not the underlying data behind their RNA seq results, it is difficult for the reader to understand the full phenotype. Volcano plots or similar representations of the underlying expression data in Figures 6 and 7 as well as supplemental datasets showing the differentially regulated genes should be included. In addition, while the paper beautifully characterizes the delayed retinal angiogenesis phenotype in CAPSL knockout mice, the authors do not return to that model to confirm their in vitro findings.

(3) In Figure S2D, the result of this vascular leak experiment is unconvincing as no dye can be seen in the vessels. What are the kinetics for biocytin tracers to enter the bloodstream after IP injection? Why did the authors choose the IP instead of the IV route for this experiment? Differences in the uptake of the eye after IP injection could confound the results, especially in the context of a model with vascular dysfunction as here.

(4) In Figure 5, it is unclear how filipodia and tip cells were identified and selected for quantification. The panels do not include nuclear or tip cell-specific markers that would allow quantification of individual tip cells, and in Figure 5C it appears that some filipodia are not highlighted in the mutant panel.

Author response:

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

Recommendations for the authors

Reviewer #1 (Recommendations For The Authors):

Below I summarize points that should be addressed in a revised version of the manuscript.

• Page 6, first paragraph: I don't understand by the signals average out to a single state. If the distribution is indeed randomly distributed, a broad signal with low intensity should be present.

We agree that this statement may cause confusion. We changed the text (marked in bold) to clarify the statement: The mobility of the undocked SBDs will be higher than the diffusion of the whole complex, allowing the sampling of varying interdomain distances within a single burst. However, these dynamic variations are subsequently averaged to a singular FRET value during FRET calculations for each burst, and may appear as a single low FRET state in the histograms.

• Page 6, third paragraph: how can the donor only be detected in the acceptor channel? Is this tailing out?

Donor only signal is not detected in the acceptor channel. As described in page 5 and in the Materials & Methods section, the dye stoichiometry value is defined for each burst/dwell using three types of photon counts: donor-based donor emission (FDD), donor-based acceptor emission (FDA) and acceptorbased acceptor emission (FAA).

When no acceptor fluorophore is present FAA=0 and S=1.

Some donor photons bleed through into the acceptor channel, but we correct for this by calculating the leakage and crosstalk factors as described in the Materials and Methods (page 20).

We changed the text (marked in bold) in the manuscript to address the question: The FRET data of both OpuA variants is best explained by a four-state model (Figure 2A,B; fourth and fifth panel) (Supplementary File 3). Two of the four states represent donor-only (S≈1) or acceptor-only (S≈0) dwells. The full bursts belonging to donor-only and acceptor-only molecules were excluded prior to mpH2MM. This means that some molecules transit to a donor-only or acceptor-only state within the burst period, which most likely reflects blinking or bleaching of one of the fluorophores. These donoronly and acceptor-only states were also excluded during further analysis. The other two states reflect genuine FRET dwells that were analyzed by mpH2MM. They represent different conformations of the SBDs.

• Page 7, "SBD dynamics ..": why was the V149Q mutant only analyzed in the K521C background and not also in the N414C background?

The two FRET states were best distinguished in OpuA-K521C. Therefore, we decided to focus on OpuA-K521C and not OpuA-N414C. OpuA-V149Q was used to show that reduced docking efficiency does not affect the transition rate constants and relative abundances of the two FRET states, and we regarded it sufficient to test the SBD dynamics in OpuA-K521C only.

• Page 8, second paragraph: why was the N414C mutant analyzed only from 0 - 600 mM and not also up to 1000 mM?

In line with the previous answer, our main focus was on OpuA-K521C, since the two FRET states were best distinguished in OpuA-K521C. OpuA-N414C was used to prove that similar states are observed when measuring with fluorophores on the opposite site of the SBD. We studied how the FRET states change in response to different conditions that correspond to different stages of the transport cycle and how it changes in response to different ionic strengths. Initially, 600 mM KCl was used to study the dynamics of the SBD at high ionic strength. Later in this study, we tested a very wide range of different salt concentrations for OpuA-K521C to get detailed insights into the dynamics of the SBDs over a wide ionic strength range. Note that 1 M KCl is a very high, non-physiological ionic strength for the typical habitat of L. lactis and was only used to show that the high FRET state occurs even under very extreme conditions.

• Page 8, third paragraph: why was the dimer (if it is the source of the FRET signal) only partially disrupted?

We acknowledge that this is a very good point. However, we purposely did not speculate on this point in the manuscript, because we have limited information on the molecular details of the interaction. As we highlight on page 8, the SBDs experience each other in a very high apparent concentration (millimolar range). This means that the interactions are most likely very weak (low affinity) and not very specific. Such interactions are in the literature referred to as the quinary structure of proteins and they occur at the high macromolecular crowding in the cell and in proteins with tethered domains, and thus at high local concentrations. Such interactions can be screened by high ionic strength. In the revised manuscript, we now present the partially disrupted dimer structure in the context of the quinary structure of a protein (page 11):

In other words, the high FRET state may comprise an ensemble of weakly interacting states rather than a singular stable conformation, resembling the quinary structure of proteins. The quinary structure of proteins is typically revealed in highly crowded cellular environments and describes the weak interactions between protein surfaces that contribute to their stability, function, and spatial organization (Guin & Gruebele, 2019). Despite the current study being conducted under dilute conditions, the local concentration of SBDs (~4 mM) mimics a densely populated environment and reveal quinary structure.

• Page 9, second paragraph: according to the EM data processing, only 20% of the particles were used for 3D reconstruction. Why? Does it mean that the remaining 80% were physiologically not relevant? If so, why were the 20% used relevant?

We note that it is a fundamental part of image processing of single particle cryo-EM data to remove false positives or low-resolution particles throughout the processing workflow. In particular when using a very low and therefore generous threshold during automated particle picking, as we did (t=0.01 and t=0.05 for the 50 mM KCl and 100 mM KCl datasets, respectively), the initial set of particles includes a significant amount of false positives – a tradeoff to avoid excluding particles belonging to low populated classes/orientations. It is thus common that more than 50% of ‘particles’ are excluded in the first rounds of 2D classification. In our case, only 30% and 52% of particles were retained after such first clean-up steps. Subsequently, the particle set is further refined, and additional false positives and low-resolution particles are excluded during extensive rounds of 3D classification. We also note that during the final steps, most of the data excluded represents particles of lower quality that do not contribute to a high-resolution, or belong to low population protein conformations. This does not mean that such a population is not physiological relevant. In conclusion, having only 5-20% of the initial automated picked particles contributing to the reconstruction of the final cryo-EM map is common, with the vast majority of excluded particles being false positives.

• Page 11, third paragraph: the way the proposed model is selected is also my main criticism. All alternative models do not fit the data. Therefore, the proposed model is suggested. However, I do not grasp any direct support for this model. Either I missed it or it is not presented.

Concerning the specific model in Figure 5, the reviewer is correct. We do not provide direct evidence for a side-ways interaction. However, we have evidence of transient interactions and our data rule out several scenarios of interaction, leaving 5C as the most likely model. This is also the main conclusion of this paper: In conclusion, the SBDs of OpuA transiently interact in a docking competent conformation, explaining the cooperativity between the SBDs during transport. The conformation of this interaction is not fixed but differs substantially between different conditions.

Because the interaction is very short-lived it was not possible to visualize molecular details of this interaction. We present Figure 5 to hypothesize the most likely type of interaction, since many possibilities can be excluded with the vast amount of presented data. To make our point more clear that we discuss models and rule out several possibilities but not demonstrate a specific interaction between the SBDs, we now write on page 10 (changes marked in bold): We have shown that the SBDs of OpuA come close together in a short-lived state, which is responsive to the addition of glycine betaine (Figure 4A). Although the occurrence of the state varies between different conditions, it was not possible to negate the high-FRET state completely, not even under very high or low KCl concentrations, or in the presence of 50 mM arginine plus 50 mM glutamate (Figure 4A,B). To evaluate possible interdomain interactions scenarios we consider the following: (1) The SBDs of OpuA are connected to the TMDs with very short linkers of approximately 4 nm, which limit their movement and allow the receptor to sample a relatively small volume near its docking site. (2) in low ionic strength condition OpuA-K521C displays a high FRET state with mean FRET values of 0.7-0.8, which correspond to inter-dye distances of approximately 4 nm. (3) The high FRET state is responsive to glycine betaine, which points toward direct communication between the two SBDs. (4) The distance between the density centers of the SBDs in the cryo-EM reconstructions (based on particles with a low and high FRET state) is 6 nm, which aligns with the dimensions of an SBD (length: ~6 nm, maximal width: ~4 nm). These findings collectively indicate that two SBDs interact but not necessarily in a singular conformation but possibly as an ensemble of weakly interacting states. Hence, we discuss three possible SBD-SBD interaction models to explain the highFRET state:

Reviewer #2 (Recommendations For The Authors):

In the abstract and elsewhere the authors suggest that the SBDs physically interact with one another, and that this interaction is important for the transport mechanism, specifically for its cooperativity.

I feel that this main claim is not well established. The authors convincingly demonstrate that the SBDs largely occupy two states relative to one another and that in one of these states, they are closer than in the other. Unless I have missed (or failed to understand) some major details of the results, I did not find any evidence of a physical interaction. Have the authors established that the high FRET state indeed corresponds to the physical engagement of the SBDs? I feel that a direct demonstration of an interaction is much missing.

Along the same lines, in the low-salt cryo-EM structure, where the SBDs are relatively closer together, the SBDs are still separated and do not interact.

See also our response to the final comment of reviewer 1. Furthermore, please carefully consider the following: (1) FRET values of 0.7-0.8 correspond to inter-dye distances of approximately 4 nm. (2) The high FRET state is responsive to glycine betaine, which points toward direct communication between the two SBDs. (3) The cryo-EM reconstruction is the average of all the particles in the final dataset, including both the particles with a low and high FRET state. Further, the local resolution of the SBDs in the cryo-EM map is low, indicative of high degree of flexibility. Thus, a potential interaction is possible within the observed range of flexibility. (4) The distance between the density centers is 6 nm, aligning with the dimensions of an SBD (length: 6 nm, maximal width: 4 nm). These factors collectively indicate SBD interactions, and we present these points now more explicitly in Figure 4 and the last part of the results section (page 9).

Once the authors successfully demonstrate that direct physical interaction indeed occurs, they will need to provide data that places it in the context of the transport cycle. Do the SBDs swap ligand molecules between them? Do they bind the ligand and/or the transporter cooperatively? What is the role of this interaction?

We acknowledge the intriguing nature of the posed questions, but they extend beyond the scope of this study. It is extremely challenging to obtain high-resolution structures of highly dynamic multidomain proteins, like OpuA, and to probe transient interactions as we do here for the SBDs of OpuA. We therefore combined cryo-TEM with smFRET studies and perform the most advanced and state-of-theart analysis tools as acknowledged by reviewer 1. We link our observations on the structural dynamics and interactions of the SBDs to a previous study, where we showed that the two SBDs of OpuA interact cooperatively. We do not have further evidence that connect the physical interactions to the transport cycle. In our view, the collective datasets indicate that the here reported physical interactions between the SBDs increase the transport efficiency.

As far as I understand, the smFRET data have been interpreted on the basis of a negative observation, i.e., that it is "likely" that none of the FRET states corresponds to a docked SBD. To convincingly show this, a positive observation is required, i.e., observation of a docked state.

The aim of this study was to study interdomain dynamics and not specifically docking. We have previously shown that docking can be visualized via cryo-EM (Sikkema et al., 2020), however the SBDs of OpuA appear to only dock in specific turnover conditions. We now show that the high FRET state of OpuA cannot represent a docked state, but that the SBDs transiently interact (see our response to the first comment). Importantly, a docked state was also not found in the cryo-EM reconstructions at low ionic strength, representing the smFRET conditions where we observe the interactions between the SBDs. The high FRET state occupies 30% of the dwells in this condition, and such a high percentage of molecules would have become apparent during cryo-EM 3D classification in case they would form a docked state. Therefore, we conclude that docking does not occur in low ionic strength apo condition. We discuss this point and our reasoning on page 11 of the revised manuscript.

In this respect, I find it troubling that in none of the tested conditions, the authors observed a FRET state which corresponds to the docked state. Such a state, which must exist for transport to occur (as mentioned in the authors' previous publications), needs to be demonstrated. This brings me to my next question: why have the authors not measured FRET between the SBDs and the transporter? Isn't this a very important piece that is missing from their puzzle?

We agree that investigating docking behavior under varied turnover conditions requires focused experiments on FRET dynamics between the SBDs and the transporter. As noted on page 5, OpuA exists as a homodimer, implying that a single cysteine mutation introduces two cysteines in a single functional transporter. To specifically implement a cysteine mutation in only one SBD and one transmembrane domain, it is necessary to artificially construct a heterodimer. We recently published initial attempts in this direction, and this will be a subject for future research but still requires years of work.

Additionally, I feel that important controls are missing. For example, how will the data presented in Fig1 look if the transporter is labeled with acceptor or donor only? How do soluble SBDs behave?

In the employed labeling method, donor and acceptor dyes are mixed in a 1:1 ratio and randomly attached to the two cysteines in the transporter. This automatically yields significant fractions of donor only and acceptor only transporters which are always present during the smFRET recordings. We can visualize those molecules on the basis of the dye stoichiometry, which we calculate by using three types of photon counts: donor-based donor emission (FDD), donor-based acceptor emission (FDA) and acceptorbased acceptor emission (FAA).

Unfiltered plots look as follows (a dataset of OpuA-K521C at 600 mM KCl):

Author response image 1.

Donor only and acceptor only molecules have a very well discernible stoichiometry of 1 and 0, respectively. The filtering procedure is described in the materials and methods section, and these plots can be found in the supplementary database. We did not add them to the main text or supplementary materials of the original manuscript, as this is a very common procedure in the field of smFRET. We now include such a dataset in the revised manuscript.

Soluble SBDs of OpuA have been studied previously (e.g. Wolters et al., 2010 & De Boer et al. 2019). For example, we have shown by SEC-MALLLS that soluble SBDs do not form dimers, which is consistent with our notion that the SBDs interact with low affinity. It is not possible to study interdomain dynamics between soluble SBDs by smFRET, because the measurements are carried out at picomolar concentrations (monomeric conditions). We emphasize that smFRET measurements with native complexes, with SBDs near each other at apparent millimolar concentrations, is physiologically more relevant.

Additional comments:

(1) "It could well be that cooperativity and transient interactions between SBDs is more common than previously anticipated" and a similar statement in the abstract. What evidence is there to suggest that the transient interactions between SBDs are a common phenomenon?

On page 11, we write: Dimer formation of SBPs has been described for a variety of proteins from different structural clusters of substrate-binding proteins [33–38,51–53]. We cite 9 papers that report SBD/SBP dimers. This suggest to us that the phenomenon of interacting substrate-binding proteins could be more common. Moreover, the concentration of maltose-binding protein and other SBPs in the periplasm of Gram-negative bacteria can reach (sub)millimolar concentrations, and low-affinity interactions may play a role not only in membrane protein-tethered SBDs (like in OpuA) but also be important in soluble substrate-receptors. Such low-affinity interactions are rarely studied in biochemical experiments.

(2) I think that the data presented in 1B-C better suits the supplementary information.

Figure 1B-D is already a summary of the supplementary information that describes the optimization of OpuA purification. We think it is valuable to show this part of the figure in the main text. A very clean and highly pure OpuA sample is essential for smFRET experiments. Quality of protein preparations and data analysis are key for the type of measurements we report in this paper.

(3) "the first peak in the SEC profile corresponds...." The peaks should be numbered in the figure to facilitate their identification.

We have changed the figure as suggested.

(4) "smFRET is a powerful tool for studying protein dynamics, but it has only been used for a handful of membrane proteins". With the growing list of membrane proteins studied by smFRET I find this an overstatement.

We removed this sentence in the new version of the manuscript.

(5) "We rationalized that docking of one SBD could induce a distance shift between the two SBDs in the FRET range of 3-10 nm (Figure 1E)" How and why was this assumed?

We realize that this is one of the sentences that caused confusion about the aim of this study. In this part of the manuscript, we should not have used docking as an example and we apologize for that. We replaced the sentence by: These variants are used to study inter-SBD dynamics in the FRET range of 310 nm (Figure 1E).

Also Figure 1E was adjusted to prevent confusion:

Author response image 2.

In addition, to avoid any confusion we changed the following sentence on page 4 (changes marked in bold): We designed cysteine mutations in the SBD of OpuA to study interdomain dynamics in the full length transporter.

(6) "However, the FRET distributions are broader than would be expected from a single FRET state, especially for OpuA-K521C" Have the authors established how a single state FRET of OpuA looks? Is there a control that supports this claim?

Below we compare two datasets from OpuA-K521C in 600 mM KCl with a typical smFRET dataset from the well-studied substrate-binding protein MBP from E. coli, which resides in a single state. Left: OpuA-K521C; Right: MBP

Author response image 3.

We agree that this cannot be assumed from the presented data. Therefore we rewrote this sentence: However, the FRET distributions tail towards higher FRET values, especially OpuA-K521C.

(7) "V149Q was designed as a mild mutation that would reduce docking efficiency and thereby substrate loading, but leave the intrinsic transport and ATP hydrolysis efficiency intact." I find this statement confusing: How can a mutation reduce docking efficiency yet leave the transport activity unchanged?

We rewrote the sentences (changes marked in bold): V149Q was designed as a mild mutation that would reduce docking efficiency and thereby substrate loading, but leave the ionic strength sensing in the NBD and the binding of glycine betaine and ATP intact. Accordingly, a reduced docking efficiency should result in a lower absolute glycine betaine-dependent ATPase activity. At the same time the responsiveness of the system to varying KCl, glycine betaine, or Mg-ATP concentrations should not change.

(8) Along the same lines: "whereas the glycine betaine-, Mg-ATP-, or KCl-dependent activity profiles remain unchanged" vs. "OpuA-V149Q-K521C exhibited a 2- to 3-fold reduction in glycine betainedependent ATPase activity".

See comment at point 7.

(9) In general, I find the writing wanting at places, not on par with the high standards set by previous publications of this group.

We recognize the potential ambiguity in our phrasing. We hope that after incorporating the feedback provided by the reviewers our manuscript will convey our findings in a clearer manner.

Extra changes to the text:

(1) Title changed: The substrate-binding domains of the osmoregulatory ABC importer OpuA physically transiently interact

(2) Second part of the abstract changed: We now show, by means of solution-based single-molecule FRET and analysis with multi-parameter photon-by-photon hidden Markov modeling, that the SBDs transiently interact in an ionic strength-dependent manner. The smFRET data are in accordance with the apparent cooperativity in transport and supported by new cryo-EM data of OpuA. We propose that the physical interactions between SBDs and cooperativity in substrate delivery are part of the transport mechanism.

(3) Page 6, third paragraph and Figure 2B: the wrong rate number was extracted from table 1. Changed this in the text and figure: 112 s-1  173 s-1. It did not affect any of the interpretations or conclusions.

(4) Page 8, last paragraph, changed: smFRET was also performed in the absence of KCl and with a saturating concentration of glycine betaine (100 µM). The mean FRET efficiency of the highFRET state of OpuA-K521C increased to 0.78, which corresponds to an inter-dye distance of about 4 nm. This indicates that the dyes at the two SBDs move very close towards each other (Figure 4A) (Table 1) (Supplementary File 34).

(5) Page 9, second paragraph changed: Due to the inherent flexibility of the SBDs, with respect to both the MSP protein of the nanodisc and the TMDs of OpuA, their resolution is limited. Furthermore, the cryo-EM reconstructions average all the particles in the final dataset, including those with a low and high FRET state. Nevertheless, in both conditions, the densities that correspond to the SBDs can be observed in close proximity (Figure 4D). The distance between the density centers is 6 nm and align with the dimensions of an SBD, providing further evidence for physical interactions between the SBDs.

eLife assessment

The OpuA Type I ABC importer uses two substrate binding domains to capture extracellular glycine betaine and present the substrate to the transmembrane domain for subsequent transport and correction of internal dehydration. This study presents valuable findings addressing the question of whether the two substrate binding domains of OpuA dock and physically interact in a salt-dependent manner. The single-molecule fluorescence resonance energy transfer and cryogenic electron microscopy data that are presented provide convincing support for the existence of a transient interaction between the substrate binding domains that depends on ionic strength, laying a foundation for future studies exploring how this interaction is involved in the overall transport mechanism.

Reviewer #1 (Public Review):

Summary: The type I ABC importer OpuA from Lactococcus lactis is the best studied transporter involved in osmoprotection. In contrast to most ABC import systems, the substrate binding protein is fused via a short linker to the transmembrane domain of the transporter. Consequently, this moiety is called the substrate binding domain (SBD). OpuA has been studied in the past in great detail and we have a very detailed knowledge about function, mechanisms of activation and deactivation as well as structure.

Strengths: Application of smFRET to unravel transient interactions of the SBDs. The method is applied at a superb quality and the data evaluation is excellent.

Weaknesses: The proposed model is not directly supported by experimental data. Rather alternative models are excluded as they do not fit to the obtained data. However, this is now clearly stated in the manuscript

Reviewer #2 (Public Review):

Summary:<br /> In this report the authors used solution-based single-molecule FRET and low resolution cryo-EM to investigate the interactions between the substrate-binding domains of the ABC-importer OpuA from Lactococcus lactis. Based on their results, the authors suggest that the SBDs interact in an ionic strength-dependent manner.

Strengths:<br /> The strength of this manuscript is the uniqueness and importance of the scientific question, the adequacy of the experimental system (OpuA), and the combination of two very powerful and demanding experimental approaches.

Weaknesses:<br /> A demonstration that the SBDs physically interact with one another, and that this interaction is important for the transport mechanism will greatly strengthen the claims of the authors. The relation to cooperativity is also unclear.

Author response:

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

We are grateful to the Editors for overseeing the review of our manuscript, and to the two reviewers for their thoughtful comments and suggestions for how it can be improved.

I submit at this time a revision, as well as a detailed response (below) to each of the points raised in the first round of review.

We feel the manuscript has been significantly improved by taking the reviewers' comments to heart. In a nutshell, we added new key pieces of data (impact of WIN site inhibition on global translation, rRNA production, as well as the requested cell biology analyses showing nucleolar stress), new analyses of the proteomics to counter potential concerns with normalization, and expanded/revised verbiage in key areas to clarify parts of the text that were confusing or problematic. The main figures have not changed; all new material is included in supplements to figures 2 and 3.

Public Reviews

Reviewer #1 (Public Review):

Building on previous work from the Tansey lab, here Howard et al. characterize transcriptional and translational changes upon WIN site inhibition of WDR5 in MLL-rearranged cancer cells. They first analyze whether C16, a newer generation compound, has the same cellular effects as C6, an early generation compound. Both compounds reduce the expression of WDR5-bound RPGs in addition to the unbound RPG RPL22L1. They then investigate differential translation by ribo-seq and observe that WIN site inhibition reduces the translational RPGs and other proteins related to biomass accumulation (spliceosome, proteasome, mitochondrial ribosome). Interestingly, this reduction adds to the transcriptional changes and is not limited to RPGs whose promoters are bound by WDR5. Quantitative proteomics at two-time points confirmed the downregulation of RPGs. Interestingly, the overall effects are modest, but RPL22LA is strongly affected. Unexpectedly, most differentially abundant proteins seem to be upregulated 24 h after C6 (see below). A genetic screen showed that loss of p53 rescues the effect of C6 and C16 and helped the authors to identify pathways that can be targeted by compounds together with WIN site inhibitors in a synergistic way. Finally, the authors elucidated the underlying mechanisms and analyzed the functional relevance of the RPL22, RPL22L1, p53, and MDM4 axis.

While this work is not conceptually new, it is an important extension of the observations of Aho et al. The results are clearly described and, in my view, very meaningful overall.

Major points:

(1) The authors make statements about the globality/selectivity of the responses in RNA-seq, ribo-seq, and quantitative proteomics. However, as far as I can see, none of these analyses have spike-in controls. I recommend either repeating the experiments with a spike-in control or carefully measuring transcription and translation rates upon WIN site inhibition and normalizing the omics experiments with this factor.

The reviewer is correct that we did not include spike-in controls in our omics experiments. We would like to emphasize that none of the omics data in this manuscript have been processed in unorthodox ways, and that the major conclusions each have independent corroborating data.

The selectivity in RPG suppression observed in RNA-Seq, for example, is supported by results from our target engagement (QuantiGene) assays; suppression of RPL22L1 mRNA levels is supported by quantitative and semi-quantitative RT-PCR, by western blotting, and by the results of our proteomic profiling; alternative splicing (and expression) of MDM4—and its dependency on RPL22—is also backed up by similar RT-PCR and western blotting data. The same applies for alternative splicing of RPL22L1.

That said, we do appreciate the point the reviewer is making here, and have done our best to respond. We do not think it is a prudent investment in resources to repeat the numerous omics assays in the manuscript. We also considered normalizing for bulk transcription and translation rates as suggested, but it is not clear in practice how this would be done, and it could introduce additional variables and uncertainties that may skew the interpretation of results. Instead, to respond to this comment, we made the following changes to the manuscript:

(1) We now explicitly state, for all omics assays, that spike-in controls were not included. These statements will prompt the reader to make their own assessment of the robustness of each of our findings and interpretations.

(2) We have added new data to the manuscript (Figure 2—figure supplement 1A–B) measuring the impact of C6 and C16 on bulk translation using the OPP labeling method. These new data demonstrate that WIN site inhibitors induce a progressive yet modest decline in protein synthesis capacity. At 24 hours, there is no significant effect of either agent on protein synthesis levels. By 48 hours, a small but significant effect is observed, and by 96 hours translation levels are ~60% of what they are in vehicle-treated control cells. These new data are important because they support the idea that normalization has not blunted the responses we observe—the magnitude of the effects are consistent between the different assays and tend to cap out at two-fold in terms of RPG suppression, translation efficiency, ribosomal protein levels, and protein synthesis capacity.

(3) We have included additional analysis regarding the LFQMS, as described below, that specifically addresses the issue of normalization in our proteomics experiments.

(2) Why are the majority of proteins upregulated in the proteomics experiment after 24 h in C6 (if really true after normalization with general protein amount per cell)? This is surprising and needs further explanation.

The reviewer is correct in noting that (by LFQMS) ~700 proteins are induced after 24 hours of treatment of MV4:11 cells with C16 (not C6, as stated). The reviewer would like us to examine whether this apparent increase in proteins is a normalization artifact. In response to this comment, we have made the following changes to the manuscript:

(1) Our new OPP labeling experiments (Figure 2—figure supplement 1A–B) show that there is no significant reduction in overall protein synthesis following 24 hours of C16 treatment. In light of this finding, it is unlikely that normalization artifacts, resulting from diminution of the pool of highly abundant proteins, create the appearance of these 700 proteins being induced. We now explicitly make this point in the text.

(2) We now clarify in the methods how we seeded identical numbers of cells for DMSO and C16-treated cultures in these experiments, and—consistent with our finding that WIN site inhibitors have little if any effect on protein synthesis or proliferation at the 24 hour timepoint— extracted comparable amounts of proteins from these two treatment conditions (DMSO: 344.75 ± 21.7 µg; C16: 366.50 ± 15.8 µg; [Mean ± SEM]).

(3) We now include in Figure 3—figure supplement 1A a plot showing the distribution of peptide intensities for each protein detected in each run of LFQMS before and after equal median normalization. This new analysis reveals that the distribution of intensities is not appreciably changed via normalization. Specifically, there is not a reduction in peptide intensities in the unnormalized data from 24 hours of C16 treatment that is reversed or tempered by normalization. This analysis provides further support for the notion that the increase we observe is not a normalization artifact.

(4) We now include in Figure 3—figure supplement 1B–D a set of new analyses examining the relationship between the initial intensity of proteins in DMSO control samples (a crude proxy for abundance) versus the fold change in response to WIN site inhibitor. This analysis shows that we have as many "highly abundant" (10th decile) proteins increasing as we do decreasing in response to WINi. Thus, it appears as though the wholesale clearance of highly abundant proteins from the cell is not occurring at this early treatment timepoint. In addition, this analysis also shows that ribosomal proteins (RP) are generally the most abundant, most suppressed, proteins and that their fold-change at the protein level at 24 hours is less than two-fold, consistent again with the magnitude of transcriptional effects of C16, as measured by RNA-Seq and QuantiGene. The fact that the drop in RP levels is consistent with expectations based on other analyses provides further empirical support for the notion that protein levels inferred from LFQMS are authentic and not skewed by global changes in the proteome.

The increase in proteins at this time point, we argue, is thus most likely genuine. It is not surprising that—at a timepoint at which protein synthesis is unaffected—several hundred proteins are induced by a factor of two. How this occurs, we do not know. It may be a transient compensatory mechanism, or it may be an early part of the active response to WIN site inhibitors. Lest the reader be confused by this finding, we have now added text to this section of the manuscript discussing and explaining the phenomenon in more detail.

(3) The description of the two CRISPR screens (GECKO and targeted) is a bit confusing. Do I understand correctly that in the GECKO screen, the treated cells are not compared with nontreated cells of the same time point, but with a time point 0? If so, this screen is not very meaningful and perhaps should be omitted. Also, it is unclear to me what the advantages of the targeted screen are since the targets were not covered with more sgRNAs (data contradictory: 4 or 10 sgRNAs per target?) than in Gecko. Also, genome-wide screens are feasible in culture for multiple conditions. Overall, I find the presentation of the screening results not favorable.

In essence, this is a single screen performed in two tiers. In Tier 1, we screened a complete GECKO library (six sgRNA/gene) with the earliest generation (less potent) inhibitor C6, and compared sgRNA representation against the time zero population. This screen would reveal sgRNAs that are specifically associated with response to C6, as well as those that are associated with general cell fitness and viability. We then identified genes connected to these sgRNAs, removed those that are pan essential, and built a custom library for the second tier using sgRNAs from the Brunello library (four sgRNA/gene). We then screened this custom library with both C6 and the more potent inhibitor C16, this time against DMSO-treated cells from the same timepoint.

We acknowledge that this is not the most streamlined setup for a screen. But our intention was to compare two inhibitors (C6 and C16) and identify high confidence 'hits' that are disconnected from general cell viability, rather than generate an exhaustive list of all genes that, when disrupted, skew the response to WIN site inhibitor. The final result of this screen (Figure 4E) is a gene list that has been validated with two chemically distinct WIN site inhibitors and up to 10 unique sgRNAs per gene. We may not have captured every gene that can modulate response to WIN site inhibitor, but those appearing in Figure 4E are highly validated.

To answer the reviewer's specific questions: (i) we cannot omit the Tier 1 screen because then there would be no rationale for what was screened in the second Tier; and (ii) the advantage of the custom Tier 2 library is that it allowed us to screen hits from the Tier 1 screen with four completely independent sgRNAs. Although there are not more sgRNAs for each gene in the Tier 2 versus the Tier 1 library, these sgRNAs are different and thus, for C6 at least, hits surviving both screens were validated with up to 10 unique sgRNAs.

We apologize that the description of the CRISPR screens was not clearer, and have reworked this section of the manuscript to make our intent and our actions clearer.

(4) Can Re-expression of RPL22 rescue the growth arrest of C6?.

We have not attempted to complement the RPL22 knock out. But we do note that evidence supporting the idea that loss of RPL22 confers resistance to WIN site inhibitor is strong—six (out of six) sgRNAs against RPL22 were significantly enriched in the Tier 1 screen, and independent knock out of RPL22 with the Synthego multi-guide system in MV4;11 and MOLM13 cells increases the GI50 for C16.

Reviewer #2 (Public Review):

Summary:

The manuscript by Howard et al reports the development of high-affinity WDR5-interaction site inhibitors (WINi) that engage the protein to block the arginine-dependent engagement with its partners. Treatment of MLL-rearranged leukemia cells with high-affinity WINi (C16) decreases the expression of genes encoding most ribosomal proteins and other proteins required for translation. Notably, although these targets are enriched for WDR5-ChIP-seq peaks, such peaks are not universally present in the target genes. High concordance was found between the alterations in gene expression due to C16 treatment and the changes resulting from treatment with an earlier, lower affinity WINi (C6). Besides protein synthesis, genes involved in DNA replication or MYC responses are downregulated, while p53 targets and apoptosis genes are upregulated. Ribosome profiling reveals a global decrease in translational efficiency due to WINi with overall ribosome occupancies of mRNAs ~50% of control samples. The magnitude of the decrements of translation for most individual mRNAs exceeds the respective changes in mRNA levels genome-wide. From these results and other considerations, the authors hypothesize that WINi results in ribosome depletion. Quantitative mass spec documents the decrement in ribosomal proteins following WINi treatment along with increases in p53 targets and proteins involved in apoptosis occurring over 3 days. Notably, RPL22L1 is essentially completely lost upon WINi treatment. The investigators next conduct a CRISPR screen to find moderators and cooperators with WINi. They identify components of p53 and DNA repair pathways as mediators of WINi-inflicted cell death (so gRNAs against these genes permit cell survival). Next, WINi are tested in combination with a variety of other agents to explore synergistic killing to improve their expected therapeutic efficacy. The authors document the loss of the p53 antagonist MDM4 (in combination with splicing alterations of RPL22L1), an observation that supports the notion that WINi killing is p53-mediated.

Strengths:

This is a scientifically very strong and well-written manuscript that applies a variety of state-ofthe art molecular approaches to interrogate the role of the WDR5 interaction site and WINi. They reveal that the effects of WINi seem to be focused on the overall synthesis of protein components of the translation apparatus, especially ribosomal proteins-even those that do not bind WDR5 by ChIP (a question left unanswered is how much the WDR5-less genes are nevertheless WINi targeted). They convincingly show that disruption of the synthesis of these proteins is accompanied by DNA damage inferred by H2AX-activation, activation of the p53pathway, and apoptosis. Pathways of possible WINi resistance and synergies with other antineoplastic approaches are explored. These experiments are all well-executed and strongly invite more extensive pre-clinical and translational studies of WINi in animal studies. The studies also may anticipate the use of WINi as probes of nucleolar function and ribosome synthesis though this was not really explored in the current manuscript.

Weaknesses:

A mild deficiency in the current manuscript is the absence of cell biological methods to complement the molecular biological and biochemical approaches so ably employed. Some microscopic observations and confirmation of nucleolar dysfunction and DNA damage would be reassuring.

We thank the reviewer for their comments. We agree that an absence of cell biological methods was a deficiency in the original manuscript. In response to this comment, we have now added immunofluorescence (IF) analyses, examining the impact of C16 on nucleolar integrity and nucleophosmin (NPM1) distribution (Figure 3—figure supplement 4). These new data clearly show that C16 induces nucleolar stress at 72 hours—as measured by the redistribution of NPM1 from the nucleolus to the nucleoplasm. These new data fill an important gap in the story, and we are grateful to the reviewer for prompting us to perform these experiments.

As part of the above study, we also probed for gamma-H2AX, expecting that we may see some signs of accumulation in the nucleoli (see comment #4 from Reviewer #2, below). We did not observe this response. Importantly, however, we did see that gamma-H2AX staining occurs only in what are overtly apoptotic cells. This is an important finding, because we had previously speculated that the induction of gamma-H2AX observed by Western blotting reflected part of a bona-fide response to DNA damage elicited by WIN site inhibitors. Instead, the IF data now leads us to conclude that this signal simply reflects the established fact that WIN site inhibitors induce apoptosis in this cell line (Aho et al., 2019). In response to this new finding, we have added additional discussion to the text and have removed or de-emphasized the potential contribution of DNA damage to the mechanism of action of WDR5 WIN site inhibitors. Again, we are grateful for this comment as it has prevented us from continuing to report/pursue erroneous observations.

Recommendations for the authors

Reviewer #1 (Recommendations For The Authors):

There is a typo in "but are are linked to mRNA instability when translation is inhibited".

Thank you for catching this typo. It has now been corrected.

Reviewer #2 (Recommendations For The Authors):

(1) The authors report that WINi initially (at 24 hrs) increases the expression of most proteins while decreasing ribosomal proteins, but at 72 hours all proteins are depressed. The transient bump-up of non-translation-related proteins seems odd. A simple resolution to this somewhat strange observation is that there is no real increase in the other proteins, but because of the loss of a large fraction of the most abundant cellular proteins (the ribosomal proteins), the relative fraction of all other proteins is increased; that is, the increase of non-ribosomal proteins may be an artifact of normalization to a lower total protein content. Can this be explored?

We are grateful to the reviewer for this comment. We have tried our best to respond, as detailed above in response to Reviewer #1 Public Comment #2.

(2) It would be really nice to assess nucleolar status microscopically. Do nucleoli get bigger? Smaller? Do they have abnormal morphology? Is there nucleolar stress? What happens to rRNA synthesis and processing?

We agree and thank the reviewer for raising this point. As noted in our response to Reviewer #2, above, we have included new IF that shows: (i) no obvious effect on nucleolar integrity, (ii) redistribution of NPM1 to the nucleoplasm (indicative of nucleolar stress), and (iii) induction of gamma-H2AX staining in apoptotic cells (indicative of apoptosis).

Additionally, in response to this comment, we also looked at the impact of WIN site inhibitors on rRNA synthesis, using AzCyd labeling. These new data appear in Figure 3—figure supplement 3. Interestingly, these new data show that there is a progressive decline in rRNA synthesis, and that by 96 hours of treatment levels of both 18S and 28S rRNAs are reduced— again by about a factor of two. Our interpretation of this finding is that in response to the progressive decline in RPG transcription there is a secondary decrease in rRNA synthesis. This result is perhaps not surprising, but it does again add an important missing piece to our characterization of WIN site inhibitors and is further support for the concept that inhibition of ribosome production is a dominant part of the response to these agents.

(3) The WINi elicited DNA damage is incompletely characterized, rather it is inferred from H2AX activation. Comet assays would help to confirm such damage.

As noted in our response to Reviewer #2, our original inference of DNA damage, prompted by gamma-H2AX activation, is erroneous, and due instead to the ability of WIN site inhibitors to induce apoptosis. We thus did not pursue comet assays, etc., and removed discussion of potential DNA damage from the manuscript.

(4) Staining and microscopic observation of H2AX would be very useful. Is the WINi provoked DNA damage nucleolar-localized? Does the deficiency of ribosomal proteins lead to localized genotoxic nucleolar stress - or alternatively does the paucity of ribosomes and decreased translation lead to imbalances in other cellular pathways, perhaps including some involved in overall genome maintenance which would provoke more global DNA damage and H2AX staining, not limited to the nucleolus.

Again, please see our response to the Public Comment from Reviewer #2.

(5) It would be important to assess the influence and effects of WINi on some p53 mutant, p53-/- and p53 wild-type cell lines. Given their prevalence, p53 status may be expected to alter WINi efficacy.

The issue of how p53 status impacts the response to WINi is interesting and important, but we feel this is beyond the scope of the current manuscript. It is likely that many factors contribute to the response of cancer cells to these agents, and thus simply surveying some cancer lines for their response and linking this to their p53 status is unlikely to be very informative. Making definitive statements about the contribution of p53, and the differences between wild-type, lossof-function mutants, gain of function mutants, and null mutants will require more extensive analyses and is fertile territory for future studies, in our opinion.

eLife assessment

This important paper reveals that one of the major roles of the WDR5 WIN site is to promote ribosome synthesis, and that by attacking the WIN site with inhibitors ribosome attrition occurs creating new vulnerabilities that can be therapeutically exploited. This deficiency of ribosomal proteins also provokes the p53 response. The data from a variety of approaches is generally very convincing, and together buttresses the authors' conclusions and interpretations quite nicely; overall, this paper will provide a justification for pre-clinical and translational studies of WDR5 interaction site inhibitors.

Reviewer #1 (Public Review):

Building on previous work from the Tansey lab, here Howard et al. characterize transcriptional and translational changes upon WIN site inhibition of WDR5 in MLL-rearranged cancer cells. They first analyze whether C16, a newer generation compound, has the same cellular effects as C6, an early generation compound. Both compounds reduce the expression of WDR5-bound RPGs in addition to the unbound RPG RPL22L1. They then investigate differential translation by ribo-seq and observe that WIN site inhibition reduces the translational RPGs and other proteins related to biomass accumulation (spliceosome, proteasome, mitochondrial ribosome). Interestingly, this reduction adds to the transcriptional changes and is not limited to RPGs whose promoters are bound by WDR5. Quantitative proteomics at two time points confirmed the downregulation of RPGs. Interestingly, the overall effects are modest, but RPL22LA is strongly affected. Unexpectedly, most differentially abundant proteins seem to be upregulated 24 h after C6 (see below). A genetic screen showed that loss of p53 rescues the effect of C6 and C16 and helped the authors to identify pathways that can be targeted by compounds together with WIN site inhibitors in a synergistic way. Finally, the authors elucidated the underlying mechanisms and analyzed the functional relevance of the RPL22, RPL22L1, p53 and MDM4 axis.

Comments on revised version:

The authors have answered my points satisfactorily and the manuscript has become clearer and more meaningful as a result. In particular, the measurement of global translation rate is important and validates the upregulation of a number of proteins following WDR5 inhibitor treatment.

Reviewer #2 (Public Review):

Summary:

The manuscript by Howard et al reports the development of high affinity WDR5-interaction site inhibitors (WINi) that engage the protein to block the arginine-dependent engagement with its partners. Treatment of MLL-rearranged leukemia cells with high-affinity WINi (C16) decreases the expression of genes encoding most ribosomal proteins and other proteins required for translation. Notably, although these targets are enriched for WDR5-ChIP-seq peaks, such peaks are not universally present in the target genes. High concordance was founded between the alterations in gene expression due to C16 treatment and the changes resulting from treatment with an earlier, lower affinity WINi (C6). Besides protein synthesis, genes involved in DNA replication or MYC responses are downregulated, while p53 targets and apoptosis genes are upregulated. Ribosome profiling reveals a global decrease in translational efficiency due to WINi with overall ribosome occupancies of mRNAs ~50% of control samples. The magnitude in the decrements of translation for most individual mRNAs exceeds the respective changes in mRNA levels genome-wide. From these results and other considerations, the authors hypothesize that WINi results in ribosome depletion. Quantitative mass spec documents the decrement in ribosomal proteins following WINi treatment along with increases in p53 targets and proteins involved in apoptosis occurring over 3 days. Notably RPL22L1 is essentially completely lost upon WINi treatment. The investigators next conduct a CRISPR screen to find moderators and cooperators with WINi. They identify components of p53 and DNA repair pathways as mediators of WINi inflicted cell death (so gRNAs against these genes permit cell survival). Next, WINi are tested in combination with a variety of other agents to explore synergistic killing to improve their expected therapeutic efficacy. The authors document loss of the p53 antagonist MDM4 (in combination with splicing alterations of RPL22L1), an observation that supports the notion that WINi killing is p53-mediated.

This is a scientifically very strong and well-written manuscript that applies a variety of state-of-the art molecular approaches to interrogate the role of the WDR5 interaction site and WINi. They reveal that the effects of WINi seem to be focused on the overall synthesis of protein components of the translation apparatus, especially ribosomal proteins-even those that do not bind WDR5 by ChIP (a question left unanswered is how such the WDR5-less genes are nevertheless WINi targeted). They convincingly show that disruption of the synthesis of these proteins occurs upon activation of p53 dependent apoptosis, likely driven by unbalanced ribosomal protein synthesis leading to MDM2 inhibition. This apoptosis is subsequently followed, as expected by ɣH2AX-activation. Pathways of possible WINi resistance and synergies with other anti-neoplastic approaches are explored. These experiments are all well-executed and strongly invite more extensive pre-clinical and translational studies of WINi in animal studies. The studies also may anticipate the use of WINi as probes of nucleolar function and ribosome synthesis though this was not really explored in the current manuscript. The current version of the manuscript documents ribosomal stress revealed by leakage of NPM1 into the nucleoplasm while nucleolar integrity is preserved. A progressive loss of rRNA synthesis occurs upon drug treatment that is presumably secondary to the decrement in ribosomal protein production.

Comments on revised version:

(1) The authors to my mind, have quite nicely and professionally addressed the comments of the reviewers and are to be congratulated on an important contribution to the elucidation of WDR5 biology and pathology.

Author response:

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

eLife assessment

This is a useful study examining the determinants and mechanisms of LRMP inhibi:on of cAMP regula:on of HCN4 channel ga:ng. The evidence provided to support the main conclusions is unfortunately incomplete, with discrepancies in the work that reduce the strength of mechanis:c insights.

Thank you for the reviews of our manuscript. We have made a number of changes to clarify our hypotheses in the manuscript and addressed all of the poten:al discrepancies by revising some of our interpreta:on. In addi:on, we have provided addi:onal experimental evidence to support our conclusions. Please see below for a detailed response to each reviewer comment.

Public Reviews

Reviewer #1 (Public Review):

Summary:

The authors use truncations, fragments, and HCN2/4 chimeras to narrow down the interaction and regulatory domains for LRMP inhibition of cAMP-dependent shifts in the voltage dependence of activation of HCN4 channels. They identify the N-terminal domain of HCN4 as a binding domain for LRMP, and highlight two residues in the C-linker as critical for the regulatory effect. Notably, whereas HCN2 is normally insensitive to LRMP, putting the N-terminus and 5 additional C-linker and S5 residues from HCN4 into HCN2 confers LRMP regulation in HCN2.

Strengths:

The work is excellent, the paper well written, and the data convincingly support the conclusions which shed new light on the interaction and mechanism for LRMP regulation of HCN4, as well as identifying critical differences that explain why LRMP does not regulate other isoforms such as HCN2.

Thank you.

Reviewer #2 (Public Review):

Summary:

HCN-4 isoform is found primarily in the sino-atrial node where it contributes to the pacemaking activity. LRMP is an accessory subunit that prevents cAMP-dependent potentiation of HCN4 isoform but does not have any effect on HCN2 regulation. In this study, the authors combine electrophysiology, FRET with standard molecular genetics to determine the molecular mechanism of LRMP action on HCN4 activity. Their study shows that parts of N- and C-termini along with specific residues in C-linker and S5 of HCN4 are crucial for mediating LRMP action on these channels. Furthermore, they show that the initial 224 residues of LRMP are sufficient to account for most of the activity. In my view, the highlight of this study is Fig. 7 which recapitulates LRMP modulation on HCN2-HCN4 chimera. Overall, this study is an excellent example of using time-tested methods to probe the molecular mechanisms of regulation of channel function by an accessory subunit.

Weaknesses:

(1) Figure 5A- I am a bit confused with this figure and perhaps it needs better labeling. When it states Citrine, does it mean just free Citrine, and "LRMP 1-230" means LRMP fused to Citrine which is an "LF" construct? Why not simply call it "LF"? If there is no Citrine fused to "LRMP 1-230", this figure would not make sense to me.

We have clarified the labelling of this figure and specifically defined all abbreviations used for HCN4 and LRMP fragments in the results section on page 14.

(2) Related to the above point- Why is there very little FRET between NF and LRMP 1-230? The FRET distance range is 2-8 nm which is quite large. To observe baseline FRET for this construct more explanation is required. Even if one assumes that about 100 amino are completely disordered (not extended) polymers, I think you would still expect significant FRET.

FRET is extremely sensitive to distance (to the 6th power of distance). The difference in contour length (maximum length of a peptide if extended) between our ~260aa fragment and our ~130 aa fragments is on the order of 450Å (45nm), So, even if not extended it is not hard to imagine that the larger fragments show a weaker FRET signal. In fact, we do see a slightly larger FRET than we do in control (not significant) which is consistent with the idea that the larger fragments just do not result in a large FRET.

Moreover, this hybridization assay is sensitive to a number of other factors including the affinity between the two fragments, the expression of each fragment, and the orientation of the fluorophores. Any of these factors could also result in reduced FRET.

We have added a section on the limitations of the FRET 2-hybrid assay in the discussion section on page 20. Our goal with the FRET assay was to provide complimentary evidence that shows some of the regions that are important for direct association and we have edited to the text to make sure we are not over-interpreting our results.

(3) Unless I missed this, have all the Cerulean and Citrine constructs been tested for functional activity?

All citrine-tagged LRMP constructs (or close derivatives) were tested functionally by coexpression with HCN (See Table 1 and pages 10-11). Cerulean-tagged HCN4 fragments are of course intrinsically not-functional as they do not include the ion conducting pore.

Reviewer #3 (Public Review):

Summary:

Using patch clamp electrophysiology and Förster resonance energy transfer (FRET), Peters and co-workers showed that the disordered N-terminus of both LRMP and HCN4 are necessary for LRMP to interact with HCN4 and inhibit the cAMP-dependent potentiation of channel opening. Strikingly, they identified two HCN4-specific residues, P545 and T547 in the C-linker of HCN4, that are close in proximity to the cAMP transduction centre (elbow Clinker, S4/S5-linker, HCND) and account for the LRMP effect.

Strengths:

Based on these data, the authors propose a mechanism in which LRMP specifically binds to HCN4 via its isotype-specific N-terminal sequence and thus prevents the cAMP transduction mechanism by acting at the interface between the elbow Clinker, the S4S5-linker, the HCND.

Weaknesses:

Although the work is interesting, there are some discrepancies between data that need to be addressed.

(1) I suggest inserting in Table 1 and in the text, the Δ shift values (+cAMP; + LRMP; +cAMP/LRMP). This will help readers.

Thank you, Δ shift values have been added to Tables 1 and 2 as suggested.

(2) Figure 1 is not clear, the distribution of values is anomalously high. For instance, in 1B the distribution of values of V1/2 in the presence of cAMP goes from - 85 to -115. I agree that in the absence of cAMP, HCN4 in HEK293 cells shows some variability in V1/2 values, that nonetheless cannot be so wide (here the variability spans sometimes even 30 mV) and usually disappears with cAMP (here not).

With a large N, this is an expected distribution. In 5 previous reports from 4 different groups of HCN4 with cAMP in HEK 293 (Fenske et al., 2020; Liao et al., 2012; Peters et al., 2020; Saponaro et al., 2021; Schweizer et al., 2010), the average expected range of the data is 26.6 mV and 39.9 mV for 95% (mean ± 2SD) and 99% (mean ± 3SD) of the data, respectively. As the reviewer mentions the expected range from these papers is slightly larger in the absence of cAMP. The average SD of HCN4 (with/without cAMP) in papers are 9.9 mV (Schweizer et al., 2010), 4.4 mV (Saponaro et al., 2021), 7.6 mV (Fenske et al., 2020), 10.0 mV (Liao et al., 2012), and 5.9 mV (Peters et al., 2020). Our SD in this paper is roughly in the middle at 7.6 mV. This is likely because we used an inclusive approach to data so as not to bias our results (see the statistics section of the revised manuscript on page 9). We have removed 2 data points that meet the statistical classification as outliers, no measures of statistical significance were altered by this.

This problem is spread throughout the manuscript, and the measured mean effects are indeed always at the limit of statistical significance. Why so? Is this a problem with the analysis, or with the recordings?

The exact P-values are NOT typically at the limit of statistical significance, about 2/3rds would meet the stringent P < 0.0001 cut-off. We have clarified in the statistics section (page 10) that any comparison meeting our significance threshold (P < 0.05) or a stricter criterion is treated equally in the figure labelling. Exact P-values are provided in Tables 1-3.

There are several other problems with Figure 1 and in all figures of the manuscript: the Y scale is very narrow while the mean values are marked with large square boxes. Moreover, the exemplary activation curve of Figure 1A is not representative of the mean values reported in Figure 1B, and the values of 1B are different from those reported in Table 1.

Y-axis values for mean plots were picked such that all data points are included and are consistent across all figures. They have been expanded slightly (-75 to -145 mV for all HCN4 channels and -65 to -135 mV for all HCN2 channels). The size of the mean value marker has been reduced slightly. Exact midpoints for all data are also found in Tables 1-3.

The GV curves in Figure 1B (previously Fig. 1A) are averages with the ±SEM error bars smaller than the symbols in many cases owing to relatively high n’s for these datasets. These curves match the midpoints in panel 1C (previously 1B). Eg. the midpoint of the average curve for HCN4 control in panel A is -117.9 mV, the same as the -117.8 mV average for the individual fits in panel B.

We made an error in the text based on a previous manuscript version about the ordering of the tables that has now been fixed so these values should now be aligned.

On this ground, it is difficult to judge the conclusions and it would also greatly help if exemplary current traces would be also shown.

Exemplary current traces have been added to all figures in the revised manuscript.

(3) "....HCN4-P545A/T547F was insensitive to LRMP (Figs. 6B and 6C; Table 1), indicating that the unique HCN4 C-linker is necessary for regulation by LRMP. Thus, LRMP appears to regulate HCN4 by altering the interactions between the C-linker, S4-S5 linker, and Nterminus at the cAMP transduction centre."

Although this is an interesting theory, there are no data supporting it. Indeed, P545 and T547 at the tip of the C-linker elbow (fig 6A) are crucial for LRMP effect, but these two residues are not involved in the cAMP transduction centre (interface between HCND, S4S5 linker, and Clinker elbow), at least for the data accumulated till now in the literature. Indeed, the hypothesis that LRMP somehow inhibits the cAMP transduction mechanism of HCN4 given the fact that the two necessary residues P545 and T547 are close to the cAMP transduction centre, remains to be proven.

Moreover, I suggest analysing the putative role of P545 and T547 in light of the available HCN4 structures. In particular, T547 (elbow) points towards the underlying shoulder of the adjacent subunit and, therefore, is in a key position for the cAMP transduction mechanism. The presence of bulky hydrophobic residues (very different nature compared to T) in the equivalent position of HCN1 and HCN2 also favours this hypothesis. In this light, it will be also interesting to see whether a single T547F mutation is sufficient to prevent the LRMP effect.

We agree that testing this hypothesis would be very interesting. However, it is challenging. Any mutation we make that is involved in cAMP transduction makes measuring the LRMP effect on cAMP shifts difficult or impossible.

Our simple idea, now clarified in the discussion, is that if you look at the regions involved in cAMP transduction (HCND, C-linker, S4-S5), there are very few residues that differ between HCN4 and HCN2. When we mutate the 5 non-conserved residues in the S5 segment and the C-linker, along with the NT, we are able to render HCN2 sensitive to LRMP. Therefore, something about the small sequence differences in this region confer isoform specificity to LRMP. We speculate that this happens because of small structural differences that result from those 5 mutations. If you compare the solved structures of HCN1 and HCN4 (there is no HCN2 structure available), you can see small differences in the distances between key interacting residues in the transduction centre. Also, there is a kink at the bottom of the S4 helix in HCN4 but not HCN1. This points a putatively important residue for cAMP dependence in a different direction in HCN4. We hypothesize in the discussion that this may be how LRMP is isoform specific.

Moreover, previous work has shown that the HCN4 C-linker is uniquely sensitive to di-cyclic nucleotides and magnesium ions. We are hypothesizing that it is the subtle change in structure that makes this region more prone to regulation in HCN4.

Reviewing Editor (recommendations for the Authors):

(1) Exemplar recordings need to be shown and some explanation for the wide variability in the V-half of activation.

Exemplar currents are now shown for each channel. See the response to Reviewer 3’s public comment 2.

(2) The rationale for cut sites in LRMP for the investigation of which parts of the protein are important for blocking the effect of cAMP is not logically presented in light of the modular schematics of domains in the protein (N-term, CCD, post-CCD, etc).

There is limited structural data on LRMP and the HCN4 N-terminus. The cut sites in this paper were determined empirically. We made fragments that were small enough to work for our FRET hybridization approach and that expressed well in our HEK cell system. The residue numbering of the LRMP modules is based on updated structural predictions using Alphafold, which was released after our fragments were designed. This has been clarified in the methods section on pages 5-6 and the Figure 2 legend of the revised manuscript.

(3) Role of the HCN4 C-terminus. Truncation of the HCN4 C-terminus unstructured Cterminus distal to the CNBD (Fig. 4 A, B) partially reverses the impact of LRMP (i.e. there is now a significant increase in cAMP effect compared to full-length HCN4). The manuscript is written in a manner that minimizes the potential role of the C-terminus and it is, therefore, eliminated from consideration in subsequent experiments (e.g. FRET) and the discussion. The model is incomplete without considering the impact of the C-terminus.

We thank the reviewer for this comment as it was a result that we too readily dismissed. We have added discussion around this point and revised our model to suggest that not only can we not eliminate a role for the distal C-terminus, our data is consistent with it having a modest role. Our HCN4-2 chimera and HCN4-S719x data both suggest the possibility that the distal C-terminus might be having some effect on LRMP regulation. We have clarified this in the results (pages 12-13) and discussion (page 19).

(4) For FRET experiments, it is not clear why LF should show an interaction with N2 (residues 125-160) but not NF (residues 1-160). N2 is contained within NF, and given that Citrine and Cerulean are present on the C-terminus of LF and N2/NF, respectively, residues 1-124 in NF should not impact the detection of FRET because of greater separation between the fluorophores as suggested by the authors.

This is a fair point but FRET is somewhat more complicated. We do not know the structure of these fragments and it’s hard to speculate where the fluorophores are oriented in this type of assay. Moreover, this hybridization assay is sensitive to affinity and expression as well. There are a number of reasons why the larger 1-260 fragment might show reduced FRET compared to 125-260. As mentioned in our response to reviewer 2’s public comment 2, we have added a limitation section that outlines the various caveats of FRET that could explain this.

(5) For FRET experiments, the choice of using pieces of the channel that do not correlate with the truncations studied in functional electrophysiological experiments limits the holistic interpretation of the data. Also, no explanation or discussion is provided for why LRMP fragments that are capable of binding to the HCN4 N-terminus as determined by FRET (e.g. residues 1-108 and 110-230, respectively) do not have a functional impact on the channel.

As mentioned in the response to comment 2, the exact fragment design is a function of which fragments expressed well in HEK cells. Importantly, because FRET experiments do not provide atomic resolution for the caveats listed in the revised limitations section on page 20-21, small differences in the cut sites do not change the interpretation of these results. For example, the N-terminal 1-125 construct is analogous to experiments with the Δ1-130 HCN4 channel.

We suspect that residues in both fragments are required and that the interaction involves multiple parts. This is stated in the results “Thus, the first 227 residues of LRMP are sufficient to regulate HCN4, with residues in both halves of the LRMP N-terminus necessary for the regulation” (page 11). We have also added discussion on this on page 21.

(6) A striking result was that mutating two residues in the C-linker of HCN4 to amino acids found in HCN channels not affected by LRMP (P545A, T547F), completely eliminated the impact of LRMP on preventing cAMP regulation of channel activation. However, a chimeric channel, (HCN4-2) in which the C-linker, the CNBD, and the C-terminus of HCN4 were replaced by that of HCN2 was found to be partially responsive to LRMP. These two results appear inconsistent and not reconciled in the model proposed by the authors for how LRMP may be working.

As stated in our answer to your question #3, we have revised our interpretation of these data. If the more distal C-terminus plays some role in the orientation of the C-linker and the transduction centre as a whole, these data can still be viewed consistent with our model. We have added some discussion of this idea in our discussion section.

(7) Replacing the HCN2 N-terminus with that from HCN4, along with mutations in the S5 (MCS/VVG) and C-linker (AF/PT) recapitulated LRMP regulation on the HCN2 background. The functional importance of the S5 mutations is not clear as no other experiments are shown to indicate whether they are necessary for the observed effect.

We have added our experiments on a midpoint HCN2 clone that includes the S5 mutants and the C-linker mutants in the absence of the HCN4 N-terminus (ie HCN2 MCSAF/VVGPT) (Fig. 7). And we have discussed our rationale for the S5 mutations as we believe they may be responsible for the different orientations of the S4-S5 linker in HCN1 and HCN4 structures that are known to impact cAMP regulation.

Reviewer #1 (Recommendations For The Authors):

A) Comments:

(1) Figure 1: Please show some representative current traces.

Exemplar currents are now shown for each channel in the manuscript.

(2) Figure 1: There appears to be a huge number of recordings for HCN4 +/- cAMP as compared to those with LRMP 1-479Cit. How was the number of recordings needed for sufficient statistical power decided? This is particularly important because the observed slowing of deactivation by cAMP in Fig. 1C seems like it may be fairly subtle. Perhaps a swarm plot would make the shift more apparent? Also, LRMP 1-479Cit distributions in Fig. 1B-C look like they are more uniform than normal, so please double-check the appropriateness of the statistical test employed.

We have revised the methods section (page 7) to discuss this, briefly we performed regular control experiments throughout this project to ensure that a normal cAMP response was occurring. Our minimum target for sufficient power was 8-10 recordings. We have expanded the statistics section (page 9) to discuss tests of normality and the use of a log scale for deactivation time constants which is why the shifts in Fig. 1D (revised) are less apparent.

(3) It would be helpful if the authors could better introduce their logic for the M338V/C341V/S345G mutations in the HCN4-2 VVGPT mutant.

See response to the reviewing editor’s comment 7.

B) Minor Comments:

(1) pg. 9: "We found that LRMP 1-479Cit inhibited HCN4 to an even greater degree than the full-length LRMP, likely because expression of this tagged construct was improved compared to the untagged full-length LRMP, which was detected by co-transfection with GFP." Co-transfection with GFP seems like an extremely poor and a risky measure for LRMP expression.

We agree that the exact efficiency of co-transfection is contentious although some papers and manufacturer protocols indicate high co-transfection efficiency (Xie et al., 2011). In this paper we used both co-transfection and tagged proteins with similar results.

(2) pg 9: "LRMP 1-227 construct contains the N-terminus of LRMP with a cut-site near the Nterminus of the predicted coiled-coil sequence". In Figure 2 the graphic shows the coiledcoil domain starting at 191. What was the logic for splitting at 227 which appears to be the middle of the coiled-coil?

See response to the reviewing editor’s comment 2.

(3) Figure 5C: Please align the various schematics for HCN4 as was done for LRMP. It makes it much easier to decipher what is what.

Fig. 5 has been revised as suggested.

(4) pg 12: I assume that the HCN2 fragment chosen aligns with the HCN4 N2 fragment which shows binding, but this logic should be stated if that is the case. If not, then how was the HCN2 fragment chosen?

This is correct. This has been explicitly stated in the revised manuscript (page 14).

(5) Figure 7: Add legend indicating black/gray = HCN4 and blue = HCN2.

This has been stated in the revised figure legend.

(6) pg 17: Conservation of P545 and T547 across mammalian species is not shown or cited.

This sentence is not included in the revised manuscript, however, for the interest of the reviewer we have provided an alignment of this region across species here.

Author response image 1.

Reviewer #2 (Recommendations For The Authors):

(1) It is not clear whether in the absence of cAMP, LRMP also modestly shifts the voltagedependent activity of the channels. Please clarify.

We have clarified that LRMP does not shift the voltage-dependence in the absence of cAMP (page 10). In the absence of cAMP, LRMP does not significantly shift the voltagedependence of activation in any of the channels we have tested in this paper (or in our prior 2020 paper).

(2) Resolution of Fig. 8b is low.

We ultimately decided that the cartoon did not provide any important information for understanding our model and it was removed.

(3) Please add a supplementary figure showing the amino acid sequence of LRMP to show where the demarcations are made for each fragment as well as where the truncations were made as noted in Fig 3 and Fig 4.

A new supplementary figure showing the LRMP sequence has been added and cited in the methods section (page 5). Truncation sites have been added to the schematic in Fig. 2A.

(4) In the cartoon schematic illustration for Fig. 3 and Fig.4, the legend should include that the thick bold lines in the C-Terminal domain represent the CNBD, while the thick bold lines in the N-Terminal domain represent the HCN domain. This was mentioned in Liao 2012, as you referenced when you defined the construct S719X, but it would be nice for the reader to know that the thick bold lines you have drawn in your cartoon indicate that it also highlights the CNBD or the HCN domain.

This has been added to figure legends for the relevant figures in the revised manuscript.

(5) On page 12, missing a space between "residues" and "1" in the parenthesis "...LRMP L1 (residues1-108)...".

Fixed. Thank you.

(6) Which isoform of LRMP was used? What is the NCBI accession number? Is it the same one from Peters 2020 ("MC228229")?

This information has been added to the methods (page 5). It is the same as Peters 2020.

Reviewer #3 (Recommendations For The Authors):

(1) "Truncation of residues 1-62 led to a partial LRMP effect where cAMP caused a significant depolarizing shift in the presence of LRMP, but the activation in the presence of LRMP and cAMP was hyperpolarized compared to cAMP alone (Fig. 3B, C and 3E; Table 1). In the HCN4Δ1-130 construct, cAMP caused a significant depolarizing shift in the presence of LRMP; however, the midpoint of activation in the presence of LRMP and cAMP showed a non-significant trend towards hyperpolarization compared to cAMP alone (Fig. 3C and 3E; Table 1)".

This means that sequence 62-185 is necessary and sufficient for the LRMP effect. I suggest a competition assay with this peptide (synthetic, or co-expressed with HCN4 full-length and LRMP to see whether the peptide inhibits the LRMP effect).

We respectfully disagree with the reviewer’s interpretation. Our results, strongly suggest that other regions such as residues 25-65 (Fig. 3C) and C-terminal residues (Fig. 6) are also necessary. The use of a peptide could be an interesting future experiment, however, it would be very difficult to control relative expression of a co-expressed peptide. We think that our results in Fig. 7E-F where this fragment is added to HCN2 are a better controlled way of validating the importance of this region.

(2) "Truncation of the distal C-terminus (of HCN4) did not prevent LRMP regulation. In the presence of both LRMP and cAMP the activation of HCN4-S719X was still significantly hyperpolarized compared to the presence of cAMP alone (Figs. 4A and 4B; Table 1). And the cAMP-induced shift in HCN4-S719X in the presence of LRMP (~7mV) was less than half the shift in the absence of LRMP (~18 mV)."

On the basis of the partial effects reported for the truncations of the N-terminus of HCN4 162 and 1-130 (Fig 3B and C), I do not think it is possible to conclude that "truncation of the distal C-terminus (of HCN4) did not prevent LRMP regulation". Indeed, cAMP-induced shift in HCN4 Δ1-62 and Δ1-130 in the presence of LRMP were 10.9 and 10.5 mV, respectively, way more than the ~7mV measured for the HCN4-S719X mutant.

As you rightly stated at the end of the paragraph:" Together, these results show significant LRMP regulation of HCN4 even when the distal C-terminus is truncated, consistent with a minimal role for the C-terminus in the regulatory pathway". I would better discuss this minimal role of the C-terminus. It is true that deletion of the first 185 aa of HCN4 Nterminus abolishes the LRMP effect, but it is also true that removal of the very Cterm of HCN4 does affect LRMP. This unstructured C-terminal region of HCN4 contains isotype-specific sequences. Maybe they also play a role in recognizing LRMP. Thus, I would suggest further investigation via truncations, even internal deletions of HCN4-specific sequences.

Please see the response to the reviewing editor’s comment 3.

(3) Figure 5: The N-terminus of LRMP FRETs with the N-terminus of HCN4.

Why didn't you test the same truncations used in Fig. 3? Indeed, based on Fig 3, sequences 1-25 can be removed. I would have considered peptides 26-62 and 63-130 and 131-185 and a fourth (26-185). This set of peptides will help you connect binding with the functional effects of the truncations tested in Fig 3.

Please see the response to the reviewing editor’s comment 2 and 5.

Why didn't you test the C-terminus (from 719 till the end) of HCN4? This can help with understanding why truncation of HCN4 Cterminus does affect LRMP, tough partially (Fig. 4A).

Please see the response to the reviewing editor’s comment 3.

(4) "We found that a previously described HCN4-2 chimera containing the HCN4 N-terminus and transmembrane domains (residues 1-518) with the HCN2 C-terminus (442-863) (Liao et al., 2012) was partially regulated by LRMP (Fig. 7A and 7B)".

I do not understand this partial LRMP effect on the HCN4-2 chimera. In Fig. 6 you have shown that the "HCN4-P545A/T547F was insensitive to LRMP (Figs. 6B and 6C; Table 1), indicating that the unique HCN4 C-linker is necessary for regulation by LRMP". How can be this reconciled with the HCN4-2 chimera? HCN4-2, "containing" P545A/T547F mutations, should not perceive LRMP.

Please see the response to the reviewing editor’s comment 6.

(5) "we next made a targeted chimera of HCN2 that contains the distal HCN4 N-terminus (residues 1-212) and the HCN2 transmembrane and C-terminal domains with 5 point mutants in non-conserved residues of the S5 segment and C-linker elbow (M338V/C341V/S345G/A467P/F469T)......Importantly, the HCN4-2 VVGPT channel is insensitive to cAMP in the presence of LRMP (Fig. 7C and 7D), indicating that the HCN4 Nterminus and cAMP-transduction centre residues are sufficient to confer LRMP regulation to HCN2".

Why did you insert also the 3 mutations of S5? Are these mutations somehow involved in the cAMP transduction mechanism?

You have already shown that in HCN4 only P545 and T547 (Clinker) are necessary for LRMP effect. I suggest to try, at least, the chimera of HCN2 with only A467P/F469T. They should work without the 3 mutations in S5.

Please see the response to the reviewing editor’s comment 7.

eLife assessment

This study identifies the molecular determinants of LRMP co-regulation of HCN 4 activity. The evidence supporting the conclusions, which is compelling, is backed by rigorous electrophysiological and spectroscopic analysis. The work is important because it greatly enhances our understanding of the mechanisms of HCN channel regulation in a tissue-specific manner and highlights a functional role for more disordered regions that have yet to be structurally resolved.

Reviewer #1 (Public Review):

Summary:

The authors use truncations, fragments, and HCN2/4 chimeras to narrow down the interaction and regulatory domains for LRMP inhibition of cAMP-dependent shifts in the voltage dependence of activation of HCN4 channels. They identify the N-terminal domain of HCN4 as a binding domain for LRMP, and highlight two residues in the C-linker as critical for the regulatory effect. Notably, whereas HCN2 is normally insensitive to LRMP, putting the N-terminus and 5 additional C-linker and S5 residues from HCN4 into HCN2 confers LRMP regulation in HCN2.

Strengths:

The work is excellent, the paper well written, and the data convincingly support the conclusions which shed new light on the interaction and mechanism for LRMP regulation of HCN4, as well as identifying critical differences that explain why LRMP does not regulate other isoforms such as HCN2.

Reviewer #2 (Public Review):

Summary:

HCN-4 isoform is found primarily in sino-atrial node where it contributes to the pacemaking activity. LRMP is an accessory subunit which prevents cAMP-dependent potentiation of HCN4 isoform but does not have any effect on HCN2 regulation. In this study, the authors combine electrophysiology, FRET with standard molecular genetics to determine the molecular mechanism of LRMP action on HCN4 activity. Their study shows parts of N- and C-termini along with specific residues in C-linker and S5 of HCN4 are crucial for mediating LRMP action on these channels. Furthermore, they show that the initial 224 residues of LRMP are sufficient to account for most of the activity. In my view, the highlight of this study is Fig. 7 which recapitulates LRMP modulation on HCN2-HCN4 chimera. Overall, this study is an excellent example of using time-tested methods to probe the molecular mechanisms of regulation of channel function by an accessory subunit.

The authors adequately addressed my earlier concerns.

Reviewer #3 (Public Review):

Summary:

Using patch clamp electrophysiology and Förster resonance energy transfer (FRET), Peters and co-workers showed that the disordered N-terminus of both LRMP and HCN4 are necessary for LRMP to interact with HCN4 and inhibit the cAMP-dependent potentiation of channel opening. Strikingly, they identified two HCN4-specific residues, P545 and T547 in the C-linker of HCN4, that are close in proximity to the cAMP transduction centre (elbow Clinker, S4/S5-linker, HCND) and account for the LRMP effect.

Strengths:

Based on these data, the Authors propose a mechanism in which LRMP specifically binds to HCN4 via its isotype-specific Nterminal sequence and thus prevents the cAMP transduction mechanism by acting at the interface between the elbow Clinker, the S4S5-linker, the HCND.

Weaknesses:

Although the work is interesting, there are some discrepancies between data that need to be addressed.

- I suggest inserting in Table 1 and in the text, the Δ shift values (+cAMP; + LRMP; +cAMP/LRMP). This will help readers.

- Figure 1 is not clear, the distribution of values is anomalously high. For instance, in 1B the distribution of values of V1/2 in the presence of cAMP goes from - 85 to -115. I agree that in the absence of cAMP, HCN4 in HEK293 cells shows some variability in V1/2 values, that nonetheless cannot be so wide (here the variability spans sometimes even 30 mV) and usually disappears with cAMP (here not).<br /> This problem is spread throughout the ms, and the measured mean effects indeed always at the limit of statistical significance. Why so? Is this a problem with the analysis, or with the recordings?<br /> There are several other problems with Figure 1 and in all figures of the ms: the Y scale is very narrow while the mean values are marked with large square boxes. Moreover, the exemplary activation curve of Fig 1A is not representative of the mean values reported in Figure 1B, and the values of 1B are different from those reported in Table 1.<br /> On this ground it is difficult to judge the conclusions and it would also greatly help if exemplary current traces would also be shown.

- "....HCN4-P545A/T547F was insensitive to LRMP (Figs. 6B and 6C; Table 1), indicating that the unique HCN4 C-linker is necessary for regulation by LRMP. Thus, LRMP appears to regulate HCN4 by altering the interactions between the C-linker, S4-S5 linker, and N-terminus at the cAMP transduction centre."

Although this is an interesting theory, there are no data supporting it. Indeed, P545 and T547 at the tip of the C-linker elbow (fig 6A) are crucial for LRMP effect, but these two residues are not involved in the cAMP transduction centre (interface between HCND, S4S5 linker and Clinker elbow), at least for the data accumulated till now in the literature. Indeed, the hypothesis that LRMP somehow inhibits the cAMP transduction mechanism of HCN4 given the fact that the two necessary residues P545 and T547 are close to the cAMP transduction centre, awaits to be proven.

Moreover, I suggest analysing the putative role of P545 and T547 in the light of the available HCN4 structures. In particular, T547 (elbow) point towards the underlying shoulder of the adjacent subunit and, therefore, it is in a key position for the cAMP transduction mechanism. The presence of bulky hydrophobic residues (very different nature compared to T) in the equivalent position of HCN1 and HCN2 is also favouring this hypothesis. In this light, it will also be interesting to see whether single T547F mutation is sufficient to prevent LRMP effect.

Author response:

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

Reviewer #1 (Public Review):

Summary:

In this study, Pan DY et al. discovered that the clearance of senescent osteoclasts can lead to a reduction in sensory nerve innervation. This reduction is achieved through the attenuation of Netrin-1 and NGF levels, as well as the regulation of H-type vessels, resulting in a decrease in pain-related behavior. The experiments are well-designed. The results are clearly presented, and the legends are also clear and informative. Their findings represent a potential treatment for spine pain utilizing senolytic drugs.

Strengths:

Rigorous data, well-designed experiments as well as significant innovation make this manuscript stand out.

Weaknesses:

Quantification of histology and detailed statistical analysis will further strengthen this manuscript.

I have the following specific comments.

(1) Since defining senescent cells solely based on one or two markers (SA-β-gal and p16) may not provide a robust characterization, it would be advisable to employ another wellestablished senescence marker, such as γ-H2AX or HMGB1, to corroborate the observed increase in senescent osteoclasts following LSI and aging.

We value the comments provided by the reviewer. In accordance with your suggestion, we have performed co-staining of HMGB1 with Trap in Supplementary Figure 1 to corroborate the observed augmentation of senescent osteoclasts following LSI and aging.

Author response image 1.

(2) The connection between heightened Netrin-1 secretion by senescent osteoclasts following LSI or aging and its relevance to pain warrants thorough discussion within the manuscript to provide a comprehensive understanding of the entire narrative.

We appreciate the reviewer's insightful comments. We have thoroughly addressed the entire narrative in the revised manuscript, as outlined below:

During lumbar spine instability (LSI) or aging, endplates undergo ossification, leading to elevated osteoclast activity and increased porosity1-4. The progressive porous transformation of endplates, accompanied by a narrowed intervertebral disc (IVD) space, is a hallmark of spinal degeneration4,5. Considering that pain arises from nociceptors, it is plausible that low back pain (LBP) may be attributed to sensory innervation within endplates. Additionally, porous endplates exhibit higher nerve density compared to normal endplates or degenerative nucleus pulposus6. Netrin-1, a crucial axon guidance factor facilitating nerve protrusion, has been implicated in this process7-9. The receptor mediating Netrin-1-induced neuronal sprouting, deleted in colorectal cancer (DCC), was found to co-localize with CGRP+ sensory nerve fibers in endplates after LSI surgery10,11. In summary, during LSI or aging, osteoclastic lineage cells secrete Netrin-1, inducing extrusion and innervation of CGRP+ sensory nerve fibers within the spaces created by osteoclast resorption. This Netrin-1/DCC-mediated pain signal is subsequently transmitted to the dorsal root ganglion (DRG) or higher brain levels.

(3) It appears that the quantitative data for TRAP staining in Figure 1j is missing.

We appreciate the reviewer's comments. We have added the statistical data of TRAP staining (Figure. 1p) to Figure 1 in the revised manuscript.

Author response image 2.

(4) Regarding Figure 6, could you please specify which panels were analyzed using a t-test and which ones were subjected to ANOVA? Alternatively, were all the panels in Figure 6 analyzed using ANOVA?

We appreciate the reviewer’s comments here. Upon careful review, we have ensured that quantitative data in panels b, c, and f are analyzed using t-tests, while panels d, e, and g are subjected to one-way ANOVA. These updates have been reflected in the revised figure legend.

Reviewer #2 (Public Review):

Summary:

This manuscript examined the underlying mechanisms between senescent osteoclasts (SnOCs) and lumbar spine instability (LSI) or aging. They first showed that greater numbers of SnOCs are observed in mouse models of LSI or aging, and these SnOCs are associated with induced sensory nerve innervation, as well as the growth of H-type vessels, in the porous endplate. Then, the deletion of senescent cells by administration of the senolytic drug Navitoclax (ABT263) results in significantly less spinal hypersensitivity, spinal degeneration, porosity of the endplate, sensory nerve innervation, and H-type vessel growth in the endplate. Finally, they also found that there is greater SnOCmediated secretion of Netrin-1 and NGF, two well-established sensory nerve growth factors, compared to non-senescent OCs. The study is well conducted and data strongly support the idea. However, some minor issues need to be addressed.

(1) In Figure 2C, "Number of SnCs/mm2", SnCs should be SnOCs.

We apologize for the oversight. This has been rectified in the revised manuscript.

Author response image 3.

(2) In Figure 3A-E, is there any statistical difference between groups Young and Aged+PBS?

We appreciate the reviewer's comments. Following your recommendation, we conducted additional statistical analyses to compare the young and PBS-treated aged mice, and we have incorporated these findings into the revised manuscript. The data reveals a significant increased paw withdrawal frequency (PWF) in aged mice treated with PBS compared with young mice, particularly at 0.4g instead of 0.07g (Figure 3a, 3b). Moreover, aged mice treated with PBS exhibited a significant reduction in both distance traveled and active time when compared to young mice (Figure. 3d, 3e). Additionally, PBS-treated aged mice demonstrated a significantly shortened heat response time relative to young mice (Figure. 3c).

Author response image 4.

(3) Again, is there any statistical difference between the Young and Aged+PBS groups in Figure 4F-K?

We appreciate the reviewer's comments. As per your suggestion, we conducted a thorough analysis to determine the statistical differences between the young and aged+PBS groups, and these statistical results have been implemented in the revised manuscript. The caudal endplates of L4/5 in PBS-treated aged mice exhibited a significant increase in endplate porosity (Figure. 4f) and trabecular separation (Tb.Sp) (Figure. 4g) compared to young mice.

Additionally, PBS-treated aged mice showed a significant elevation in endplate score (Figure. 4h), as well as an increased distribution of MMP13 and ColX within the endplates when compared to young mice (Figure. 4i, 4j). Furthermore, TRAP staining revealed a significant increase in TRAP+ osteoclasts within the endplates of PBS-treated aged mice as compared to young mice (Figure. 4k).

Author response image 5.

(4) What is the figure legend of Figure 7?

The legend for Figure 7 (as below) is included in a separate PDF file labeled 'Figures and Legends.' We have carefully checked the revised manuscript and made sure all the legends are included.

“Fig. 7. (a) Representative images of immunofluorescent analysis of CD31, an angiogenesis marker (green), Emcn, an endothelial cell marker (red) and nuclei (DAPI; blue) of adult sham, LSI and aged mice injected with PBS or ABT263. (b) Quantitative analysis of the intensity mean value of CD31 per mm2 in sham, LSI mice treated with PBS or ABT263. (c) Quantitative analysis of the intensity mean value of CD31 per mm2 in aged mice treated with PBS or ABT263. (d) Quantitative analysis of the intensity mean value of Emcn per mm2 in sham, LSI mice treated with PBS or ABT263. (e) Quantitative analysis of the intensity mean value of Emcn per mm2 in aged mice treated with PBS or ABT263. n ≥ 4 per group. Statistical significance was determined by one-way ANOVA, and all data are shown as means ± standard deviations. “

(5) In "Mice" section, an Ethical code is suggested to be added.

We appreciate the reviewer's comments. In accordance with your suggestion, we have included the Johns Hopkins University animal protocol number in the revised manuscript. The relevant paragraph has been updated to read: “All mice were maintained at the animal facility of The Johns Hopkins University School of Medicine (protocol number: MO21M276).”

(6) In "Methods" section, please indicate the primers of GAPDH.

We apologize for the absence of the GAPDH primers. Upon review, the GAPDH primers used were as follows: forward primer 5'-ATGTGTCCGTCGTGGATCTGA-3' and reverse primer 5'-ATGCCTGCTTCACCACCTTCTT-3'. These primer sequences have been included in the revised manuscript.

(7) Preosteoclasts are regarded to be closely related to H-type vessel growth, so do the authors have any comments on this? Any difference or correlation between SnCs and preosteoclasts?

The pre-osteoclast plays a crucial role in secreting anabolic growth factors that facilitate H-type vessel formation, osteoblast chemotaxis, proliferation, differentiation, and mineralization. The osteoclast represents the terminal differentiation phase, ultimately leading to the induction of resorption.

Senescent cells, including senescent osteoclasts, are characterized by permanent cell cycle arrest and changes in their secretory profile, which can impact their function. In the context of osteoclasts, senescence can lead to a reduction in bone resorption capacity and impaired bone remodeling. Senescent osteoclasts are believed to contribute to age-related bone loss and bonerelated diseases, such as osteoporosis.

Reviewer #3 (Public Review):

Summary:

This research article reports that a greater number of senescent osteoclasts (SnOCs), which produce Netrin-1 and NGF, are responsible for innervation in the LSI and aging animal models.

Strengths:

The research is based on previous findings in the authors' lab and the fact that the IVD structure was restored by treatment with ABT263. The logic is clear and clarifies the pathological role of SnOCs, suggesting the potential utilization of senolytic drugs for the treatment of LBP. Generally, the study is of good quality and the data is convincing.

Weaknesses:

There are some points that can be improved:

(1) Since this work primarily focuses on ABT263, it resembles a pharmacological study for this drug. It is preferable to provide references for the ABT263 concentration and explain how the administration was determined.

Thank you for your comment. ABT263 has been extensively employed in diverse research studies12-15. The concentration and administration of ABT263 followed the protocol outlined in the published paper13. The reference on how to use ABT263 is cited in the method section: “ABT263 was administered to mice by gavage at a dosage of 50 mg per kg body weight per day (mg/kg/d) for a total of 7 days per cycle, with two cycles conducted and a 2-week interval between them39”.

(2) It would strengthen the study to include at least 6 mice per group for each experiment and analysis, which would provide a more robust foundation.

Thank you for your comment here. In response, we conducted a new set of experiments, augmenting the majority of the sample size to six, and updated the corresponding statistical data in the revised manuscript.

(3) In Figure 4, either use "adult" or "young" consistently, but not both. Additionally, it's important to define "sham," "young," and "adult" explicitly in the methods section.

Thank you for your comment. We have addressed the inconsistency in the labeling of Figure 4. Additionally, we have explicitly defined "sham," "young," and "adult" in the methods section as follows: The control group (sham group) for the LSI group refers to C57BL/6J mice that did not undergo LSI surgery, while the control group (young group) for the Aged group refers to 4-month-old C57BL/6J mice.

Author response image 6.

(4) Assess the protein expression of Netrin 1 and NGF.

Thank you for your comment here. We employed ELISA to assess the protein expression of Netrin-1 and NGF in the L3 to L5 endplates. The data revealed that compared to the young sham mice, LSI was associated with significantly greater protein expression of Netrin1 and NGF, which was substantially attenuated by ABT263 treatment in LSI mice (Supplementary Fig. 2a, 2b)

Author response image 7.

Reference

(1) Bian, Q. et al. Excessive Activation of TGFbeta by Spinal Instability Causes Vertebral Endplate Sclerosis. Sci Rep 6, 27093, doi:10.1038/srep27093 (2016).

(2) Bian, Q. et al. Mechanosignaling activation of TGFbeta maintains intervertebral disc homeostasis. Bone Res 5, 17008, doi:10.1038/boneres.2017.8 (2017).

(3) Papadakis, M., Sapkas, G., Papadopoulos, E. C. & Katonis, P. Pathophysiology and biomechanics of the aging spine. Open Orthop J 5, 335-342, doi:10.2174/1874325001105010335 (2011).

(4) Rodriguez, A. G. et al. Morphology of the human vertebral endplate. J Orthop Res 30, 280-287, doi:10.1002/jor.21513 (2012).

(5) Taher, F. et al. Lumbar degenerative disc disease: current and future concepts of diagnosis and management. Adv Orthop 2012, 970752, doi:10.1155/2012/970752 (2012).

(6) Fields, A. J., Liebenberg, E. C. & Lotz, J. C. Innervation of pathologies in the lumbar vertebral end plate and intervertebral disc. Spine J 14, 513-521, doi:10.1016/j.spinee.2013.06.075 (2014).

(7) Hand, R. A. & Kolodkin, A. L. Netrin-Mediated Axon Guidance to the CNS Midline Revisited. Neuron 94, 691-693, doi:10.1016/j.neuron.2017.05.012 (2017).

(8) Moore, S. W., Zhang, X., Lynch, C. D. & Sheetz, M. P. Netrin-1 attracts axons through FAK-dependent mechanotransduction. J Neurosci 32, 11574-11585, doi:10.1523/JNEUROSCI.0999-12.2012 (2012).

(9) Serafini, T. et al. Netrin-1 is required for commissural axon guidance in the developing vertebrate nervous system. Cell 87, 1001-1014, doi:10.1016/s0092-8674(00)81795-x (1996).

(10) Forcet, C. et al. Netrin-1-mediated axon outgrowth requires deleted in colorectal cancer-dependent MAPK activation. Nature 417, 443-447, doi:10.1038/nature748 (2002).

(11) Shu, T., Valentino, K. M., Seaman, C., Cooper, H. M. & Richards, L. J. Expression of the netrin-1 receptor, deleted in colorectal cancer (DCC), is largely confined to projecting neurons in the developing forebrain. J Comp Neurol 416, 201-212, doi:10.1002/(sici)1096-9861(20000110)416:2<201::aid-cne6>3.0.co;2-z (2000).

(12) Born, E. et al. Eliminating Senescent Cells Can Promote Pulmonary Hypertension Development and Progression. Circulation 147, 650-666, doi:10.1161/CIRCULATIONAHA.122.058794 (2023).

(13) Chang, J. et al. Clearance of senescent cells by ABT263 rejuvenates aged hematopoietic stem cells in mice. Nat Med 22, 78-83, doi:10.1038/nm.4010 (2016).

(14) Lim, S. et al. Local Delivery of Senolytic Drug Inhibits Intervertebral Disc Degeneration and Restores Intervertebral Disc Structure. Adv Healthc Mater 11, e2101483, doi:10.1002/adhm.202101483 (2022).

(15) Yang, H. et al. Navitoclax (ABT263) reduces inflammation and promotes chondrogenic phenotype by clearing senescent osteoarthritic chondrocytes in osteoarthritis. Aging (Albany NY) 12, 12750-12770, doi:10.18632/aging.103177 (2020).

eLife assessment

This fundamental study advances our understanding of the role of senescent osteoclasts (SnOCs) in the pathogenesis of spine instability. The authors provide compelling evidence for the SnOCs to induce sensory nerve innervation. Accordingly, reduction of SnOCs by the senolytic drug Navitoclax markedly reduces spinal pain sensitivity. This work will be of broad interest to regenerative biologists working on spinal pain.

Reviewer #1 (Public Review):

Summary:

In this study, Pan DY et al. discovered that the clearance of senescent osteoclasts can lead to a reduction in sensory nerve innervation. This reduction is achieved through the attenuation of Netrin-1 and NGF levels, as well as the regulation of H-type vessels, resulting in a decrease in pain-related behavior. The experiments are well-designed. The results are clearly presented, and the legends are also clear and informative. Their findings represent a potential treatment for spine pain utilizing senolytic drugs.

Strengths:

Rigorous data, well-designed experiments as well as significant innovation make this manuscript stand out.

Weaknesses:

All my concerns have been well addressed, no further comments.

Reviewer #2 (Public Review):

Summary:

This manuscript examined the underlying mechanisms between senescent osteoclasts (SnOCs) and lumbar spine instability (LSI) or aging. They first showed that greater numbers of SnOCs are observed in mouse models of LSI or aging, and these SnOCs are associated with induced sensory nerve innervation, as well as the growth of H-type vessels, in the porous endplate. Then, the deletion of senescent cells by administration of the senolytic drug Navitoclax (ABT263) results in significantly less spinal hypersensitivity, spinal degeneration, porosity of the endplate, sensory nerve innervation, and H-type vessel growth in the endplate. Finally, they also found that there is greater SnOC-mediated secretion of Netrin-1 and NGF, two well-established sensory nerve growth factors, compared to non-senescent OCs. The study is well conducted and data strongly support the idea.

Reviewer #3 (Public Review):

Summary:

This research article reports that a greater number of senescent osteoclasts (SnOCs), which produce Netrin-1 and NGF, are responsible for innervation in the LSI and aging animal models.

Strengths:

The research is based on previous findings in the authors' lab and the fact that the IVD structure was restored by treatment with ABT263. The logic is clear and clarifies the pathological role of SnOCs, suggesting the potential utilization of senolytic drugs for the treatment of LBP. Generally, the study is of good quality and the data is convincing.

Weaknesses:

All my concerns have been well addressed, no further comments.

Author response:

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

Reviewer #1

The authors should include experiments such as Cryo-EM and genetically modified animals to demonstrate the physiological importance of the TMEM81 complex.

While we intend to pursue cryo-EM studies of the putative complex (or subcomplexes thereof), this is clearly not a straightforward endeavor and goes beyond the scope of the present manuscript. Concerning the generation of genetically modified animals, we would like to underline that the majority of the proteins that we used for AlphaFold-Multimer complex predictions were precisely chosen based on the fact that - as detailed in the publications referenced in the Introduction - ablation of the respective genes caused sex-specific infertility due to defects in gamete fusion (the other criterion used for inclusion being structural similarity to IZUMO1 coupled with expression in the testis (IZUMO2-4 and TMEM81), or evidence from other kinds of experiments in the case of human-specific MAIA). Concerning TMEM81, experimental evidence for a direct involvement in gamete fusion is described in the referenced preprint by Daneke et al., which was submitted to bioRxiv concomitantly with the present work.

Reviewer #2

I believe that the manuscript would benefit from the authors providing more information about the systematic search (Figure 4). For example, by indicating for each pair tested the average pDock score in a 2D plot (or table) and as raw data in the supplementary information.

Figure 4 has been modified to report both the top and the mean ranking scores for every interaction. Furthermore, additional metrics for the systematic search summarized in Figure 4, including pDockQ scores, are provided in this manuscript revision as supplementary Table S1.

A global search, such as including all membrane proteins expressed in eggs or sperm, could not only be more informative but could also allow the reader to understand the pDock score discrimination power for this particular subset.

The possibility of carrying out a global search was evaluated by performing preliminary computational experiments on an extended ensemble of sperm and egg proteins. In order to do so, we compiled a list of sperm membrane proteins by referring to 4 proteomic datasets (PMIDs 36384108, 36896575, 31824947, 24082039) and identifying ~600 proteins that were found in at least two of them; among these, 250 were single-pass type I or type II membrane proteins, or GPI-anchored proteins. Similarly, a list of 160 egg surface membrane proteins, excluding multipass and secreted ones, was obtained by comparing oocyte cDNA library NIH_MGC_257_N (Express Genomics, USA) with 4 proteomic datasets (PMIDs 35809850, 36042231, 29025019, 27215607). As we briefly commented at the beginning of the section “Prediction of interactions between human proteins associated with gamete fusion” of the revised manuscript, the tests carried out using the resulting list of sperm and egg proteins suggested that interpreting the results of a global search would be severely complicated by a relatively large number of putative false positives. Moreover, the tests showed that performing a complete systematic search would be beyond our current access to computing power. Based on these observations, we preferred to maintain the present study limited to proteins that had been previously clearly implicated in gamete fusion and/or matched specific structural features of IZUMO1.

Figure 5 could be improved in clarity by schematically indicating to which cell each protein is anchored.

This has been done in the revised version of the manuscript.

Reviewer #3

Major comments

(1) In Figure 1, how the protein of mouse/human IZUMO1 and JUNO is purified is not mentioned in the main text nor in the Methods. Are the mouse IZUMO1-His and mouse JUNO-His transfected together or separately? Are human JUNO-His and human IZUMO1-Myc transfected together into HEK293 cells? And purified by IMAC?

Transfection information has been included in the Methods section “Protein expression, purification and analysis” (previously “Protein expression and purification”). Concerning the purification procedure, we had already stated in the legend of Figure 1 that human JUNOE-His/IZUMO1E-Myc had been purified by IMAC before SEC, and have now done the same for mouse JUNOE-His and IZUMO1E-His.

(2) It would be easier to understand the figure if the author could run a WB to indicate which band above JUNO is specifically IZUMO1-Myc in Figure 1.

This has been done and reported in a new Figure S1 (with the original Figure S1 having now become Figure S2). Details about the antibodies used for immunoblot have been included in both Methods section “Protein expression, purification and analysis” and the Key Resources Table.

(3) Figure 4: Analysis of more proteins that have been suggested as possible candidates for sperm-egg interaction will help to highlight the following results. Also, providing a score for the possibility of interaction might help in selecting those proteins in Figures 5 and 6.

Please refer to the answer to the first question of Reviewer #2.

(4) Figure 7: The authors take advantage of the latest developments in protein structure and interaction to model protein complex formation. However, some experimental experiments such as Co-IP, pull down to support the prediction to verify some of this predicated interaction is necessary.

We agree with the reviewer; however, for the reasons we discussed during our comparison of the biochemical properties of the JUNO/IZUMO1 interaction between mouse and human, pursuing this line of inquiry will likely necessitate an extensive set of parallel experiments using proteins from different species. This work is being planned and will be the focus of future studies. However, as we mentioned at the end of the Abstract, one should also consider that some of these complexes are likely to be highly transient. Because of this, while they may have important regulated roles in vivo (function at a specific time and place), they could be very challenging to detect using standard approaches in vitro. We thus see this as a significant advance that structural modeling could contribute to the identification of such functionally important but transient interactions.

Minor points

(1) In the abstract, "three sperm (IZUMO1, SPACA6 and TMEM81) "should be "three sperm proteins."

The Abstract has been condensed to fit within the suggested 200-word limit and, as part of this, the sentence has been changed to “complex involving sperm IZUMO1, SPACA6, TMEM81 and egg JUNO, CD9”.

(2) How do the predictions of the binary complex IZUMO1/CD9 (Figure S1B) or IZUMO1/CD81 (Figure S1C) suggest "the two egg tetraspanins are interchangeable"? Was it because they are quite similar? Please provide more explanation for this speculation. Interchangeable by function or for complex formation? To support the conclusion, biochemical data is required. Otherwise, it needs to be toned down.

This is because, in the AlphaFold-Multimer predictions of the pentameric complex, CD9 and CD81 are placed in essentially the same way relative to the other subunits.

We have now clarified this at the end of page 6:

“(...) suggest that the two egg tetraspanins are interchangeable because they are predicted to bind to the same region of IZUMO1; (...)”

(3) It would be more reader-friendly if the author could label the name of each protein in the figure in Figure S1, especially when the name is not written in the figure legend.

This has been done in Figure S2 of the revised manuscript (corresponding to original Figure S1).

eLife assessment

This study offers valuable insights into the structural architecture of the mammalian egg-sperm fusion synapse, shedding light on the role of specific proteins in fertilization. The significance of the findings lies in the potential identification of a pentameric complex involved in gamete fusion by AlphaFold Multimer. The strength of evidence for the approach/methodology is solid, while the experimental validation is incomplete in supporting these interactions. This work will be of interest to biomedical researchers working on fertility and reproductive health.

Reviewer #2 (Public Review):

Summary:

Fertilization is a crucial event in sexual reproduction, but the molecular mechanisms underlying egg-sperm fusion remain elusive. Elofsson A et al. used AlphaFold to explore possible synapse-like assemblies between sperm and egg membrane proteins during fertilization. Using a systematic search of protein-protein interactions, the authors proposed a pentameric complex of three sperm (IZUMO1, SPACA6, and TMEM81) and two egg (JUNO and CD9) proteins, providing a new structural model to be used in future structure-function studies.

Strengths:

(1) The study uses the AlphaFold algorithm to predict higher-order assemblies. This approach could offer insights into a highly transient protein complex, which are challenging to detect experimentally.<br /> (2) The article predicts a pentameric complex between proteins involved in fertilization, shedding light on the architectural aspects of the egg-sperm fusion synapse.

Weaknesses:

The proposed model, which is a prediction from a modeling algorithm, lacks experimental validation of the identity of the components and the predicted contacts.

It is noteworthy that in an independent study, Deneke et al. provides experimental evidence of the interaction between IZUMO1/SPACA6/TMEM81 in zebrafish. This is an important element that supports the findings presented in this manuscript

Regarding the authors response on the question of a global search:<br /> I understand that a global search might be difficult to interpret because a large number of putative false positives. But it is this type of information that is needed to assess the validity of the model and the scoring power in the absence of any experimental validation. At minimum, the search should include a negative control set of proteins known to be unrelated to sperm fertilization or homologous egg-sperm fusion complexes from incompatible species to account for species-specific interactions.

I acknowledge that experimentally validating highly transient complexes presents technical hurdles. However, a high-confidence structural model could enable the design of point mutations specifically disrupting the predicted interactions. Subsequent rescue experiments could then validate the directionality of these interactions. Ultimately, such experiments are crucial for robust model validation.

Reviewer #3 (Public Review):

Summary:

Sperm-egg fusion is a critical step in successful fertilization. Although several proteins have been identified in mammals that are required for sperm-egg adhesion and fusion, it is still unclear whether there are other proteins involved in this process and how the reported proteins complex co-operate to complete the fusion process. In this study, the authors first identified TMEM81 as a structural homologue of IZUMO1 and SPACA6, and predicted the interactions with a pool of human proteins associated with gamete fusion, using AlphaFold-Multimer, a recent advance in protein complex structure prediction. The prediction is compelling and well discussed, and the experimental evidence to verify this interaction is lacking in this study but supported by a complementary and independent study by another group.

Strengths:

The authors present a pentameric complex formation of four previously reported proteins involved in egg/sperm interaction together with TMEM181 using a deep learning tool, AlphaFold-Multimer.

Weaknesses:

It is intriguing to see that some of the proteins involved in sperm-egg interaction are successfully predicted to be assembled into a single multimeric structure by AlphaFold-Multimer. The experimental validation of the interactions is not directly supported in this study. As there are more candidate proteins in the process, testing other possible protein interactions more comprehensively will provide more rationale for the current 3D multi-protein modeling.

eLife assessment

This important study sheds light on the mechanisms underlying a rare brain disease, offering insight into the role of microglia in this complex pathophysiology. The evidence presented is solid, utilizing state-of-the-art laboratory models to explore cellular interactions and disease development. While further research is needed, this study will be of interest to neuroscientists and clinicians aiming to understand and combat similar neurodegenerative disorders.

Reviewer #1 (Public Review):

Here, using an organoid system, Wong et al generated a new model of hereditary diffuse leukoencephalopathy with axonal spheroids, with which they investigated how CSF1R-mutaions affect the phenotypes of microglia/macrophages, and revealed metabolic changes in microglia/macrophages associated with a proinflammatory phenotype.

In general, this paper is interesting and well-written, and tackles important issues to be addressed.

This study suffers from several major concerns and limitations that dampen the value of the study. As the authors also mentioned, models that perfectly recapitulate the complexity of the HDLS brain the models would be required to better understand the molecular mechanisms of the disease. In this regard, it is unclear how nicely the organoid system in this study can recapitulate the condition in patients with HDLS (e.g. reduced microglia density, downregulated expression of P2YR12, pathological alterations). In addition, the authors used two different models with distinct mutations that could produce different readouts in CSF1R-mediated cellular responses.

Although the reviewer does understand the importance of providing several options/tools to study rare diseases like HDLS and the difficulty of generating stable organoids with less variation, it is unclear if the different outcomes between HD1 and HD2 are generated through different mutations or simply due to different differentiation efficiency from iMacs (e.g. Figure 2B), which needs to be confirmed. Lastly, there is an over-interpretation regarding the results in Figure 6A. There is no difference between isoHD1 iMac control and HD1 Mut iMac.

Reviewer #2 (Public Review):

Summary:

This paper investigates a rare and severe brain disease called Hereditary Diffuse Leukoencephalopathy with Axonal Spheroids (HDLS). The authors aimed to understand how mutations in the gene CSF-1R affect microglia, the resident immune cells in the brain, and which alterations and factors lead to the specific pathophysiology. To model the human brain with the pathophysiology of HDLS, they used the human-specific model system of induced pluripotent stem cell (iPSC)-derived forebrain organoids with integrated iPSC-derived microglia (iMicro) from patients with the HDLS-causing mutation and an isogenic cell line with the corrected genome. They found that iPSC-derived macrophages (iMac) with HDLS mutations showed changes in their response, including increased inflammation and altered metabolism. Additionally, they studied these iMacs in forebrain organoids, where they differentiate into iMicro, and showed transcriptional differences in isolated iMicro when carrying the HDLS mutation. In addition, the authors described the influence of the mutation within iMicro on the transcriptional level of neurons and neural progenitor cells (NPCs) in the organoid. They observed that the one mutation showed implications for impaired development of neurons, possibly contributing to the progression of the disease.

Overall, this study provides valuable insights into the mechanisms underlying HDLS and emphasizes the importance of studying diseases like these with a suitable model system. These findings, while promising, represent only an initial step towards understanding HDLS and similar neurodegenerative diseases, and thus, their direct translation into new treatment options remains uncertain.

Strengths:

The strength of the work lies in the successful reprogramming of two HDLS patient-derived induced pluripotent stem cells (iPSCs) with different mutations, which is crucial for the study of HDLS using human forebrain organoid models. The use of corrected isogenic iPSC lines as controls increases the validity of the mutation-specific observations. In addition, the model effectively mimics HDLS, particularly concerning deficits in the frontal lobe, mirroring observations in the human brain. Obtaining iPSCs from patients with different CSF1R mutations is particularly valuable given the limitations of rodent and zebrafish models when studying adult-onset neurodegenerative diseases. The study also highlights significant metabolic changes associated with the CSF1R mutation, particularly in the HD2 mutant line, which is confirmed by the HD1 line. In addition, the work shows transcriptional upregulation of the proinflammatory cytokine IL-1beta in cells carrying the mutation, particularly when they phagocytose apoptotic cells, providing further insight into disease mechanisms.

Weaknesses:

The authors have not elucidated the significance of the increased CSF1 dosage in Figure 2F, aside from its effect on cell viability, lacking a thorough discussion of this result. Additionally, while transcriptomic and metabolic alterations related to the mutation were demonstrated in iMac models, similar investigations in iMicros are absent, necessitating further experiments to validate the findings across cell models. The conclusion drawn regarding cytokine levels lacks robust support from the data, particularly considering the varied responses observed in different mutant lines. Further analysis of the secretome (e.g. via ELISA) could provide additional insights. Moreover, the characterization of iMicros is incomplete, with limited protein-level analysis (e.g. validate RNA-seq via flow cytometry). Additionally, the claim of microglial-like morphology lacks adequate evidence, as the provided image is insufficient for such an assessment. RNA-seq experiments should be represented better, it is not possible to read the legends or gene names in the figures. Maybe the data sets can be combined into PCAone and one overall analysis, e.g. via WGCNA-like analyses? This would make it easier for the reader to compare the two cell lines side by side. Furthermore, inaccuracies and omissions in the figure legends compromise the clarity of data representation. Statistical test information is missing. Finally, inconsistent terminology usage throughout the paper may confuse readers (iMac versus iMicros).

eLife assessment

In this study, the authors provide valuable evidence that the LGE is not a significant source of oligodendrocytes for the cortex. The reviewers did find some technical considerations that call for some modulation of the strength of the authors' conclusions and also pointed out some aspects of the data that were incomplete as presented.

Reviewer #1 (Public Review):

Summary:

In this study, the authors generated a novel transgenic mouse line OpalinP2A-Flpo-T2A-tTA2 to specifically label mature oligodendrocytes, and at the same time their embryonic origins by crossing with a progenitor cre mouse line. With this clever approach, they found that LGE/CGE-derived OLs make minimum contributions to the neocortex, whereas MGE/POA-derived OLs make a small but lasting contribution to the cortex. These findings are contradictory to the current belief that LGE/CGE-derived OPCs make a sustained contribution to cortical OLs, whereas MGE/POA-derived OPCs are completely eliminated. Thus, this study provides a revised and more comprehensive view on the embryonic origins of cortical oligodendrocytes. To specifically label mature oligodendrocytes, and at the same time their embryonic origins by crossing with a progenitor cre mouse line. With this clever approach, they found that LGE/CGE-derived OLs make minimum contributions to the neocortex, whereas MGE/POA-derived OLs make a small-but-lasting contribution to to cortex. These findings are contradictory to the current belief that LGE/CGE-derived OPCs make a sustained contribution to cortical OLs, whereas MGE/POA-derived OPCs are completely eliminated. Thus, this study has provided a revised and updated view on the embryonic origins of cortical oligodendrocytes.

Strengths:

The authors have generated a novel transgenic mouse line to specifically label mature differentiated oligodendrocytes, which is very useful for tracing the final destiny of mature myelinating oligodendrocytes. Also, the authors carefully compared the distribution of three progenitor cre mouse lines and suggested that Gsh-cre also labeled dorsal OLs, contrary to the previous suggestion that it only marks LGE-derived OPCs. In addition, the author also analyzed the relative contributions of OLs derived from three distinct progenitor domains in other forebrain regions (e.g. Pir, ac). Finally, the new transgenic mouse lines and established multiple combinatorial genetic models will facilitate future investigations of the developmental origins of distinct OL populations and their functional and molecular heterogeneity.

Weaknesses:

Since OpalinP2A-Flpo-T2A-tTA2 only labels mature oligodendrocytes but not OPCs, the authors can not suggest that the lack of LGE/CGE-derived-OLs in the neocortex is less likely caused by competitive postnatal elimination, but more likely due to limited production and/or allocation (line 118-9). It remains possible that LGE/CGE-derived OPCs migrate into the cortex but are later eliminated.

Reviewer #2 (Public Review):

Summary:

In this manuscript, Cai et al use a combination of mouse transgenic lines to re-examine the question of the embryonic origin of telencephalic oligodendrocytes (OLs). Their tools include a novel Flp mouse for labelling mature oligodendrocytes and a number of pre-existing lines (some previously generated by the last author in Josh Huang's lab) that allowed combinatorial or subtractive labelling of oligodendrocytes with different origins. The conclusion is that cortically-derived OLs are the predominant OL population in the motor and somatosensory cortex and underlying corpus callosum, while the LGE/CGE generates OLs for the piriform cortex and anterior commissure rather than the cerebral cortex. Small numbers of MGE-derived OLs persist long-term in the motor, somatosensory and piriform cortex.

Strengths:

The strength and novelty of the manuscript lies in the elegant tools generated and used and which have the potential to elegantly and accurately resolve the issue of the contribution of different progenitor zones to telencephalic regions.

Weaknesses:

(1) Throughout the manuscript (with one exception, lines 76-78), the authors quantified OL densities instead of contributions to the total OL population (as a % of ASPA for example). This means that the reader is left with only a rough estimation of the different contributions.

(2) All images and quantifications have been confined to one level of the cortex and the potential of the MGE and the LGE/CGE to produce oligodendrocytes for more anterior and more posterior cortical regions remains unexplored.

(3) Hence, the statement that "In summary, our findings significantly revised the canonical model of forebrain OL origins (Figure 4A) and provided a new and more comprehensive view (Figure 4B )." (lines 111, 112) is not really accurate as the findings are neither new nor comprehensive. Published manuscripts have already shown that (a) cortical OLs are mostly generated from the cortex [Tripathi et al 2011 (https://doi.org/10.1523/JNEUROSCI.6474-10.2011), Winker et al 2018 (https://doi.org/10.1523/JNEUROSCI.3392-17.2018) and Li et al (https://doi.org/10.1101/2023.12.01.569674)] and (b) MGE-derived OLs persist in the cortex [Orduz et al 2019 (https://doi.org/10.1038/s41467-019-11904-4) and Li et al 2024 (https://doi.org/10.1101/2023.12.01.569674)]. Extending the current study to different rostro-caudal regions of the cortex would greatly improve the manuscript.

Reviewer #3 (Public Review):

In the manuscript entitled "Embryonic Origins of Forebrain Oligodendrocytes Revisited by Combinatorial Genetic Fate Mapping," Cai et al. used an intersectional/subtractional strategy to genetically fate-map the oligodendrocyte populations (OLs) generated from medial ganglionic eminence (NKX2.1+), lateral ganglionic eminences, and dorsal progenitor cells (EMX1+). Specifically, they generated an OL-expressing reporter mouse line OpalinP2A-Flpo-T2A-tTA2 and bred with region-specific neural progenitor-expressing Cre lines EMX1-Cre for dOL and NKX2.1-Cre for MPOL. They used a subtractional strategy in the OpalinFlp::Emx1Cre::Nkx2.1Cre::RC::FLTG mouse line to predict the origins of OLs from lateral/caudal ganglionic eminences (LC). With their genetic tools, the authors concluded that neocortical OLs primarily consist of dOLs. Although the populations of OLs (dOLs or MP-OLs) from Emx1+ or Nkx2.1+ progenitors are largely consistent with previous findings, they observed that MP-OLs contribute minimally but persist into adulthood without elimination as in the previous report (PMID: 16388308).

Intriguingly, by using an indirect subtraction approach, they hypothesize that both Emx1-negative and Nkx2.1-negative cells represent the progenitors from lateral/caudal ganglionic eminences (LC), and conclude that neocortical OLs are not derived from the LC region. This is in contrast to the previous observation for the contribution of LC-expressing progenitors (marked by Gsx2-Cre) to neocortical OLs (PMID: 16388308). The authors claim that Gsh2 is not exclusive to progenitor cells in the LC region (PMID: 32234482). However, Gsh2 exhibits high enrichment in the LC during early embryonic development. The presence of a small population of Gsh2-positive cells in the late embryonic cortex could originate/migrate from Gsh2-positive cells in the LC at earlier stages (PMID: 32234482). Consequently, the possibility that cortical OLs derived from Gsh2+ progenitors in LC could not be conclusively ruled out. Notably, a population of OLs migrating from the ventral to the dorsal cortical region was detected after eliminating dorsal progenitor-derived OLs (PMID: 16436615).

The indirect subtraction data for LC progenitors drawn from the OpalinFlp-tdTOM reporter in Emx1-negative and Nkx2.1-negative cells in the OpalinFlp::Emx1Cre::Nkx2.1Cre::RC::FLTG mouse line present some caveats that could influence their conclusion. The extent of activity from the two Cre lines in the OpalinFlp::Emx1Cre::Nkx2.1Cre::RC::FLTG mice remains uncertain. The OpalinFlp-tdTOM expression could occur in the presence of either Emx1Cre or Nkx2.1Cre, raising questions about the contribution of the individual Cre lines. To clarify, the authors should compare the tdTOM expression from each individual Cre line, OpalinFlp::Emx1Cre::RC::FLTG or OpalinFlp::Nkx2.1Cre::RC::FLTG, with the combined OpalinFlp::Emx1Cre::Nkx2.1Cre::RC::FLTG mouse line. This comparison is crucial as the results from the combined Cre lines could appear similar to only one Cre line active.

Overall, the authors provided intriguing findings regarding the origin and fate of oligodendrocytes from different progenitor cells in embryonic brain regions. However, further analysis is necessary to substantiate their conclusion about the fate of LC-derived OLs convincingly.

eLife assessment

This useful work describes a novel microscopy-based method to correlate the degree of pigmentation with the gene expression profile of human-induced pluripotent stem cell-derived Retinal Pigmented Epithelial (iPSC-RPE) cells at the single cell level. The presented evidence is solid in showing that there is heterogeneous gene expression in iPSC-derived RPE cells, and there is no significant correlation with the pigmentation. By analyzing the expression of some genes related to function, lysosomal- and complement-related pathways were partially enriched in darker cells. This methodology can be used by other researchers interested in analyzing gene expression related to microscopic images.

eLife assessment

This useful study reports that a week or more of hypoxia exposure in mice increases erythropoiesis and decreases the number of iron-recycling macrophages in the spleen, compromising their capacity for red blood cell phagocytosis – reflected by increased mature erythrocyte retention in the spleen. Compared to an earlier version, the study has been strengthened with mouse experiments under hypobaric hypoxia and complemented by extensive ex vivo analyses. Unfortunately, while some of the evidence is solid, the work as it currently stands only incompletely supports the authors' hypotheses. While the study would benefit from additional experiments that more directly buttress the central claims, it should be of interest to the fields of hemopoiesis and bone marrow biology and possibly also blood cancer.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

This study examined a universal fractal primate brain shape. However, the paper does not seem well structured and is not well written. It is not clear what the purpose of the paper is. And there is a lack of explanation for why the proposed analysis is necessary. As a result, it is challenging to clearly understand what novelty in the paper is and what the main findings are.

We have now restructured the paper, including a summary of the main purpose and findings as follows:

“Compared to previous literature, we can summarise our main contribution and advance as follows:

(i) We are showing for the first time that representative primate species follow the exact same fractal scaling – as opposed to previous work showing that they have a similar fractal dimension [Hofman1985, Hofman1991], i.e. slope, but not necessarily the same offset, as previous methods had no consistent way of comparing offsets.

(ii) Previous work could also not show direct agreement in morphometrics between the coarse-grained brains of primate species and other non-primate mammalian species.

(iii) Demonstrating in proof-of-principle that multiscale morphometrics, in practice, can have much larger effect sizes for classification applications. This moves beyond our previous work where we only showed the scaling law across [Mota2015] and within species [Wang2016], but all on one (native) scale with comparable effect sizes for classification applications [Wang2021].

In simple terms: we know that objects can have the same fractal dimension, but differ greatly in a range of other shape properties. However, we demonstrate here, that representative primate brains and mammalian brain indeed share a range of other key shape properties, on top of agreeing in fractal dimension. This suggests a universal blueprint for mammalian brain shape and a common set of mechanisms governing cortical folding. As a practical additional outcome of our study, we could show that our novel method of deriving multiscale metrics of brain shape can differentiate subtle shape changes much better than the metrics we have been using so far at a single native scale.”

We plan to use the second paragraph as a plain-language summary of our work.

Additionally, several terms are introduced without adequate explanation and contextualization, further complicating comprehension.

We have now made sure that potential jargon is introduced with context and explanation. For example in Introduction: “This scaling law, relating powers of cortical thickness and surface area metrics, […]”

Does the second section, "2. Coarse-graining procedure", serve as an introduction or a method?

We have now renamed this section to “Coarse-graining Method” to indicate that this is a section about methods. However, to describe the methods adequately, we also expanded this section with introductory texts around the history and motivation of the method to provide context and explanations, as the reviewer rightly requested.

Moreover, the rationale behind the use of the coarse-graining procedure is not adequately elucidated. Overall, it is strongly recommended that the paper undergoes significant improvements in terms of its structure, explanatory depth, and overall clarity to enhance its comprehensibility.

To specifically explain the rationale behind the coarse-graining method, we added several clarifications, including the following paragraph:

“As a starting point for such a coarse-graining procedure, we suggest to turn to a well-established method that measures fractal dimension of objects: the so-called box-counting algorithm [Kochunov2007, Madan2019]. Briefly, this algorithm fills the object of interest (say the cortex in our case) with boxes, or voxels of increasingly larger sizes and counts the number of boxes in the object as a function of box size. As the box size increases, the number of boxes decreases; and in a log-log plot, the slope of this relationship indicates the fractal dimension of the object. In our case, this method would not only provide us with the fractal dimension of the cortex, but, with increasing box size, the filled cortex would also contain less and less detail of the folded shape of the cortex. Intuitively, with increasing box size, the smaller details, below the resolution of a single box, would disappear first, and increasingly larger details will follow -- precisely what we require from a coarse-graining method. We therefore propose to expand the traditional box-counting method beyond its use to measure fractal dimension, but to also analyse the reconstructed cortices as different realisations of the original cortex at the specified spatial scale.”

Reviewer #2 (Public Review):

In this manuscript, Wang and colleagues analyze the shapes of cerebral cortices from several primate species, including subgroups of young and old humans, to characterize commonalities in patterns of gyrification, cortical thickness, and cortical surface area. The work builds on the scaling law introduced previously by co-author Mota, and Herculano-Houzel. The authors state that the observed scaling law shares properties with fractals, where shape properties are similar across several spatial scales. One way the authors assess this is to perform a "cortical melting" operation that they have devised on surface models obtained from several primate species. The authors also explore differences in shape properties between the brains of young (~20 year old) and old (~80) humans. My main criticism of this manuscript is that the findings are presented in too abstract a manner for the scientific contribution to be recognized.

We recognise that our work is at the intersection of complex mathematical concepts and a perplexing biological phenomenon. Therefore, our paper has to strike a balance between scientifically accurate and succinct descriptions whilst giving sufficient space to provide context and explanations.

Throughout, we have now added text to provide more context, but also repeat key statements in plain-english terms.

For example, we added the following text to highlight our key contributions.

“In simple terms: we know that objects can have the same fractal dimension, but differ greatly in a range of other shape properties. However, we demonstrate here, that representative primate brains and mammalian brain indeed share a range of other key shape properties, on top of agreeing in fractal dimension. This suggests a universal blueprint for mammalian brain shape and a common set of mechanisms governing cortical folding. As a practical additional outcome of our study, we could show that our novel method of deriving multiscale metrics of brain shape can differentiate subtle shape changes much better than the metrics we have been using so far at a single native scale.”

(1) The series of operations to coarse-grain the cortex illustrated in Figure 1, constitute a novel procedure, but it is not strongly motivated, and it produces image segmentations that do not resemble real brains.

To specifically explain the rationale behind the coarse-graining method, we added several clarifications, including the following paragraph:

“As a starting point for such a coarse-graining procedure, we suggest to turn to a well-established method that measures fractal dimension of objects: the so-called box-counting algorithm [Kochunov2007, Madan2019]. Briefly, this algorithm fills the object of interest (say the cortex in our case) with boxes, or voxels of increasingly larger sizes and counts the number of boxes in the object as a function of box size. As the box size increases, the number of boxes decreases; and in a log-log plot, the slope of this relationship indicates the fractal dimension of the object. In our case, this method would not only provide us with the fractal dimension of the cortex, but, with increasing box size, the filled cortex would also contain less and less detail of the folded shape of the cortex. Intuitively, with increasing box size, the smaller details, below the resolution of a single box, would disappear first, and increasingly larger details will follow -- precisely what we require from a coarse-graining method. We therefore propose to expand the traditional box-counting method beyond its use to measure fractal dimension, but to also analyse the reconstructed cortices as different realisations of the original cortex at the specified spatial scale.”

We also note in several places in the text that the coarse-grained brains are not to be understood as exact reconstructions of actual brains, but serve the purpose of a model:

“[…] nor are the coarse-grained versions of human brains supposed to exactly resemble the location/pattern/features of gyri and sulci of other primates. The similarity we highlighted here are on the level of summary metrics, and our goal was to highlight the universality in such metrics to point towards highly conserved quantities and mechanisms.”

“Note, of course, that the coarse-grained brain surfaces are an output of our algorithm alone and not to be directly/naively likened to actual brain surfaces, e.g. in terms of the location or shape of the folds. Our comparisons here between coarse-grained brains and actual brains is purely on the level of morphometrics across the whole cortex.”

The process to assign voxels in downsampled images to cortex and white matter is biased towards the former, as only 4 corners of a given voxel are needed to intersect the original pial surface, but all 8 corners are needed to be assigned a white matter voxel (section S2). This causes the cortical segmentation, such as the bottom row of Figure 1B, to increase in thickness with successive melting steps, to unrealistic values. For the rightmost figure panel, the cortex consists of several 4.9-sided voxels and thus a >2 cm thick cortex. A structure with these morphological properties is not consistent with the anatomical organization of a typical mammalian neocortex.

Specifically on the point on increasing cortical thickness with increased level of coarse-graining, we have now added the following paragraph:

“The observation that with increasing voxel sizes, the coarse-grained cortices tend to be smoother and thicker is particularly interesting: the scaling law in Eq. 1 can be understood as thicker cortices (T) form larger folds (or are smoother i.e. less surface area At) when brain size is kept constant (Ae). This way of understanding has also been vividly illustrated by using the analogy of forming paper balls with papers of varying thickness in [Mota2015]: to achieve the same size of a paper ball (Ae), the one that uses thicker paper (T) will show larger folds (or is smoother i.e. less surface area At) than the one using thinner paper. The scaling law can therefore be understood as a physically and biologically plausible statement, and here, we are encouraged that our algorithm yields results in line with the scaling law.”

(2) For the comparison between 20-year-old and 80-year-old brains, a well-documented difference is that the older age group possesses more cerebral spinal fluid due to tissue atrophy, and the distances between the walls of gyri becomes greater. This difference is born out in the left column of Figure 4c. It seems this additional spacing between gyri in 80-year-olds requires more extensive down-sampling (larger scale values in Figure 4a) to achieve a similar shape parameter K as for the 20-year-olds. A case could be made that the familiar way of describing brain tissue - cortical volume, white matter volume, thickness, etc. - is a more direct and intuitive way to describe differences between young and old adult brains than the obscure shape metric described in this manuscript. At a minimum, a demonstration of an advantage of the Figure 4a and 4b analyses over current methods for interpreting age-related differences would be valuable.

We have demonstrated the utility of our new shape metrics in a separate paper [Wang2021]. However, we agree with the reviewer that, in this specific instance, it is much easier to understand the key message without considering the less traditional metrics. We have therefore completely revised this part of the Results section to highlight the advantage of multiscale morphometrics, and used the traditional metric of surface area to illustrate the point. The reasoning in surface area is much easier to follow, both visually and conceptually, exactly as the reviewer described.

(3) In Discussion lines 199-203, it is stated that self-similarity, operating on all length scales, should be used as a test for existing and future models of gyrification mechanisms. First, the authors do not show, (and it would be surprising if it were true) that self-similarity is observed for length scales smaller than the acquired MRI data for any of the datasets analyzed. The analysis is restricted to coarse (but not fine)-graining.

To clarify this point, we have added a supplementary section and the following sentence: “Note this method has also no direct dependency on the original MR image resolution, as the inputs are smooth grey and white matter surface meshes reconstructed from the images using strong (bio-)physical assumptions and therefore containing more fine-grained spatial information than the raw images (also see Suppl. Text 3).”

We are indeed sampling at resolutions down to 0.2mm, which is below MR image resolution. The reviewer is, however, correct that we are only coarse-graining, not “fine-graining”. Coarse-graining, here, relates to more coarse than the smooth surface meshes though, not the MR image.

Therefore, self-similarity on all length scales would seem to be too strong a constraint. Second, it is hard to imagine how this test could be used in practice. Specific examples of how gyrification mechanisms support or fail to support the generation of self-similarity across any length scale, would strengthen the authors' argument.

We agree that spatial scales much below 0.2mm resolution may not be of interest, as these scales are only measuring the fractal properties, or “bumpiness”, of the surface meshes at the vertex level. We have therefore revised our statement in Discussion and clarified it with an example: “Finally, this dual universality is also a more stringent test for existing and future models of cortical gyrification mechanisms at relevant scales, and one that moreover is applicable to individual cortices. For example, any models that explicitly simulate a cortical surface could be directly coarse-grained with our method and compared to actual human and primate data provided here.”

Some additional, specific comments are as follows:

(4) The definition of the term A_e as the "exposed surface" was difficult to follow at first. It might be helpful to state that this parameter is operationally defined as the convex hull surface area.

We agree and introduced this term now at first use: “The exposed surface area can be thought of as the surface area of a piece of cling film wrapped around the brain. Mathematically, for the remaining paper it is the convex hull of the brain surface.”

Also, for the pial surface, A_t, there are several who advocate instead for the analysis of a cortical mid-thickness surface area, as the pial surface area is subject to bias depending on the gyrification index and the shape of the gyri. It would be helpful to understand if the same results are obtained from mid-thickness surfaces.

This point is indeed being investigated independently of this study. Our provisional understanding is that in healthy human brains, at native scale, using the mid (or the white matter) surface introduced a systematic offset shift in the scaling law, but does not affect the scaling slope of 1.25. However, this requires a more in-depth investigation in a range of other conditions, and in the context of the coarse-grained shapes, which is on-going. Nevertheless, the scaling law, at first introduction already, has been using the pial surface area [Mota2015] and all subsequent follow-up studies followed this convention. To make our paper here accessible and directly comparable, we therefore used the same metric. Future work will investigate the utility of other metrics.

(5) In Figure 2c, the surfaces get smaller as the coarse-graining increases, making it impossible to visually assess the effects of coarse-graining on the shapes. Why aren't all cortical models shown at the same scale?

The purpose of rescaling the surfaces is to match the scaling plot (Fig 2A) directly, which are showing shrinking surface areas Ae and At with increasing coarse-graining. Here, we are effectively keeping the size of the box constant and resizing the cortical surface instead, which is mathematically equivalent to changing the box size and keeping the cortical surface constant.

An alternative interpretation of the “shrinking” is, therefore, that with increasingly smaller cortical surfaces, the folding details disappear, as we require from our coarse-graining method. This is also visually apparent, as the reviewer points out. We have added this to the explanation in the text.

If we, however, changed the box size instead, the scaling law plot would be meaningless: for example, Ae would barely change with coarse-graining. We would therefore have needed to introduce more complexity in our analysis in terms of how we can measure the scaling law. Thus, we opted to present the simpler method and interpretation here.

Nevertheless, we agree that a direct comparison would be beneficial and have thus added the videos for each species in supplementary under this link: https://bit.ly/3CDoqZQ Upon completed peer-review, we hope to integrate these directly into eLife’s interactive displays for this figure.

(6) Text in Section 3.2 emphasizes that K is invariant with scale (horizontal lines in Figure 3), and asserts this is important for the formation of all cortices. However, I might be mistaken, but it appears that K varies with scale in Figure 4a, and the text indicates that differences in the S dependence are of importance for distinguishing young vs. old brains. Is this an inconsistency?

We agree that it may be confusing to emphasise a “constant K” in the first set of results across species, and then later highlight a changing K in the human ageing results. To clarify, in the first set of results, we find a constant K relative to a changing S: the range in K across melted primate brains is less than 0.1, whereas in S it is over 1.2. In other words, S changes are an order of magnitude higher than K changes. Hence, we described K as “constant” relative to S.

Nevertheless, K shows subtle changes within individuals, which is what we were describing in the human ageing results. These changes are within the range of K values described in the across species results.

However, in the interest of clarity, we followed the reviewer’s suggestion of simplifying the last set of results on human ageing and therefore the variable K in human ageing now only appears in Supplementary. We have now added clarifications to the supplementary on this point.

Reviewer #3 (Public Review):

Summary:

Through a detailed methodology, the authors demonstrated that within 11 different primates, the shape of the brain matched a fractal of dimension 2.5. They enhanced the universality of this result by showing the concordance of their results with a previous study investigating 70 mammalian brains, and the discordance of their results with other folded objects that are not brains. They incidentally illustrated potential applications of this fractal property of the brain by observing a scale-dependent effect of aging on the human brain.

Strengths:

• New hierarchical way of expressing cortical shapes at different scales derived from the previous report through the implementation of a coarse-graining procedure.

Positioning of results in comparison to previous works reinforcing the validity of the observation.

• Illustration of scale-dependence of effects of brain aging in the human.

Weaknesses:

• The impact of the contribution should be clarified compared to previous studies (implementation of new coarse graining procedure, dimensionality of primate brain vs previous studies, and brain aging observations).

We have now made these changes, particularly by adding two paragraphs to the start of Discussion. One summarising the main contributions above previous work, and one paraphrasing the former in plain English for accessibility.

• The rather small sample sizes, counterbalanced by the strength of the effect demonstrated.

We have now increased the sample size of the human ageing analysis substantially to over 100 subjects and observe the same trends, but with an even stronger effect. We therefore believe that this revision serves as an additional internal validation of our data and methods.

• The use of either averaged or individual brains for the different sub-studies could be made clearer.

We have now added this to our Suppl methods: with the exception of the Marmoset, all brain surface data were derived from healthy individual brains.

• The model discussed hypothetically in the discussion is not very clear, and may not be state-of-the-art (axonal tension driving cortical folding? cf. https://doi.org/10.1115/1.4001683).

We have now added this citation to our Discussion and given it context:

“Indeed, our previously proposed model [Mota2015] for cortical gyrification is very simple, assuming only a self-avoiding cortex of finite thickness experiencing pressures (e.g. exerted by white matter pulling, or by CSF pressure). The offset K, or 'tension term', precisely relates to these pressures, leading us to speculate that subtle changes in K correlate with changes in white matter property [Wang2016, Wang2021]. In the same vein of speculation, the scale-dependence of K shown in this work might therefore be related to different types of white matter that span different length scales, such as superficial vs. deep white matter, or U-fibres vs. major tracts. However, there are also challenges to the axonal tension hypothesis [Xu2010]. Indeed, white matter tension differentials in the developed brain may not explain location of folds, but instead white matter tension may contribute to a whole-brain scale 'pressure' during development that drives the folding process overall.”

Reviewer #3 (Recommendations For The Authors):

Many thanks to the authors for this elegant article. I will only report here on the cosmetics of the article.

We thank the reviewer for their kind words and attention to detail and have made all the suggested changes and revised the paper generally for readability, grammar and spelling.

p2: last line of abstract: 'for a range of conditions in the future'.

p3 l.37: I would not self-describe this method as elegant as this is a subjective property .

p3 l.38: 'that will render' -> I wouldn't use the future here.

p.4 l.59: double spacing before ref [9]?

p.6 l.99: 'approximate a fractal' -> why is 'a' italicized?

p.7 fig.2: I would expect the colours to be detailed in the legend. Are there two data points per species because both hemispheres are treated separately?

p.9 l.134-135: 'similar to and in terms of the universal law 'as valid as' -> please add commas for reading comfort: 'similar to, and, in terms of the universal law, 'as valid as'.

p.9 l. 141: For all the cortices we analysed.

p.9 Fig 3: I find the colours a bit confusing in Figs B and C. I find Fig C a bit confusing: what are all the lines representative of, and more specifically, the two lower lines with a different trajectory?

p.10 l.155: '1̃500' -> '~1500'.

p.13 l. 209: either 'speculate that' of 'wonder if'.

p.14 l.232: 'neuron numbers' -> 'number of neurons'.

p.26 S2 second paragraph: 'gryi' -> 'gyri'.

p.30 l.3: please refrain from starting a sentence with I.e..

p.30 last line before S3.2: 'The algorithmic implementation in MATLAB can be found on Zenodo: TBA' - I guess this is linked to you disclosing the code upon acceptance, but please complete before final submission.

p.34 middle/bottom of page: 'The scheme described in Sec. S3.1' -> double spacing before S3.1?

p.35 l.1: 'We simply replace' -> 'we simply replace' (no capital).

p.36 Fig S5.1: explicit the same colouring of the points and boxes in legend

p.38 Fig. S6.1: briefly describe the use of colours in the legend.

p.39 Fig. S7.1: detail colours in the legend.

p.41 Fig. S7.3: detail colours in the legend.

eLife assessment

This study presents valuable framework and findings to our understanding of the brain as a fractal object by observing the stability of its shape property within 11 primate species and by highlighting an application to the effects of aging on the human brain. The evidence provided is solid but the link between brain shape and the underlying anatomy remains unclear. This study will be of interest to neuroscientists interested in brain morphology, and to physicists and mathematicians interested in modeling the shapes of complex objects.

Author response:

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

Reviewer #1:

Figure 1

• The "matched primary tumors" from TCGA include n=424 from cutaneous melanoma; but it is unclear where this is coming from; the PanCan Atlas for melanoma shows n=81 primary and 367 metastatic tumors. There are also additional large cohorts of ICI-treated metastatic tumors with RNAseq data (e.g. a metastatic melanoma cohort with 100+ patients https://doi.org/10.1038/s41591-019-0654-5) that would increase the numbers here.

We thank the reviewer for their observation. We have replaced references to “primary” cancers as “TCGA” cancers as appropriate. While the TCGA analyses included metastatic samples, the majority of the TCGA tumors in most cohorts correspond to primary cancers or local metastases, a point which we added to the text. We retained Fig. 1D as the representative examples are actual primary samples. We have decided to defer analysis of additional melanoma cohorts for future inquiry.

Figure 2

• What is the basis for the split between high and low Dux4 expressing tumors at 1 TPM? Is it arbitrary, or based on some structure in the distribution? (e.g. bimodal distribution)

Our previous analyses of RNA-seq datasets derived from early embryogenesis samples (PMID: 3132774, 28459457) showed that physiologic levels of DUX4 range from approximately 2 to 10 TPM. We added a description in the methods section, under “Genome annotations, gene expression, and Gene Ontology (GO) enrichment analyses,” of our conservative choice for the threshold: DUX4-positivity defined as expression levels > 1 TPM.

Figure 3

• Overall claim is that Dux4 expression is associated with worse survival in metastatic urothelial carcinomas treated with PD-L1 inhibitor. However, the rationale for the choice of split (Dux4 expression < 0.5 and > 1 TPM) to show is unclear (is this the 25th percentile? 75th percentiles?), and the rationale/interpretation of the "partial adjustment" for TMB by removing the bottom quartile of TMB feels non-rigorous and prone to bias. It doesn't feel like Fig 3bc contributes very much; Figure 4 really is the more rigorous analysis.

We thank the reviewer for these comments and suggestions. We adjusted the analyses in Fig. 3C and Fig. S3 to be consistent with Fig. 1 and Fig. 2, in terms of the choice of split. We also clarified in the text how our initial, crude TMB adjustment served as an important indication for us to pursue more rigorous statistical approaches.

Figure 4

• Dux4 expression is independently associated with worse survival considering other clinical and molecular characteristics

• I would include TGFB in the features considered in the table (in the supplementary but not the main table or forest plots, not sure why not?)

• The choice of Dux4 expression split ( < 0.25 and > 1 TPM) feels arbitrary and is different than the split in Figure 3; what is the rationale for this? Also, how many patients does this exclude? (TPM between 0.25 and 1). What does the continuous value or median split for Dux4 expression give you for the CoxPH model?

• Re: building a predictive model, excluding patients (e.g. between <0.25 and > 1 Dux4 TPM) makes the model difficult to apply (e.g. cannot apply to patients with Dux4 levels in the missing interval); a better predictive model would include all patients in the cohort.

We thank the reviewer for their other suggestions. We have clarified in the text that our choice to define DUX4negative samples as those with DUX4 expression levels < 0.25 TPM was made to preemptively address potential misclassifications due to decreased sensitivity of bulk RNA-seq at very low expression levels (PMID: 18516045). We believe our classifications with the new scheme are more reliable. We have also now specified in the text that our categorization excludes 126 patients. We have decided to not pursue the addition of TGFB or exploration of the use of an alternative split or continuous version of DUX4 expression in the Cox Proportional Hazards analyses but appreciate the suggestions, which we will keep in mind for future studies.

Figure 5

• An RSF (randomized survival forest) model predicts survival in Dux4+ vs Dux4- patient, and the Shapley values for landmark time analyses show time-varying effects of different features.

• In some sense, the authors have already demonstrated that Dux4+ is associated with survival differences in ICI treated patients; so a model that predicts survival applied to Dux4+ and Dux4- patients that shows a difference in survival is unsurprising (even in a training/test set setting given that there is a difference in survival across the entire cohort). The quantified marginal effect (from a predictive perspective) of different features is what is interesting here. In that light, I'd like to see more validation of the model up front, specifically how close the predicted survival is to the actual survival of patients (e.g. the survival curves in Fig 5a but with actual survival of the Dux4- and Dux4+ cohorts superimposed on the predicted probabilities).

We thank the reviewer for this suggestion. We have added a plot showing the superimposed survival probability estimates over time for the RSF and KM models for patients assigned to either the test or training sets in Fig. 5.

SFig 5

• Unclear how the authors got estimates of the # of expected deaths associated with covariates (e.g. "...we measured an increase in the number of predicted deaths associated with DUX4-positivity by approximately 16, over DUX4negative status (Fig S5F-G).") from Shapley values as shown in the indicated figure - is this 16 out of the entire cohort? At a given time point? Would recommend perhaps showing the inferred absolute change in mortality (e.g. 8% absolute increase in mortality)

Mortality is the expected number of deaths for the cohort over the observation window, measured as the sum of the CHF over time. We have clarified this in the Methods section, under “Random Survival Forest, feature importance, and partial dependence.” We have also changed the quantification to show the absolute mortality differences comparing patients with DUX4-negative and -positive tumors; we thank the reviewer for this suggestion. We have also clarified in the text that adjusted mortality was estimated via partial dependence, which operates using the correct units, as opposed to Shapley values, where attribution is scaled. Finally, we changed the referenced figure when discussing changes in mortality associated with TMB and DUX4 status (Fig. S5H-I); we appreciate the reviewer pointing out this error.

Figure S1B-C

• The authors argue that Dux4 expression is not an artifact of FFPE tissue by analyzing a mixed tumor cohort sequenced with both poly-A and hybrid probe capture in matched flash-frozen and FFPE tumor samples, showing that it is 1) detectible both FFPE and flash-frozen tissue and 2) higher levels are detected in polyA sequencing/frozen tissue. However, the reference for this section (D. Robinson et al 2015) is a study of a cohort of prostate cancers with polyA bulk RNAseq sequencing; is this correct/is the data coming from a different study?

• Analysis of scRNAseq (if available) would strengthen their analyses by better delineating the expression and response of interferon-gamma and downstream (e.g. antigen presentation) pathways in specific cell compartments, and potential differences in cell-cell interactions (e.g. using CellPhoneDB) associated with Dux4+ vs Dux4- tumors.

• Do the investigators find similar findings in primary and metastatic tumors sequenced the same way (e.g. tcga primary vs met melanoma, albeit most of the met melanoma are Stage III lymph nodes)?

We thank the reviewer for finding the citation error. We have corrected the manuscript to reflect the correct study we analyzed (PMID: 28783718). We also thank the reviewer for their additional suggestions, which undoubtedly would strengthen the current study. However, we have respectfully decided to defer these additional analyses for future study.

Reviewer #2:

It is strange as a statistician to see BIC and AIC represented as barplots, e.g. Figure 4B. There is no knowledge to be gained through this visual representation that would not otherwise be conveyed by just giving the numbers.

We thank the reviewer for this suggestion. We understand that simply stating the numbers would be equally informative. However, we respectfully decided to retain our current versions of Figures 4 and S4 so that the numbers can be illustrated in a visual manner in the figures, rather than just stated in the text.

eLife assessment

This study presents a valuable finding on the association between DUX4 expression with features of immune evasion in human tissue and clinical outcomes in patients with advanced urothelial cancer. The evidence supporting the claims of the authors is convincing, using a range of corroborative statistical techniques. Compared to an earlier version, the quality of the manuscript has been enhanced, for example Figure 5 now illustrates the key features of survival probability estimates over time for patients assigned to with the test or training set.

Reviewer #1 (Public Review):

Pineda et al investigate the association of the hypothesis that Dux4, an embryonic transcription factor, expression in tumor cells is associated with immune evasion and resistance to immunotherapy. They analyze existing cohorts of bulk RNAseq sequenced tumors across cancer types to identify Dux4 expression and association with survival. They find that Dux4 expression is detected in a higher proportion of metastatic tumors compared to primary tumors, is associated with decreased immune infiltrate and a variety of immune metrics and previously nominated immune signatures, and do an in depth evaluation of a cohort of metastatic urothelial cell carcinoma, finding that Dux4 expression is associated with a more immunodeficient tumor microenvironment (desert or excluded microenvironment) and worse survival in this aPDL1 treated cohort. They then find that Dux4 expression is a major independent predictor of survival in this cohort using different types of survival analyses (KM, Cox PH, and random survival forests). With prior existing biological data supporting the hypothesis (in prior work, the senior author has demonstrated Dux4 expression causally suppresses MHC-I expression in interferon-gamma treated cell lines), the current work links Dux4 expression with less immune activity in clinical tumor samples and with survival in ICI treated urothelial carcinomas, and demonstrates that Dux4 expression provides independent information towards survival including other molecular and clinical characteristics (TMB, ECOG PS as the other strongest markers), and provides interesting resolution on landmark analyses with TMB and Dux4 expression providing greater informativeness at later survival landmarks (e.g. 1 year and later), while ECOG PS has strong informativeness already at earlier time points. This work provides impetus towards more mechanistic and functional dissection of the mechanism of Dux4-associated changes with the tumor microenvironment (e.g. in vivo mouse studies) as well as potential interventional studies (e.g. Dux4 as a target in combination therapies). What the work does not provide is additional resolution on the mechanism of how Dux4 may be associated with a more immunodeficient microenvironment.

The conclusions are generally well supported, but there are issues that would benefit from clarification and extension:

- The finding that Dux4 expression is detected in a higher proportion of metastatic tumors and at higher levels compared to TCGA samples (Fig 1BC) is striking. However, at least for one tumor type (melanoma), the TCGA cohort is comprised of mostly locoregional metastatic (n=81 primary and 367 metastatic tumors in the PanCan Atlas). Since there are annotations for primary and (locoregional) metastatic samples in TCGA, an analysis of the primary vs. locoregional metastasis vs distant metastatic samples seems reasonable and likely informative. The analysis of tumors with matched FFPE and flash frozen samples with hybrid probe capture and polyA sequencing, respectively is a nice validation to show that the difference in Dux4 expression is not due to differences in preservation of starting material/sequencing in the metastatic samples vs TCGA samples (S1BC).<br /> - The findings that Dux4 expression in the metastatic urothelial carcinoma setting is associated with a more immunodeficient microenvironment (Figure 2) is clear and unambiguous using multiple lines of data and analyses (bulk RNAseq, DUX4-positive vs DUX4-negative tumors, different immune cell and cytokine signatures; IHC showing an association with immune deserts and immune excluded phenotypes). However, this is an association and does not demonstrate causality.<br /> - The survival analyses (Fig 3,4,5) show fairly convincingly that Dux4 provide independent predictive information beyond clinical variables and TMB towards survival in the aPDL1 treated metastatic urothelial carcinoma cohort. However, the choice to split the cohort into Dux4 negative (defined as < 0.25 TPM) and Dux4 positive (> 1 TPM) while excluding a large number of patients (n=126 pts) that fall in between has significant impact on the rigor of conclusions. This would benefit from showing all the data (e.g. including the 3rd group of in-betweens in the survival analyses as a separate group).<br /> - The authors demonstrate that adding Dux4 to clinical markers and TMB results in an improved predictive model for survival, but there are a few questions regarding this model as a clinical biomarker<br /> o Is Dux4 expression better than other correlated immune signatures/markers (e.g. interferon gamma, T effector signature, overall immune infiltrate) in providing additional information?<br /> - The use of random survival forests to quantify the (predictive) marginal effect of Dux4+ vs Dux4- expression on survival in a non-parametric model as well as shed light on association with survival at different landmark times using Shapley values is quite interesting and well conducted.

Reviewer #2 (Public Review):

Summary:

This article takes an expansive look at the potential role of DUX4 in cancer treatment and prognosis, including its correlation with other key biomarkers, the potential for cancer to be resistant to treatment, and risk prediction.

Strengths:

The primary strength of this work is the breadth of the analyses. The authors have linked DUX4 to not just one but multiple points in the trajectory of cancer, which increases the face validity of their conclusion that DUX4 is meaningfully related to the course of a cancer as well as the prognosis for a patient.

Statistically, the authors have taken care to properly validate their findings using appropriate bootstrapping and testing strategies.

Weaknesses:

Several weaknesses are noted. First, there is little-to-no description of the underlying sample population. It is only stated that "several large cohorts of patients with different metastatic cancers" were analyzed, and that a cohort of patients with advanced urothelial cancer was used for estimating associations with clinical outcomes. Lacking is information on the sampling mechanism, inclusion/exclusion criteria, treatment modalities, the definition of 'time = 0', the number of events observed, or even the sample size. Knowledge about the underlying study design would help explain some counterintuitive results, e.g. that the hazard of death among patients with Stage IV cancer is half that of those with Stage I cancer (Table 1); presumably this is not because Stage IV is actually protective but rather an artifact of the sampling scheme for these data. Second, the definition of negative versus positive DUX4 expression varies throughout the paper. In Figure 2B, Figure 3A, and Figure 3C, it is defined as >1 TPM vs. <= 1 TPM; in Figure 4A and Figure 5A, it is defined as >1 TPM vs. < 0.25 TPM; in Figure S1C it is partitioned into four groups, with boundaries defined at 0.25 TPM, 1 TPM, and 5 TPM. If categorization is needed, a rationale should be provided (ideally prospectively and not based upon the observed data, so as to avoid the perception of forking paths analyses), and it should be consistently applied. Third and finally, data seem to be occasionally excluded without rationale. For example, as mentioned above, the Cox model presented in Figure 4A seems to exclude all patients with DUX4 TPM between 0.25 and 1. Figure 3C excludes patients with TMB in the lowest quartile (although the decision was ostensibly to control for TMB confounding, there are more appropriate ways to do so that don't result in loss of data, e.g. a stratified KM plot). Excluding patients based upon a particular region of the covariate space makes interpreting the resulting model awkward.

Author response:

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

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

• Line 144, after eq. (1). Vectors d_i need to be defined. Are these the mapping of vectors e_i due to the active deformation? It would be useful to state then that d_3 is aligned with r'.

Thank you for your suggestion, and the definition has been added to lines 146-149 for a better understanding of the model.

• Line 144.Authors state a_i(0,0,Z)=0. Shouldn't this be true also for any angle, i.e., a_i(0,Theta, Z)=0?

Thank you, we have revised it in line 144.

• Line 156. G_0 is defined as Diag(1,g_0(t), 1), which seems to be using cylindrical coordinates. Previously, in line 147, vector argument X of \chi is defined with Cartesian coordinates (X,Y,Z). Shouldn't these be also cylindrical?

We are very sorry for this error, our initial configuration is defined with cylindrical coordinates, we have revised it in the manuscript line 151.

• Line 162. "where alpha and beta lie in the range [-pi/2, pi/2]" has already been indicated.

Thank you for your mention, we have deleted duplicate information in line 166.

• Line 171. W is defined as the strain energy density, while in equation (2), symbol W is the total energy (which depends on the previous W). Letters for total elastic and strain energy must be distinguished.

Thank you, we have changed the letter for total energy in Eq.(2).

• Line 176. "we take advantage of the weakness of" -> "we take advantage of the small value of".

We have revised it in line 179.

• Line 177. Why is there a subscript i in p_i? If these do not correspond to penalty p, but to parameters in eqn (3), the latter should have been introduced before this line.

We have revised this error in line 180.

• Line 186. "as the overall elongation \zeta". This parameter, axial extension, has not been defined yet.

Thank you for your mention, the definition of \zeta is now given in line 146.

• Figure 4. Why are the values of g_0 from the elastic model and equations (30)-(32) so non-smooth? Clarify what is being fit and what is the input in the latter equations. Final external radius R_3? Final internal radius R_1'?

(1) To mimic the embryo, we consider a multi-layered cylindrical body so that the shear modulus of each layer is different. The continuity of both deformations and stresses is imposed (see Eq.(26)-Eq.(30). This is the usual treatment for complex morpho-elastic systems. Obviously, $g_0$ originates from the actomyosin cortex so it appears only in the corresponding layer. Finally, all physical quantities such as deformations and stresses must be continuous.

(2) The final outer radius is R_3, which represents the outer radius of C. elegans embryos. In addition to R_3, what we need to consider in this model are R_1’=0.7, R_1’=0.768, R_2=0.8 and R_2’=0.96, these definitions have been added in the caption of Appendix 2—figure 1.

• Line 663, equation (19). Parameter mu is multiplying penalisation term with p, while in equation (2) mu is only affecting the elastic part.

These two different ways of expressing the energy function will ultimately affect the value of p, but the two p are not the same quantities, so they will not affect our results. To avoid misunderstandings, we will replace p in equation (19) with q.

Reviewer #2 (Recommendations For The Authors):

As mentioned in my public summary, I find the writing really not adequate. I provide here a list of specific points that the authors should in my opinion address. As a general comment, I would delete many instances of 'the'.

First, here are figures and whole paragraphs that do not seem to bring anything to the understanding of the phenomenon of C. elegans elongation, notably, Figs. 2, 3C-H, 5m, and 6. Figures 6G and 7 are the only figures containing results it seems. Some elements of the figures are repeated, for example, the illustration of the system's cross-section in Figs 3 and 5.

Thank you for your suggestion, we have made some adjustments to our images to remove some of the duplicate information.

Second, and this is my most important criticism: the mechanism of elongation by releasing elastic stress introduced by muscle contraction is not explained in clear terms anywhere in the text. At least, I was unable to understand it. On p 10 you write "This energy exchange causes the torsion-bending energy to convert into elongation energy, (...)" How this is done is not explained. I assume that the reference state is somehow changed through muscle contraction. The new reference state probably has a longer axis than the one before, but this would then be a plastic deformation and not purely elastic as claimed by the authors (ll 76: "This work aims to answer this paradox within the framework of finite elasticity without invoking cell plasticity (...)"). Is torsion important for this process or is it 'just' another way to store elastic energy in the system?

We perfectly explain most of the exchange of energy between bending, torsion and elongation: indeed, we quantify all aspects of this transformation as the elastic elongation energy, and the dissipation processes which will cost energy. The dissipation evaluated here concerns the rotation of the worm due to the muscle geometry and the viscous friction at the inner surface of the egg. Torsion seems to appear in the late stages and only in some cases. As we show, it comes from a torque induced by the muscles which are not vertical. vertical. Finally, our quantitative predictions of the modelling which recovers most of the experimental published results.

Third, there are a number of strange phrasings and the notation is not helpful in places.

We feel sorry for that, the manuscript is now more precise.

Fourth, the title promises to explain how cyclic muscle contractions reinforce acto-myosin motors. I can't see this done in this work.

The fact that the acto-myosin is reorganized between two sequences of contraction justifies the title. The complete reorganization of the actomyosin network would require a chemico-mechanical model that is not achieved here, perhaps in future work as data become available.

In addition:

We have chosen to respond globally rather than point by point to the referee’s recommendations.

Typographic errors and vocabulary

All English corrections and typos are now included in the main text.

Figures and captions:

Figures and captions have been improved.

• Figure 1: Make the caption and the illustration more coherent. For example, only two cell types are distinguished; in the caption, you mention lateral cells, in the sketch seam cells. What is the difference between acto-myosin and muscle contraction? Muscle contraction is also auto-myosin-based.

(1) The caption for Fig.1 is revised.

(2) From a mechanical point of view, actomyosin bundles in C elegans are orthoradial, whereas muscles are essentially parallel to the main axis of the body are essentially parallel to the main axis of the body, so the geometry is completely different and of extreme importance for deformation. Muscle contractions are quasi-periodic, we do not know the dynamics of the attached molecular motor of myosin. So of course, both contain actin and myosin (not exactly the same proteins), but our model is sensitive to more macroscopic properties.

• Figure 2: I do not find this figure helpful. I might expect such a figure in a grant proposal, but much less in an article.

Figure 2 shows the strategy of our work, we hope that readers can see at a glance what kind of analysis has been done through this figure: since our work is divided into several parts, readers can also unravel the logic through this scheme after reading the whole manuscript. So, this diagram is a guide, and it may be helpful and necessary.

• Figure 3: Figure 3 A, right: What is the dashed line? B You indicate fibers, but your model does not contain fibers, does it? How do I get from the cube to the deformed object? What is the relation of C-H with the rest of the work? Furthermore, you mention seam cells in Fig. 1, but they are absent here. Why can you neglect them? Why introduce them in the first place? E What is a plant vine? F-H What rods are you referring to? Plants do not have muscles, right?

We have modified this figure, and the original Figure 3 now corresponds to Figures 3 and 4.

(1) The dashed line is the centerline after deformation.

(2) The referee is wrong: our model represents the fibers by a higher shear modulus for the actomyosin cortex and for the muscles (see Table Appendix 1) and G_1 reflects the activities of the muscle and actin fibers.

(3) The cube in Figure 3 is a mathematical 3D volume element that is subjected to stresses. Hyperelasticity modelling is based on such a representation.

(4) C-H(new version: Fig.4 A-F): These images show similar deformations: bending and torsion as our C. elegans study. These figures indicate that such deformations are quite common in nature, even if the underlying mechanism is different.

(5) This is a point we have already mentioned: we ignore the difference between the different types of epidermal cells and average their role in the early and second stages of elongation.

(6) The plant vine is the 'botanical vine', see Goriely's article and book.

(7) F-H(new version: Fig.4 D-F) do not have fixed rods, we set a curvature and torsion to fit the actual biological behavior.

(8) Plants do not have muscles, but they grow, and our formalism for growth, pre-strain and material plasticity is very similar to the hyper-elasticity formalism.

• Figure 4: Fig .4 A: "The central or inner part (0 < 𝑅 < 𝑅2, shear modulus 𝜇𝑖) except the muscles which are stiffer." I do not understand.

In the new version, this figure corresponds to Fig.5. The shear modulus of the intrinsic part is very small, but the muscles are harder so we have to consider them separately, we have revised this sentence to avoid misunderstanding.

• Figure 5: Fig 5 A and D: The schematic of the cross-section has appeared already in the previous figure. No need to repeat it here. The same holds for the schematic of the cylindrical embryo. Caption: "But, the yellow region is not an actual tissue layer and it is simply to define the position of muscles." Why do you introduce the yellow region at all? I do not think that it clarifies anything. "Deformation diagram, when left side muscles M_1 and M_2." Something seems to be missing here. Similarly in the next sentence. "the actin fiber orientation changes from the 'loop' to the 'slope'" Do the rings break up and form a helix?

In the new version, this figure corresponds to Fig.6.

(1) We have made revisions to these figures.

(2) The yellow part can show the accurate location of four muscles, which is important for our model and further calculations.

(3) We have revised this sentence in the caption of Fig. 6.

(4) Actin rings do not change to a helix pattern, they will be only sloping.

• Figure 6: Fig 6 A-C These panels do not go beyond Fig 5B. Fig 6D: what are these images supposed to show? They are not really graphs, but microscopy images. The caption is not helpful to understand, what the reader is supposed to see here. Fig 6F: do you really want to plot a linear curve?

In the new version, Fig.5 and Fig.6 respectively correspond to Fig.6 and Fig.7.

(1) Fig.6 shows the simulated images, and Fig.7 A-C is the real calculation results, they are different.

(2) Fig.7 D can show the real condition during C. elegans late elongation, here, we would like to show the torsion of the C. elegans.

(3) Yes, it is our result.

Discussions concerning the biological referee questions:

Ll 75: “how the muscle contractions couple to the acto-myosin activity" Again I find this misleading because muscle contraction relies on auto-myosin activity. Probably, you can find a better expression to refer to the activity of the actomyosin network in the epidermis. Do you propose any mechanism for how muscle contraction increases epidermal contractility? This does not seem to be the mechanism that you propose for elongation, is it?

The actomyosin activity will not stop because of the muscle contraction. Obviously, these two processes cannot be independent. The energy released by a muscle contraction event can and must contribute to the reorganization of the actomyosin network that occurs during the elongation process. Indeed, despite the fact that the embryo elongates, the density of actin cables appears to be maintained, which automatically requires a redistribution of actin monomers. We propose a scenario in which muscle contraction increases actomyosin contractility via energy conversion. We show that after unilateral contraction there is an energy release for this once all dissipation factors are eliminated. We invite the reviewer to re-examine Figure 2 and invite biologists to seriously evaluate the density of molecular motors attached to the circumferential actin cable throughout the stretch process.

Ll 133: "we decide to simplify the geometrical aspect because of the mechanical complexity" This is hardly a justification. Why is it appropriate?

Yes, we would like to offer the reader the simplest modelling with a limiting technicity and a limited number of unknown parameters.

L 135: "active strains" Why not active stress?

The two are equivalent, the choice is dictated by the simplicity of deriving quantitative results for comparison with experiments.

L 170: "hyperelastic" Please, explain this term.

It is the elasticity of very soft samples subjected to large deformations. For classic references, see the books of Ogden, Holzapfel and Goriely, all of which are mentioned in our paper.

Major criticism

Eq. 3 and Ll 227: "𝑝1 is the ratio between the free available myosin population and the attached ones divided by the time of recruitment" Why is the time of recruitment the same for all motors? "inverse of the debonding time" Is it the same as the unbinding rate? Why use the symbol p_2 for it? What is p_3?

The model proposed to justify the increase in the activity of the actomyosin motors during the first phase is a mean-field model: thus all quantities are averaged: we are not considering the theory of a single molecular motor, but a collection in a dynamic environment, so we do not need stochasticity here. Equation (3) concerns the compressive pre-strain, which by definition is a quantity varying between $0$ and $1$ and $X_g=1-G$. ... The debonding time is not the same as the debonding rate. The term $p_3$ indicates saturation and is derived from the law of mass action. The good agreement with the experimental data is shown in Fig.5 (A) and (B). An equivalent model has been developed by (M. Serra et al.).

Serra M, Serrano Nájera G, Chuai M, et al. A mechanochemical model recapitulates distinct vertebrate gastrulation modes[J]. Science Advances, 2023, 9(49)

Ll 275: "This energy exchange causes the torsion-bending energy to convert into elongation energy, leading to a length increase during the relaxation phase, as shown in Fig.1 of Appendix 5." You have posed the puzzle of how contraction leads to elongation, and now that you resolve the puzzle, you simply say that torsion and bending energy are converted into elongation. How? Usually, if I deform an elastic object, it will return to its original configuration after releasing the external forces. Why is this not the case here?

Furthermore, the central result of your work is presented in an Appendix!?

We agree with the referee that an elastic object will return to its initial configuration by releasing stress, i.e. by giving up its accumulated elastic energy to the environment. But the elastic energy has to go somewhere, such as heat. We do not dare to say that the temperature of the worm increases during the muscle contractions.

In fact, the referee's comment also assumes that full relaxation of the stresses is possible, so the object is not a multi-layered specimen and/or it is not enclosed in a box. Most living species are under stress, usually called residual stress. Our skin is under stress. Our fingerprints result from an elastic instability of the epidermis, occurring on foetal life as our brain circumvolutions or our vili. . So, it is obvious that stresses are maintained in multilayered living systems. Closer to the case of C. elegans, the existence of stresses has been demonstrated by experiments with laser ablation fractures in the first stage. The fact that the fractures open proves the existence of stress: if not, there is no opening and only a straight line.

Ll 379: "Although a special focus is made on late elongation, its quantitative treatment cannot avoid the influence of the first stage of elongation due to the acto-myosin network, which is responsible for a prestrain of the embryo." This statement is made repeatedly through the manuscript, but I do not understand, why you could not use an initial state without pre-strain.

This is the basic concept of hyperelasticity. The reference state must be free of stress, so we cannot evaluate the first muscle contraction without treating the first elongation stage.

Grammar, vocabulary and writing errors

ll 31: "the influence of mechanical stresses (...) becomes more complex to be identified and quantified" Is the influence of mechanical stress too complex or too difficult to be identified/quantified?

We have revised it in line 31, “The superposition of mechanical stresses, cellular processes (e.g., division, migration), and tissue organization is often too complex to identify and quantify.”

Ll 41: "The embryonic elongation of C. elegans represents an attractive model of matter reorganization without a mass increase before hatching." Maybe "Embryonic elongation of C. elegans before hatching represents an attractive model of matter reorganization in the absence of growth.".

We have revised it in line 41.

L 42: "It happens after the ventral enclosure (...)" Maybe "It happens after ventral enclosure (...)".

We have revised it in line 42.

Ll 52: "The transition is well defined since the muscle participation makes the embryo rather motile impeding any physical experiments such as laser ablation (...)" Ablation of what?

We have revised it in line 53:The transition is well defined, because the muscle involvement makes the embryo rather motile, and any physical experiments such as laser fracture ablation of the epidermis, which could be performed and achieved in the first period (\cite{vuong2017interplay}), become difficult,.

Ll 59: "a hollow cylinder composed of four parts (seam and dorso-ventral cells)" It is not clear, what the four parts are - in the parenthesis, two are mentioned.

We have revised it in line 59. Fig.1 shows the whole structure, dorsal, ventral and seam cells form four parts of the epidermis.

L 78: "several important issues at this stage remain unsettled" At which stage?

It means the late elongation stage, we have added this information in line 78.

Ll 85: "but how it works at small scales remains a challenge." Maybe "but how it works at small scales remains to be understood.".

We have revised it in line 86.

Ll 99: "the osmolarity of the interstitial fluid" The comes out of the blue. Before you only talked about mechanics, why now osmolarity? Also, the interstitial fluid is only mentioned now. It is important for the dissipative effects that you discuss later, right? If yes, then you should probably introduce it earlier.

For a better understanding, we have change osmolarity into viscosity in line 99.

l 120: "The cortex is composed of three distinct cells" Maybe "distinct cell types".

Thank you, and we have revised it in line 120.

L 121: "cytoskeleton organization and actin network configurations" What is the difference between cytoskeleton organization and actin network configuration? Also, either both should be plural or both singular, I guess.

(1) Cytoskeleton (which involves microtubules) forms the epidermis of C. elegans embryos, and the actin network surrounds the epidermis.

(2) Thank you for your suggestion, we have revised it in line 121.

L 130: "which will be introduced hereafter" Maybe "which will be used hereafter".

We have revised it in line 130.

Ll 148: "The geometric deformation gradient" You usually denote vectors in bold face, so \chi should be bold, right? Define d_i in Eq.(1).

Yes, we have added this information in line 147.

L 172: "auxiliary energy density" Please, explain this term.

We have changed "auxiliary energy density" into "associated energy density" in line 175. Energy density is the amount of energy stored in a given system or region of space per unit volume, the associated energy density in our manuscript can help us to do some calculations.

Ll 188: "Similar active matter can be found in biological systems, from animals to plants as illustrated in Fig.3(C)-(E), they have a structure that generates internal stress/strain when growing or activity. (...)" Why such a general statement during the presentation of the results? The second part of the sentence seems to be incomplete.

Answers: We would like to show our method is general, and can be used in many situations. We have revised the wrong sentence in line 192.

Ll 243: "a bending deformation occurs on the left for active muscles localized on left" Maybe "bending to the left occurs if muscles on the left are activated".

Thank you, we have revised it in line 247.

L 250: "we assume them are perfectly synchronous" Maybe "we assume them to contract simultaneously". We have revised it in line 252.

L 258: "the muscle and acto-myosin activities are assumed to work almost simultaneously." Before it was simultaneously, now only almost!? What does almost mean?

Sorry, we would like to express the same meaning in theses two sentences, we have deleted the word ‘almost’ in line 261.

Ll 294: "one can hypothesize several scenarios" After that, only one scenario is described it seems.

Thank you, we have revised this sentence in line 299.

L 341: "and then is more viscous than water" Maybe "and that is more viscous than water".

We have revised it in line 345.

L 373: "before the egg hatch" Maybe "before the embryo (or larva) hatches"?

We have revised the sentence in line 367.

L 409: "elephant trunk elongated" maybe "elephant trunk elongation".

We have revised it in line 412.

Ll 417: "As one imagines, it is far from triviality (...)" Does this remake help in any way to understand better C. elegans elongation? Also maybe "it is far from trivial".

We have revised it in line 423.

Ll 428: "can map the initial stress-free state B_0 to a state B_1, which reflects early elongation process" Maybe: "maps the initial stress-free state B_0 to a state B_1, which describes early elongation".

We have revised it in line 428.

L 429: "After in the residually stressed (...)" Maybe "Subsequently, we impose an incremental strain filed G_1 that maps the state B_1 to the state B_2, which represents late elongation".

We have revised it in line 429.

l 763: "Modelling details of without pre-strain case" Maybe "Case without pre-strain" or "Modelling in the absence of pre-strain" Similarly for l 784.

We have revised them in line 763 and line 784.

Some questions of definition and understanding

Ll 71: "We can imagine that once the muscle is activated on one side, it can only contract, and then the contraction forces will be transmitted to the epidermis on this side." I do not understand the sentence. Muscle activation leads to contraction, there is nothing to imagine here. Maybe you hypothesize that the muscles are attached to the epidermis such that muscle contraction leads to epidermis deformation?

Yes, four muscle bands are attached to the epidermis, as shown in Fig.1. The deformation does not concern only the epidermis but the whole embryo during the bending events. We have modified the sentence to avoid misunderstanding, the sentence change to “Once the muscle is activated on one side, it can only contract, and then the contraction forces will be transmitted to the epidermis on this side.” in line 71.

Ll 110: "However, it is less widely known that its internal striated muscles share similarities with skeletal muscles found in vertebrates in terms of both function and structure" Is it important for what you report, whether this fact is widely known?

Yes, it is our opinion.

Ll 112: "the role of the four axial muscles (...) is nearly contra-intuitive" Is it or is it not? If yes, why?

Yes it is. Muscles exert contractions, so compressive deformations. Their localization are along the axis of symmetry (up to a small deviation) so they cannot mechanically realize the expected elongation, contrary to the orthoradial actomyosin network.

However, elongation of the C. elegans is observed experimentally, so yes, we think the result contraintuitive.

L 116: "fully heterogeneous cylinder" What is this?

It means that the C. elegans embryo does not have the same elastic properties in different parts (or layers).

L 129: "will collaborate to facilitate further elongation" To facilitate or to drive? If the former, what drives elongation?

Contraction of muscles and actin bundles together drive elongation

Ll 141: "the deformation in each section can be quantified since the circular geometry is lost with the contractions" The deformation could also be quantified if the sections remained circular, right?

Yes. However, circularity is lost during each bending event.

Ll 151: "we need to evaluate the influence of the C. elegans actin network during the early elongation before studying the deformation at the late stage. So, the deformation gradient can be decomposed into: (...) where (...) is the muscle-actomyosin supplementary active strain in the late period" I thought you were now studying the early stage?

In this part, we are outlining how we can study the whole elongation (early and late), not just the early elongation stage. To evaluate the deformation induced by the first contraction of the muscles, we need to know the state of stress of the worm prior to this event, so we also need to recover the early period using the same formalism for the same structure.

L 160: "When considering a filamentary structure with different fiber directions" Which filamentary structure are you talking about?

Fig.3 B shows this model and the filamentary structure, which contains the actin and muscle fibers.

Ll 174: "When the cylinder involves several layers with different shear modulus 𝜇 and different active strains, the integral over 𝑆 covers each layer" I do not understand this sentence. Also, you should probably write 'moduli' instead of modulus.

This implies that when integrating over the whole cross-section S, we need to take into account each layer independently with its own shear modulus and sum the results.

L 176: "weakness of 𝜀" Do you mean \epsilon << 1?

Yes

Ll 178: "Given that the Euler-Lagrange equations and the boundary conditions are satisfied at each order, we can obtain solutions for the elastic strains at zero order 𝐚(𝟎) and at first order 𝐚(𝟏)." Are you thinking about different orders in an \epsilon expansion or the early and the late stages of elongation?

Answers: Different orders are considered only for the late elongation study, the early elongation is treated exactly so do not need a correction in \epsilon.

L 197: "fracture ablation" Please, define.

This is an experiment in which a laser is used to make a cut in a small-scale object of study and then the internal stresses are obtained based on the morphology of the cut, please see the Ref ‘Assessing the contribution of active and passive stresses in C. elegans elongation’. We have added this definition in line 200.

Ll 203: What motivated your choice of notations for the radii R_2'? The inner part of the cylinder is fluid? But above you wrote about a solid cylinder. Why should the inner part be compressible?

(1) We need to define the location of actin cables, which concentrate at the outer periphery.

(2) Our model is a hollow cylinder, and the inner part of the cylinder contains internal organs, tissues, fluids, and so on, so we consider it to be a compressible extremely soft material (Line 213).

Ll 212: "𝑟(𝑅) is the radius after early elongation." And during?

R is variable, r(R) depends on R but also on time t, it represents the radius of C. elegans embryos after the onset of elongation, i.e., after acto-myosin and muscle activities begin.

L 232: \tau_p is probably t_p?

Yes.

L 240: "quite simultaneously" Please, be precise.

In practice, it is difficult to define the concept of simultaneous occurrence unless there is rigorous experimental data to show it, but all we can get in the Ref ‘Remodelage des jonctions sous stress mécanique’, is that it occurs almost simultaneously, which we define as quite simultaneously.

Ll 246: "a short period" What does short mean? Why is it relevant?

From the experimental observations and data, we know that each contraction occurs very rapidly: a few seconds so we define a short period for one contraction.

L 263: "the bending of the model will be increased" Is it really the model that is bent?

Yes, the bending deformation predicted by the model, we have revised in line 266.

Ll 265: "we observed a consistent torsional deformation (Fig.6(E)) that agrees with the patterns seen in the video" In which sense do these configurations agree? I do not see any similarity between panels D and E.

Both show a torsion deformation.

L 267: "torsion as the default of symmetry of the muscle axis" I do not understand.

We discuss two cases in this research, one where the muscle follows the axis of the C. elegans in the initial configuration, and the other where the muscle has a slight angle of deflection, and we have added more information in the manuscript (line 270).

Ll 274: "Each contraction of a pair increases the energy of the system under investigation, which is then rapidly released to the body." Do you mean the elastic energy stored in the epidermis and central part of the embryo?

Yes, the whole body.

Ll 284: "The activation of actin fibers 𝑔𝑎1 after muscle relaxation can be calculated and determined by our model." Have you done it?

Yes, we can obtain the value of g_a1, and then calculate the elongation.

Ll 286 I do not understand, why you write about mutants at this place. Am I supposed to have already understood the basic mechanism of elongation? Why do you now write about the first stage?

I would like to show our formalism can model wild-type and mutant C.elegans, and the comparison results are good.

L 302: "The result is significantly higher than our actual size 210𝜇𝑚." How was significance assessed? Your actual size is probably more than 210µm.

Here, we have considered two situations, one is that the accumulated energy is totally applied to the elongation so that the length will be much larger than the experimental result of 210 µm, the length value that we have obtained by calculation. In the other case, we have considered the energy dissipation, which leads to 210 µm.

L 433: "where 𝜆 is the axial extension due to the pre-strained" Maybe ""where 𝜆 is the axial extension due to the pre-stress".

In our manuscript, we define the pre-strain, not the pre-stress.

L 438: "active filamentary tensor" Please, define.

Active filamentary tensor defines the tensor representing the activities of a cylindrical model composed of different orientations fibers.

Reviewer #1 (Public Review):

The authors have made a novel and important effort to distinguish and include different sources of active deformations for fitting C elegans embryo development: cyclic muscle contractions and actomyosion circumferential stresses. The combination and synchronisation of both contributions are, according to the model, responsible for different elongation rates, and can induce bending and torsion deformations, which are a priori not expected from purely contractile forces. The model can be applied to other growth processes in initially cylindrical shapes.

The tilt of the fibers is an important assumption of the model. However, fiber direction in Figure 3B is not fully clear for explaining the tilting. The fiber in 3B has not very much in common with the fibers in the color part of the figure. Also, is vector m supposed to be tangent to the fiber? In the figure does not seem to be so. It should be expected that alpha is a consequence of the deformation, not as an input parameter, as it seems in the tests of Figure 6A. How is the value of alpha chosen? According to Figure 6, torsion is expected for alpha>0, but for beta=pi/2 and alpha>0 no torsion may be obtained. In fact, it seems that torsion should appear when cos(beta)*sin(alpha)>0. As a consequence, value of beta should be given in Figure 6. Can the amount of torsion be tested as a function of alpha and beta?

The transfer of energy and deformation is a very interesting aspect of the paper, and also crucial for the model and predicting elongation. However, the modelling of this transfer remains very obscure and only explained in the Appendix. Some more details on how the transfer is selected should be given in the main text. Can the transfer of energy interpreted as a change of the relaxed reference configuration? Once a ratio of the energy transferred is fixed, the assumption on elongation distribution should be stated. (Uniformly? ) The authors should also define in the main text the factor g_a1, and explain how this value is computed from condition W_c=W_r .

Given the convoluted shape of the embryo in the egg, contact may be a crucial mechanism for determining growth and torsion. The model does not include this contact, and this limitation should be reflected in the article.

Minor comment:<br /> -Line 300: "we determine the optimal values for the activation parameters". the optimal with respect to which objective? Norm of difference between experimental and computational displacements? How this is quantified needs to be specified.

Reviewer #2 (Public Review):

Summary:

During C. elegans development, embryos undergo elongation of their body axis in absence of cell proliferation or growth. This process relies in an essential way on periodic contractions of two pairs muscles that extend along the embryo's main axis. How contraction can lead to extension along the same direction is unknown.

To address this question, the authors use a continuum description of a multicomponent elastic solid. The various components are the interior of the animal, the muscles, and the epidermis. The different components form separate compartments and are described as hyperelastic solids with different shear moduli. For simplicity, a cylindrical geometry is adopted. The authors consider first the early elongation phase, which is driven by contraction of the epidermis, and then late elongation, where contraction of the muscles injects elastic energy into the system, which is then transferred into elongation. The authors get elongation that can be successfully fitted to the elongation dynamics of wild type worms and two mutant strains.

Strengths:

The work proposes a physical mechanism underlying a puzzling biological phenomenon. The framework developed by the authors could be used to explain phenomena in other organisms and could be exploited in the design of soft robots.

Weaknesses:

(1) The manuscript is hard to read without being very familiar with continuum descriptions of elastic media. This might make the work difficult to access for biologists. This is a real pity because the findings are potentially of great interest to developmental biologists and engineers alike.

(2) The discussion of the worm's mechanical properties could go deeper. The authors hardly justify their assumptions.

Author response:

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

Reviewer #1 (Public Review):

Summary:

This study presents careful biochemical experiments to understand the relationship between LRRK2 GTP hydrolysis parameters and LRRK2 kinase activity. The authors report that incubation of LRRK2 with ATP increases the KM for GTP and decreases the kcat. From this, they suppose an autophosphorylation process is responsible for enzyme inhibition. LRRK2 T1343A showed no change, consistent with it needing to be phosphorylated to explain the changes in G-domain properties. The authors propose that phosphorylation of T1343 inhibits kinase activity and influences monomer-dimer transitions.

Strengths:

The strengths of the work are the very careful biochemical analyses and the interesting result for wild-type LRRK2.

Weaknesses:

A major unexplained weakness is why the mutant T1343A starts out with so much lower activity--it should be the same as wild-type, non-phosphorylated protein. Also, if a monomer-dimer transition is involved, it should be either all or nothing. Other approaches would add confidence to the findings.

We thank the reviewer for these suggestions. We are aware that the T1343A has generally a lower activity compared to the wild type. Therefore, we would like to emphasize that this mutant is the only one not showing an increase in Km values after ATP treatment. Other mutants, also having lower kcat values like T1503A, still show this characteristic change in Km. Our favored explanation for the lower kcat of T1343A is that this mutation lays within a critical region, the so-called ploop, of the Roc domain and is very likely structurally not neutral. Concerning the dimer-monomer transition, we are convinced that there is more than one factor involved in this equilibrium. Most likely, including, but not limited to other LRRK2 domains (e.g. the WD40 domain), binding of co-factors (e.g. Rab29/Rab32 or 14-3-3) and membrane binding. Consistently, also with stapled peptides targeting the Roc or Cor domains we were not able to shift the equilibrium completely to the monomer (Helton et al., ACS Chem Biol. 2021, 16:2326-2338; Pathak et al. ACS Chem Neurosci. 2023, 14(11):1971-1980) We have addressed these points in a revised version of the manuscript.

Reviewer #2 (Public Review):

This study addresses the catalytic activity of a Ras-like ROC GTPase domain of LRRK2 kinase, a Ser/Thr kinase linked to Parkinson's disease (PD). The enzyme is associated with gain-of-function variants that hyper-phosphorylate substrate Rab GTPases. However, the link between the regulatory ROC domain and activation of the kinase domain is not well understood. It is within this context that the authors detail the kinetics of the ROC GTPase domain of pathogenic variants of LRRK2, in comparison to the WT enzyme. Their data suggest that LRRK2 kinase activity negatively regulates the ROC GTPase activity and that PD variants of LRRK2 have differential effects on the Km and catalytic efficiency of GTP hydrolysis. Based on mutagenesis, kinetics, and biophysical experiments, the authors suggest a model in which autophosphorylation shifts the equilibrium toward monomeric LRRK2 (locked GTP state of ROC). The authors further conclude that T1343 is a crucial regulatory site, located in the P-loop of the ROC domain, which is necessary for the negative feedback mechanism. Unfortunately, the data do not support this hypothesis, and further experiments are required to confirm this model for the regulation of LRRK2 activity.

Specific comments are below:

• Although a couple of papers are cited, the rationale for focusing on the T1343 site is not evident to readers. It should be clarified that this locus, and perhaps other similar loci in the wider ROCO family, are likely important for direct interactions with the GTP molecule.

To clarify this point: We, have not only have focused on this specific locus, but instead systematically mutated all known auto-phosphorylation sites with the RocCOR domain (see. supplemental information). Furthermore, it has been shown that this site, at least in the RCKW (Roc to WD40) construct, is quantitatively phosphorylated (Deniston et al., Nature 2020, 588:344-349). We are aware that the T1343 residue is located within the p-loop and that this can impact nucleotide binding capacities (see response to reviewer 1).

We have clarified and addressed these points in a revised version of the manuscript.

• Similar to the above, readers are kept in the dark about auto-phosphorylation and its effects on the monomer/dimer equilibrium. This is a critical aspect of this manuscript and a major conceptual finding that the authors are making from their data. However, the idea that auto-phosphorylation is (likely) to shift the monomer/dimer equilibrium toward monomer, thereby inactivating the enzyme, is not presented until page 6, AFTER describing much of their kinetics data. This is very confusing to readers, as it is difficult to understand the meaning of the data without a conceptual framework. If the model for the LRRK2 function is that dimerization is necessary for the phosphorylation of substrates, then this idea should be presented early in the introduction, and perhaps also in the abstract. If there are caveats, then they should be discussed before data are presented. A clear literature trail and the current accepted (or consensus) mechanism for LRRK2 activity is necessary to better understand the context for these data.

We agree on the reviewer’s opinion. We have revised the introduction accordingly and added a paragraph on page 3 starting from line 27.

• Following on the above concepts, I find it interesting that the authors mention monomeric cytosolic states, and kinase-active oligomers (dimers??), with citations. Again here, it would be useful to be more precise. Are dimers (oligomers?) only formed at the membrane? That would suggest mechanisms involving lipid or membrane-attached protein interactions. Also, what do the authors mean by oligomers? Are there more than dimers found localized to the membrane?

There are multiple studies that have shown that LRRK2 is mainly monomeric in the cytosol while it forms mainly dimeric or higher oligomeric states at membrane (James et al., Biophys. J. 2012, 102, L41–L43; Berger et al., Biochemistry, 2010, 49, 5511–5523). However, we agree with the reviewer that it remains to be determined if the dimeric form is the most active state at the membrane, or a higher oligomeric state. Espescially since a recent study shows that LRRK2 can form active tetramers only when bound to Rab29 (Zhu et al., bioRxiv, 2022, DOI: 10.1101/2022.04.26.489605). We have clarified these points in the introduction of the revised version of the manuscript (page 3, line 27ff).

• Fig 5 is a key part of their findings, regarding the auto-phosphorylation induced monomer formation of LRRK2. From these two bar graphs, the authors state unequivocally that the 'monomer/dimer equilibrium is abolished', and therefore, that the underlying mechanism might be increased monomerization (through maintenance of a GTP-locked state). My view is that the authors should temper these conclusions with caveats. One is that there are still plenty of dimers in the auto-phosphorylated WT, and also in the T1343A mutant. Why is that the case? Can the authors explain why only perhaps a 10% shift is sufficient? Secondly, the T1343A mutant appears to have fewer overall dimers to begin with, so it appears to readers that 'abolition' is mainly due to different levels prior to ATP treatment at 30 deg. I feel these various issues need to be clarified in a revised manuscript, with additional supporting data. Finally, on a minor note, I presume that there are no statistically significant differences between the two sets of bar graphs on the right panel. It would be wise to place 'n.s.' above the graphs for readers, and in the figure legend, so readers are not confused.

Starting with the monomer-dimer equilibrium we are convinced that there is more than the phosphorylation of T1343 (see response to reviewer 1). Therefore a 10% shift in our assay most likely underestimate the effect seen in cells. Consistently, the T1343A mutants show a similar increase in Rab10 phosphorylation assay as the G2019S mutant. This thus shows that the identified feedback mechanism plays an important role in a cellular context. We have addressed this point in the revised manuscript on page 6, line 8ff. As long as the significance indicators in the bar charts are concerned, we agree with reviewer. In order not to overload the figure, we finally decided to include all pairwise comparisons (post-hoc tests) in the supplement.

• Figure 6B, Westerns of phosphorylation, the lanes are not identified and it is unclear what these data mean.

We apologize for this mistake and have added the correct labeling in the revised version of the manuscript.

eLife assessment

This valuable manuscript reports on the relationship between GTP hydrolysis parameters and kinase activity of LRRK2, which is associated with Parkinson's disease. The authors provide a detailed accounting of the catalytic efficiency of the ROC GTPase domain of pathogenic variants of LRRK2, in comparison with the wild-type enzyme. The authors propose that phosphorylation of T1343 inhibits kinase activity and influences monomer-dimer transitions, but the experimental evidence is currently incomplete.

Reviewer #1 (Public Review):

Summary:

This study presents careful biochemical experiments to understand the relationship between LRRK2 GTP hydrolysis parameters and LRRK2 kinase activity. The authors report that incubation of LRRK2 with ATP increases the KM for GTP and decreases the kcat. From this they suppose an autophosphorylation process is responsible for enzyme inhibition. LRRK2 T1343A showed no change, consistent with it needing to be phosphorylated to explain the changes in G-domain properties. The authors propose that phosphorylation of T1343 inhibits kinase activity and influences monomer-dimer transitions.

Strengths:

Strengths of the work are the very careful biochemical analyses and interesting result for wild type LRRK2.

Weaknesses:

The conclusions related to involvement of a monomer-dimer transition are to this reviewer, premature and an independent method needs to be utilized to bolster this aspect of the story.

Reviewer #2 (Public Review):

As discussed in the original review, this manuscript is an important contribution to a mechanistic understanding of LRRK2 kinase. Kinetic parameters for the GTPase activity of the ROC domain have been determined in the absence/presence of kinase activity. A feedback mechanism from the kinase domain to GTP/GDP hydrolysis by the ROC domain is convincingly demonstrated through these kinetic analyses. However, a regulatory mechanism directly linking the T1343 phospho-site and a monomer/dimer equilibrium is not fully supported. The T1343A mutant has reduced catalytic activity and can form similar levels of dimer as WT. The revised manuscript does point out that other regulatory mechanisms can also play a role in kinase activity and GTP/GDP hydrolysis (Discussion section). The environmental context in cells cannot be captured from the kinetic assays performed in this manuscript, and the introduction contains some citations regarding these regulatory factors. This is not a criticism, the detailed kinetics here are rigorous, but it is simply a limitation of the approach. Caveats concerning effects of membrane localization, Rab/14-3-3 proteins, WD40 domain oligomers, etc... should be given more prominence than a brief (and vague) allusion to 'allosteric targeting' near the end of the Discussion.

Specific comments

(1) The revised version is better organized with respect to the significance of monomer/dimer equilibrium and the relevance of the GTP-binding region of ROC domain that encompasses the T1343 phospho-site. The relevance of monomers/dimers of LRRK2 from previous studies is better articulated and readers are able to follow the reasoning for the various mutations.

(2) As a suggestion I would change the following on page 6 to clarify for readers:<br /> "...would show no change in kcat and KM values upon in vitro ATP treatment" to:<br /> "...would show no change in kcat and KM values for GTP hydrolysis upon in vitro ATP treatment"

(3) The levels of dimer in WT (+ATP) and T1343A (+/- ATP) are the same, about 40-45%. These data are cited when the authors state that ATP-induced monomerization is 'abolished' (page 6). My suggestion is to re-phrase this conclusion for consistency with data (Fig 5). For example, one can state that 'ATP incubation does not affect the percentage of dimer for the T1343A variant of LRRK2'. This would be similar to the authors' description of these data on page 8 - 'no difference in dimer formation upon ATP treatment'.

eLife assessment

In this valuable study, the authors examine the role of the cytoskeletal regulatory protein Abba in governing the process of cell genesis in the developing cortex. This study provides insights into the mechanisms of microcephaly, a developmental malformation. The evidence supporting the study was felt to be solid, but the reviewers did note some technical weaknesses that limit the strength of some of the interpretations.

eLife assessment

This is an important study on changes in newborns' neural abilities to distinguish auditory signals at 37 weeks of gestation. The evidence of change in neural discrimination as a function of gestational age is convincing, but further analysis of the acoustic signals and description of the infants' language environment would strengthen the interpretation of the results. The work contributes to the field of neurodevelopment and suggests potential clinical applications in neurodevelopmental disorders.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

The author studies a family of models for heritable epigenetic information, with a focus on enumerating and classifying different possible architectures. The key aspects of the paper are:

• Enumerate all 'heritable' architectures for up-to 4 constituents.

• A study of whether permanent ("genetic") or transient ("epigenetic") perturbations lead to heritable changes

• Enumerated the connectivity of the "sequence space" formed by these heritable architectures

• Incorporating stochasticity, the authors explore stability to noise (transient perturbations)

• A connection is made with experimental results on C elegans.

The study is timely, as there is a renewed interest in the last decade in non-genetic, heritable heterogeneity (e.g., from single-cell transcriptomics). Consequently, there is a need for a theoretical understanding of the constraints on such systems. There are some excellent aspects of this study: for instance, the attention paid to how one architecture "mutates" into another. Unfortunately, the manuscript as a whole does not succeed in formalising nor addressing any particular open questions in the field. Aside from issues in presentation and modelling choices (detailed below), it would benefit greatly from a more systematic approach rather than the vignettes presented.

Despite being foundational, this work was systematic in that (1) for the simple architectures modeled using ordinary differential equations (ODEs) with continuity assumptions, parameters that support steady states were systematically determined for each architecture and then every architecture was explored using genetic changes exhaustively, although epigenetic perturbations were not examined exhaustively because of their innumerable variety; and (2) for the more realistic modeling of architectures as Entity-Sensor-Property systems, the behavior of systems with respect to architecture as well as parameter space that lead to particular behaviors (persistence, heritable epigenetic change, etc.) was systematically explored. A more extensive exploration of parameter space that also includes the many ways that the interaction between any two entities/nodes could be specified using an equation is a potentially ever-expanding challenge that is beyond the scope of any single paper.

Specific aspects that remain to be addressed include the application of multiple notions of heritability to real networks of arbitrary size, considering different types of equations for change of each entity/node, and classifying different behavioral regimes for different sets of parameters.

The key contribution of the paper is an articulation of the crucial questions to ask of any regulatory architecture in living systems rather than the addressing of any question that a field has recognized as ‘open’. Specifically, through the exhaustive listing of small regulatory architectures that can be heritable and the systematic analysis of arbitrary Entity-Sensor-Property systems that more realistically capture regulatory architectures in living systems, this work points the way to constrain inferences after experiments on real living systems. Currently, most experimental biologists engaged in reductionist approaches and some systems biologists examining the function or prevalence of network motifs do not explicitly constrain their models for heritability or persistence. It is hoped that this paper will raise awareness in both communities and lead to more constrained models that minimize biases introduced by incomplete knowledge of the network, which is always the case when analyzing living systems.

Terminology

The author introduces a terminology for networks of interacting species in terms of "entities" and "sensors" -- the former being nodes of a graph, and the latter being those nodes that receive inputs from other nodes. In the language of directed graphs, "entities" would seem to correspond to vertices, and "sensors" those vertices with positive indegree and outdegree. Unfortunately, the added benefit of redefining accepted terminology from the study of graphs and networks is not clear.

The Entities-Sensors-Property (ESP) framework is based on underlying biology and not graph theory, making an ESP system not entirely equivalent to a network or graph, which is much less constrained. The terms ‘entity’, ‘sensor’, and ‘property’ were defined and justified in a previous paper (Jose, J R. Soc. Interface, 2020). While nodes of a network can be parsed arbitrarily and the relationship between them can also be arbitrary, entities and sensors are molecules or collections of molecules that are constrained such that the sensors respond to changes in particular properties of other entities and/or sensors. When considered as digraphs, sensors can be seen as vertices with positive indegree and outdegree. The ESP framework can be applied across any scale of organization in living systems and this specific way of parsing interactions also discretizes all changes in the values of any property of any entity. In short, ESP systems are networks, but not all networks are ESP systems. Therefore, the results of network theory that remain applicable for ESP systems need further investigation.

The key utility of the ESP framework is that it is aligned with the development of mechanistic models for the functions of living systems while being consistent with heredity. In contrast, widely analyzed networks like protein-interaction networks, signaling networks, gene regulatory networks, etc., are not always constrained using these principles.

Model

The model seems to suddenly change from Figure 4 onwards. While the results presented here have at least some attempt at classification or statistical rigour (i.e. Fig 4 D), there are suddenly three values associated with each entity ("property step, active fraction, and number"). Furthermore, the system suddenly appears to be stochastic. The reader is left unsure what has happened, especially after having made the effort to deduce the model as it was in Figs 1 through 3. No respite is to be found in the SI, either, where this new stochastic model should have been described in sufficient detail to allow one to reproduce the simulation.

The Supplementary Information section titled ‘Simulation of simple ESP systems’ provides the requested detailed information and revisions to the writing provide the biologically grounded justification for parsing interacting regulators as ESP systems.

Perturbations

Inspired especially by experimental manipulations such as RNAi or mutagenesis, the author studies whether such perturbations can lead to a heritable change in network output. While this is naturally the case for permanent changes (such as mutagenesis), the author gives convincing examples of cases in which transient perturbations lead to heritable changes. Presumably, this is due the the underlying multistability of many networks, in which a perturbation can pop the system from one attractor to another.

Unfortunately, there appears to be no attempt at a systematic study of outcomes, nor a classification of when a particular behaviour is to be expected. Instead, there is a long and difficult-to-read description of numerical results that appear to have been sampled at random (in terms of both the architecture and parameter regime chosen). The main result here appears to be that "genetic" (permanent) and "epigenetic" (transient) perturbations can differ from each other -- and that architectures that share a response to genetic perturbation need not behave the same under an epigenetic one. This is neither surprising (in which case even illustrative evidence would have sufficed) nor is it explored with statistical or combinatorial rigour (e.g. how easy is it to mistake one architecture for another? What fraction share a response to a particular perturbation?)

As an additional comment, many of the results here are presented as depending on the topology of the network. However, each network is specified by many kinetic constants, and there is no attempt to consider the robustness of results to changes in parameters.

The systematic study of all arbitrary regulatory architectures is beyond the scope of this paper and, indeed, beyond the scope of any one paper. Nevertheless 225,000 arbitrary Entity-Sensor-Property systems were systematically explored and collections of parameters that lead to different behaviors provided (e.g., 78,285 are heritable). These ESP systems more closely mimic regulation in living systems than the coupled ODE-based specification of change in a regulatory architecture.

The example questions raised here are not only difficult to answer, but subjective and present a moving target for future studies. One, ‘how easy is it to mistake one architecture for another?’. Mistaking one architecture for another clearly depends on the number of different types of experiments one can perform on an architecture and the resolution with which changes in entities can be measured to find distinguishing features. Two, ‘What fraction share a response to a particular perturbation?’. ‘Sharing a response’ also depends on the resolution of the measurement after perturbation.

DNA analogy

At two points, the author makes a comparison between genetic information (i.e. DNA) and epigenetic information as determined by these heritable regulatory architectures. The two claims the author makes are that (i) heritable architectures are capable of transmitting "more heritable information" than genetic sequences, and (ii) that, unlike DNA, the connectivity (in the sense of mutations) between heritable architectures is sparse and uneven (i.e. some architectures are better connected than others).

In both cases, the claim is somewhat tenuous -- in essence, it seems an unfair comparison to consider the basic epigenetic unit to be an "entity" (e.g., an entire transcription factor gene product, or an organelle), while the basic genetic unit is taken to be a single base-pair. The situation is somewhat different if the relevant comparison was the typical size of a gene (e.g., 1 kb).

Considering every base being the unit of stored information in the DNA sequence results in the maximal possible storage capacity of a genome of given length. Any other equivalence between entity and units within the genome (e.g., 1 kb gene) will only reduce the information stored in the genome.

Nevertheless, the claim was modified to say that the information content of an ESP system can [italics added] be more extensive than the information content of the genome. This accounts for the possibility of an organism that has an inordinately large genome such that maximal information that can be stored in a particular genome sequence exceeds that stored in a particular configuration of all the contents in a cell.

I thank the reviewer for providing further explanation of this misunderstanding in the second round of review, which helps draw future readers to the sections in the paper that discusses this important point (also see response to Recommendations for the authors).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

I thank the author for their efforts in replying to the comments. I have updated my review accordingly; in particular, I have:

(1) Removed my complaint that Heritability is nowhere defined

(2) Removed issues with the presentation of the ODE model in the supplementary information.

I thank the reviewer for raising these issues and acknowledging the improvements made.

However, given that the manuscript is broadly unchanged from the initial one, many of my prior comments remain justified. Some key points:

(1) The manuscript continues to be difficult to read, for the same reasons as I mentioned when reviewing the paper previously.

(2) The utility of the "ESP" formalism is still unclear.

• As the author notes, continuous ODEs are of course an idealisation of a system with discrete copy number.

• However, discussing this is standard fare in any textbook dealing with chemical dynamics and stochastic processes -- see, for instance, the standard textbook by van Kampen.

• This seems little reason to reject ODEs and implement a poorly defined formalism/simulation scheme.

(3) The author claims that many questions raised are "beyond the scope of this study". Indeed, answering all of these questions are beyond the scope of any one study. However, as I initially wrote, the paper would be much stronger if it focused on a particular problem rather than the many vignettes depicted.

The broad scope of this foundational paper necessitates addressing many issues, which may make it a difficult read for some readers. I hope that future work where each paper focuses on one of the aspects raised here will enable the extensive treatment of limited scope as suggested by the reviewer.

The utility of ODEs is much appreciated and was indeed a computationally efficient way of exploring the vast space of regulatory architectures. As stated in the response to the public reviews, the Entity-Sensors-Property framework provides a biologically grounded way of parsing interacting regulators. This approach is aligned with the development of mechanistic models for the functions of living systems while being consistent with heredity. In contrast, widely analyzed networks like protein-interaction networks, signaling networks, gene regulatory networks, etc., are not always constrained using these principles.

On a final note, on the subject of the comparison with DNA:

Perhaps I have misunderstood something. I simply meant that comparing the "maximal information" with 4 HRAs (12.45 bits) is certainly more than the "maximal information" with 4 basepairs (8 bits), but definitely less than the "maximal information" for four 1-kb genes (4^(4000) combinations, so 8000 bits...)

Perhaps the author means that the growth in information of HRAs is faster than exponential. If so, that should be shown and then remarked on.

For this reason, I maintain my comment that the comparison is tenuous.

This issue was addressed once in the results section and again in the discussion section.

The results section states that “The combinatorial growth in the numbers of HRAs with the number of interactors can thus provide vastly more capacity for storing information in larger HRAs compared to that afforded by the proportional growth in longer genomes.”

The discussion section states that “Despite imposing heritability, regulated non-isomorphic directed graphs soon become much more numerous than unregulated non-isomorphic directed graphs as the number of interactors increase (125 vs. 5604 for 4 interactors, Table 1). With just 10 interactors, there are >3x1020 unregulated non-isomorphic directed graphs [60] and HRAs are expected to be more numerous. This tremendous variety highlights the vast amount of information that a complex regulatory architecture can represent and the large number of changes that are possible despite sparsity of the change matrix (Fig. 3).”

Thus, indeed as the reviewer surmises, the combinatorial explosion in information of HRAs with increases in interacting entities is faster than the proportional growth in information of genome sequence with increases in length.

In summary, I thank the reviewers and editors for their help in improving the paper and would like to make the current manuscript the Version of Record.

---

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

Public Reviews:

Reviewer #1 (Public Review):

The author studies a family of models for heritable epigenetic information, with a focus on enumerating and classifying different possible architectures. The key aspects of the paper are:

• Enumerate all 'heritable' architectures for up to 4 constituents.

• A study of whether permanent ("genetic") or transient ("epigenetic") perturbations lead to heritable changes.

• Enumerated the connectivity of the "sequence space" formed by these heritable architectures.

-Incorporating stochasticity, the authors explore stability to noise (transient perturbations). - A connection is made with experimental results on C elegans.

The study is timely, as there has been a renewed interest in the last decade in nongenetic, heritable heterogeneity (e.g., from single-cell transcriptomics). Consequently, there is a need for a theoretical understanding of the constraints on such systems. There are some excellent aspects of this study: for instance:

• The attention paid to how one architecture "mutates" into another, establishing the analogue of a "sequence space" for network motifs (Fig 3).

• The distinction is drawn between permanent ("genetic") and transient ("epigenetic") perturbations that can lead to heritable changes.

• The interplay between development, generational timescales, and physiological time (as in Fig. 5).

I thank the reviewer for highlighting these aspects of the work.

The manuscript would be very interesting if it focused on explaining and expanding these results. Unfortunately, as a whole, it does not succeed in formalising nor addressing any particular open questions in the field. Aside from issues in presentation and modelling choices (detailed below), it would benefit greatly from a more systematic approach rather than the vignettes presented.

This first paper is foundational and therefore cannot be expected to solve all aspects of the problem of heredity. The work was nevertheless systematic in that (1) for the simple architectures modeled using ordinary differential equations (ODEs) with continuity assumptions, parameters that support steady states were systematically determined for each architecture and then every architecture was explored using genetic changes exhaustively, although epigenetic perturbations were not examined exhaustively because of their wide variety; and (2) for the more realistic modeling of architectures as Entity-Sensor-Property systems, the behavior of systems with respect to architecture as well as parameter space that lead to particular behaviors (persistence, heritable epigenetic change, etc.) was systematically explored. A more extensive exploration of parameter space that also includes the many ways that the interaction between any two entities/nodes could be specified using an equation is a potentially ever-expanding challenge that is beyond the scope of any single paper (see response to additional comments below).

Specific aspects that remain to be addressed include the application of multiple notions of heritability to real networks of arbitrary size, considering different types of equations for change of each entity/node, and classifying different behavioral regimes for different sets of parameters. As is evident from this list of combinatorial possibilities, the space to be explored is vast and beyond the scope of this foundational paper.

The key contribution of the paper is an articulation of the crucial questions to ask of any regulatory architecture in living systems rather than the addressing of any question that a field has recognized as ‘open’. Specifically, through the exhaustive listing for small regulatory architectures that can be heritable and the systematic analysis of arbitrary Entity-Sensor-Property systems that more realistically capture regulatory architectures in living systems, this work points the way to constrain inferences after experiments on real living systems. Currently, most experimental biologists engaged in reductionist approaches and some systems biologists examining the function or prevalence of network motifs do not explicitly constrain their models for heritability or persistence. It is hoped that this paper will raise awareness in both communities and lead to more constrained models that minimize biases introduced by incomplete knowledge of the network, which is always the case when analyzing living systems.

Terminology

The author introduces a terminology for networks of interacting species in terms of "entities" and "sensors" -- the former being nodes of a graph, and the latter being those nodes that receive inputs from other nodes. In the language of directed graphs, "entities" would seem to correspond to vertices, and "sensors" those vertices with positive indegree and outdegree. Unfortunately, the added benefit of redefining accepted terminology from the study of graphs and networks is not clear.

The Entities-Sensors-Property (ESP) framework is based on underlying biology and not graph theory, making an ESP system not entirely equivalent to a network or graph, which is much less constrained. The terms ‘entity’, ‘sensor’, and ‘property’ were defined and justified in a previous paper (Jose, J R. Soc. Interface, 2020). While nodes of a network can be parsed arbitrarily and the relationship between them can also be arbitrary, entities and sensors are molecules or collections of molecules that are constrained such that the sensors respond to changes in particular properties of other entities and/or sensors. When considered as digraphs, sensors can be seen as vertices with positive indegree and outdegree. The ESP framework can be applied across any scale of organization in living systems and this specific way of parsing interactions also discretizes all changes in the values of any property of any entity. In short, ESP systems are networks, but not all networks are ESP systems. Therefore, the results of network theory that remain applicable for ESP systems need further investigation. This justification is now repeated in the paper.

The key utility of the ESP framework is that it is aligned with the development of mechanistic models for the functions of living systems while being consistent with heredity. In contrast, widely analyzed networks like protein-interaction networks, signaling networks, gene regulatory networks, etc., are not always constrained using these principles. In addition, the language of digraphs where sensors can be seen as vertices with positive indegree and outdegree has been also added to aid readers who are familiar with graph theory.

Heritability

The primary goal of the paper is to analyse the properties of those networks that constitute "heritable regulatory architectures". The definition of heritability is not clearly stated anywhere in the paper, but it appears to be that the steady-state of the network must have a non-zero expression of every entity. As this is the heart of the paper, it would be good to have the definition of heritable laid out clearly in either the main text or the SI.

I have now defined the term as used in this paper early, which is indeed as surmised by the reviewer simply the preservation of the architecture and non-zero levels of all entities. I have also highlighted additional notions of heredity that are possible, which will be the focus of future work. These can range from precise reproduction of the concentration and the localization of every entity to a subset of the entities being reproduced with some error while the rest keep varying from generation to generation (as illustrated in Fig. 2 of Jose, BioEssays, 2018). Importantly, it is currently unclear which of these possibilities reflects heredity in real living systems.

Model

As described in the supplementary, but not in the main text, the author first chooses to endow these networks with simple linear dynamics; something like $\partial_t \vec{x} = A x - T x$, where the vector $x$ is the expression level of each entity, $A$ has the structure of the adjacency matrix of the directed graph, and $T$ is a diagonal matrix with positive entries that determines the degradation or dilution rate of each entity. From a readability standpoint, it would greatly aid the reader if the long list of equations in the SI were replaced with the simple rule that takes one from a network diagram to a set of ODEs.

I have abridged the description by eliminating the steady state expression for every HRA as suggested and simply pointed to the earlier version of the paper for those readers who might prefer the explicit derivations of these simple expressions. An overview is now provided for going from any network diagram to a set of ODEs.

The implementation of negative regulation is manifestly unphysical if the "entities" represent the expression level of, say, gene products. For instance, in regulatory network E, the value of the variable z can go negative (for instance, if the system starts with z= and y=0, and x > 0).

Negative values for any entity were avoided in simulations by explicitly setting all such values to zero. This constraint has been added as a note in the section describing the equations for the change of each node/entity in each regulatory network. Specifically, the levels of each entity/sensor was set to zero during any time step when the computed value for that entity/sensor was less than zero. This bounding of the function allows for any approach to zero while avoiding negative values. I apologize for the omission of this constraint from the supplemental material in the last submission. This constraint was used in all the simulations and therefore this change does not affect any of the results presented. In this way, it is ensured that the presence of negative regulation does not lead to negative values.

Formally, the promotion or inhibition of an entity or sensor can be modeled using any function that is either increasing (for promotion) or decreasing (for inhibition). This diversity of possibilities is one of the challenges that prevents exhaustive exploration of all functions. In fact, the use of ODEs after assuming a continuous function is an idealization that facilitates understanding of general principles but is not in keeping with the discreteness of entities or step changes in their values (amount, localization, etc.) observed in living systems. Other commonly used continuous functions include Hill functions for the rate of production of y given as xn/(k + xn) for x activating y, which increases to ~1 as x increases, or given as k/(k + xn) for x inhibiting y, which decreases to ~0 as x increases. Increasing values of ‘n’ result in steeper sigmoidal curves. In reality, levels of all entities/sensors are expected to be discretized by measurement in living systems and the form of the function for any regulation needs empirical measurement in vivo (see response to comment below).

The model seems to suddenly change from Figure 4 onwards. While the results presented here have at least some attempt at classification or statistical rigour (i.e. Fig 4 D), there are suddenly three values associated with each entity ("property step, active fraction, and number"). Furthermore, the system suddenly appears to be stochastic. The reader is left unsure of what has happened, especially after having made the effort to deduce the model as it was in Figs 1 through 3. No respite is to be found in the SI, either, where this new stochastic model should have been described in sufficient detail to allow one to reproduce the simulation.

While ODEs are easier to simulate and understand, they are less realistic as explained above. I have now added more explanation justifying the need for the subsequent simulation of Entity-Sensor-Property systems. I have also expanded the information provided for each aspect of the model (previously outlined in Fig. 4A and detailed within the code) in a Supplementary Information section titled ‘Simulation of simple ESP systems’.

Perturbations

Inspired especially by experimental manipulations such as RNAi or mutagenesis, the author studies whether such perturbations can lead to a heritable change in network output. While this is naturally the case for permanent changes (such as mutagenesis), the author gives convincing examples of cases in which transient perturbations lead to heritable changes. Presumably, this is due the the underlying mutlistability of many networks, in which a perturbation can pop the system from one attractor to another.

Unfortunately, there appears to be no attempt at a systematic study of outcomes, nor a classification of when a particular behaviour is to be expected. Instead, there is a long and difficult-to-read description of numerical results that appear to have been sampled at random (in terms of both the architecture and parameter regime chosen). The main result here appears to be that "genetic" (permanent) and "epigenetic" (transient) perturbations can differ from each other -- and that architectures that share a response to genetic perturbation need not behave the same under an epigenetic one. This is neither surprising (in which case even illustrative evidence would have sufficed) nor is it explored with statistical or combinatorial rigour (e.g. how easy is it to mistake one architecture for another? What fraction share a response to a particular perturbation?)

The systematic study of all arbitrary regulatory architectures is beyond the scope of this paper and, as stated earlier, beyond the scope of any one paper. Nevertheless 225,000 arbitrary Entity-Sensor-Property systems were systematically explored and collections of parameters that lead to particular behaviors provided (e.g., 78,285 are heritable). These ESP systems more closely mimic regulation in living systems than the coupled ODE-based specification of change in a regulatory architecture.

The example questions raised here are not only difficult to answer, but subjective and present a moving target for future studies. One, ‘how easy is it to mistake one architecture for another?’. Mistaking one architecture for another clearly depends on the number of different types of experiments one can perform on an architecture and the resolution with which changes in entities can be measured to find distinguishing features. Two, ‘What fraction share a response to a particular perturbation?’. ‘Sharing a response’ also depends on the resolution of the measurement of entities after perturbation.

As an additional comment, many of the results here are presented as depending on the topology of the network. However, each network is specified by many kinetic constants, and there is no attempt to consider the robustness of results to changes in parameters.

The interpretations presented are conservative determinations of heritability based on the topology of the architecture. In other words, architectures that can be heritable for some set of parameters. Of course, parameter sets can be found that make any regulatory architecture not heritable. As stated earlier, exploring all parameters for even one architecture is beyond the scope of a single study because of the infinitely many ways that the interaction between any two entities can be specified.

DNA analogy

At two points, the author makes a comparison between genetic information (i.e. DNA) and epigenetic information as determined by these heritable regulatory architectures. The two claims the author makes are that (i) heritable architectures are capable of transmitting "more heritable information" than genetic sequences, and (ii) that, unlike DNA, the connectivity (in the sense of mutations) between heritable architectures is sparse and uneven (i.e. some architectures are better connected than others).

In both cases, the claim is somewhat tenuous -- in essence, it seems an unfair comparison to consider the basic epigenetic unit to be an "entity" (e.g., an entire transcription factor gene product, or an organelle), while the basic genetic unit is taken to be a single base-pair. The situation is somewhat different if the relevant comparison was the typical size of a gene (e.g., 1 kb).

Considering every base being the unit of stored information in the DNA sequence results in the maximal possible storage capacity of a genome of given length. Any other equivalence between entity and units within the genome (e.g., 1 kb gene) will only reduce the information stored in the genome.

Nevertheless, the claim has been modified to say that the information content of an ESP system can [italics added] be more extensive than the information content of the genome. This accounts for the possibility of an organism that has an inordinately large genome such that maximal information that can be stored in a particular genome sequence exceeds that stored in a particular configuration of all the contents in a cell.

Reviewer #2 (Public Review):

Summary:

This manuscript uses an interesting abstraction of epigenetic inheritance systems as partially stable states in biological networks. This follows on previous review/commentary articles by the author. Most of the molecular epigenetic inheritance literature in multicellular organisms implies some kind of templating or copying mechanisms (DNA or histone methylation, small RNA amplification) and does not focus on stability from a systems biology perspective. By contrast, theoretical and experimental work on the stability of biological networks has focused on unicellular systems (bacteria), and neglects development. The larger part of the present manuscript (Figures 1-4) deals with such networks that could exist in bacteria. The author classifies and simulates networks of interacting entities, and (unsurprisingly) concludes that positive feedback is important for stability. This part is an interesting exercise but would need to be assessed by another reviewer for comprehensiveness and for originality in the systems biology literature. There is much literature on "epigenetic" memory in networks, with several stable states and I do not see here anything strikingly new.

The key utility of the initial part of the paper is the exhaustive enumeration of all small heritable regulatory architectures. The implications for the abundance of ‘network motifs’ and more generally any part of a network proposed to perform a particular function is that all such parts need to be compatible with heredity. This principle is generally not followed in the literature, resulting in incomplete networks being interpreted as having motifs or modules with autonomous function. Therefore, while the need for positive feedback for stability is indeed obvious, it is not consistently applied by all. For example, the famous synthetic circuit ‘the repressilator’ (Elowitz and Leibler, “A synthetic oscillatory network of transcriptional regulators”, Nature, 2000), which is presented as an example of ‘rational network design’, has three transcription factors that all sequentially inhibit the production of another transcription factor in turn forming a feedback loop of inhibitory interactions. Therefore, the contributions of the factors that promote the expression of each entity is unknown and yet essential for heritability. The comprehensive listing of the heritable regulatory architectures that are simple provide the basis for true synthetic biology where the contributing factors for observed behavior of the network are explicitly considered only after constraining for heredity. Using this principle, the minimal autonomous architecture that can implement the repressilator is the HRA ‘Z’ (Fig. 1).

An interesting part is then to discuss such networks in the framework of a multicellular organism rather than dividing unicellular organisms, and Figure 5 includes development in the picture. Finally, Figure 6 makes a model of the feedback loops in small RNA inheritance in C. elegans to explain differences in the length of inheritance of silencing in different contexts and for different genes and their sensitivity to perturbations. The proposed model for the memory length is distinct from a previously published model by Karin et al. (ref 49).

I thank the reviewer for appreciating this aspect of the paper.

Strengths:

A key strength of the manuscript is to reflect on conditions for epigenetic inheritance and its variable duration from the perspective of network stability.

I thank the reviewer for appreciating the importance of the overall topic.

Weaknesses:

• I found confusing the distinction between the architecture of the network and the state in which it is. Many network components (proteins and RNAs) are coded in the genome, so a node may not disappear forever.

I have added language to clarify the many states of a network versus its architecture (also illustrated in Fig. 4 for ESP systems). Even loss of expression below a threshold can lead to permanent loss if there is not sufficient noise to induce re-expression. For example, consider the simple case of a transcription factor that binds to its own promoter, requiring 10 molecules for the activation of the promoter and thus production of more of the same transcription factor. If an epigenetic change (e.g., RNA interference) reduces the levels to fewer than 10 molecules and if the noise in the system never results in the numbers of the transcription factor increasing beyond 10, the transcription factor has been effectively lost permanently. In this way, reduction of a regulator can lead to permanent change despite the presence of the DNA. Many papers in the field of RNA silencing in C. elegans have provided strong experimental evidence to support this assertion.

• From the Supplementary methods, the relationship between two nodes seems to be all in the form of dx/dt = Kxy . Y, which is just one way to model biological reactions. The generality of the results on network architectures that are heritable and robust/sensitive to change is unclear. Other interactions can have sigmoidal effects, for example. Is there no systems biology study that has addressed (meta)stability of networks before in a more general manner?

Indeed, the relationship between any two entities can in principle be modeled using any function. Extensive exploration of the behavior of any regulatory architecture – even the simplest ones – require simplifications. For example, early work by Stuart Kauffman explored Boolean networks (see ref. 10 in the paper for history and extensive explanations). However, allowing all possible ways of specifying the interactions between components of a network makes analysis both a computational and conceptual challenge.

• Why is auto-regulation neglected? As this is a clear cause of metastable states that can be inherited, I was surprised not to find this among the networks.

Auto-regulation in the sense of some molecule/entity ultimately leading to the production of more of itself is present in every heritable regulatory architecture. Specifically, all auto-regulatory loops rely on a sequence of interactions between two or more kinds of molecules. For example, a transcription factor (TF) binding to the promoter of its own gene sequence, resulting in the production of more TF protein is a positive feedback loop that relies on many interacting factors (transcription, translation, nuclear import, etc.) and can be considered as ‘auto-regulation’ as it is sometimes referred to in the literature. In this sense, every HRA (A through Z) includes ‘auto-regulation’ or more appropriately positive feedback loops. For example, in the HRA ‘A’, x ‘auto-regulates’ itself via y.

• I did not understand the point of using the term "entity-sensor-property". Are they the same networks as above, now simulated in a computer environment step by step (thus allowing delays)?

Please see response to the other reviewer regarding the need for the Entity-SensorProperty framework and how it is distinct from generic networks. Briefly, the ODE-based simple networks, while easy to analyze, are not realistic because of the assumptions of continuity. In contrast ESP systems are more realistic with measurement discretizing changes in property values as is expected in real living systems.

• The final part applies the network modeling framework from above to small RNA inheritance in C. elegans. Given the positive feedback, what requires explanation is how fast the system STOPs small RNA inheritance. A previous model (Karin et al., ref. 49) builds on the fact that factors involved in inheritance are in finite quantity hence the different small RNAs "compete" for amplification and those targeting a given gene may eventually become extinct.

The present model relies on a simple positive feedback that in principle can be modulated, and this modulation remains outside the model. A possibility is to add negative regulation by factors such as HERI-1, that are known to limit the duration of the silencing.

The duration of silencing differs between genes. To explain this, the author introduces again outside the model the possibility of piRNAs acting on the mRNA, which may provide a difference in the stability of the system for different transcripts. At the end, I do not understand the point of modeling the positive feedback.

The previous model (Karin et al., Cell Systems, 2023) can describe populations of genes that are undergoing RNA silencing but cannot explain the dynamics of silencing particular genes. Furthermore, this model also cannot explain cases of effectively permanent silencing of genes that have been reported (e.g., Devanapally et al., Nature Communications, 2021 and Shukla et al., Current Biology, 2021). Finally, the observations of susceptibility to, recovery from, and even resistance to trans silencing (e.g., Fig. 5a in Devanapally et al., Nature Communications, 2021) require an explanation that includes modulation of the HRDE-1-dependent positive feedback loop that maintains silencing across generations.

The specific qualitative predictions regarding the relationship between piRNA-mediated regulation genome-wide and HRDE-1-dependent silencing of a particular gene across generations could guide the discovery of potential regulators of heritable RNA silencing. The equations (4) and (5) in the paper for the extent of modulation needed for heritable epigenetic change provide specific quantitative predictions that can be tested experimentally in the future. I have also revised the title of the section to read ‘Tuning of positive feedback loops acting across generations can explain the dynamics of heritable RNA silencing in C. elegans’ to emphasize the above points.

• From the initial analysis of abstract networks that do not rely on templating, I expected a discussion of possible examples from non-templated systems and was a little surprised by the end of the manuscript on small RNAs.

The heritability of any entity relies on regulatory interactions regardless of whether a templated mechanism is also used or not. For example, DNA replication relies on the interactions between numerous regulators, with only the sequence being determined by the template DNA. The field of small RNA-mediated silencing facilitates analysis of epigenetic changes at single-gene resolution (Chey and Jose, Trends in Genetics, 2022). It is therefore likely to continue to provide insights into heritable epigenetic changes and how they can be modulated. Unfortunately, there are currently no known cases of epigenetic inheritance where the role of any templated mechanism has been conclusively excluded. Future research will improve our understanding of epigenetic states and their modulation in terms of changes in positive feedback loops as proposed in this study and potentially lead to the discovery of such mechanisms that act entirely independent of any template-dependent entity.

Recommendations for the authors:

I thank the reviewers for their specific suggestions to improve the paper.

Reviewer #1 (Recommendations For The Authors):

The paper has many long paragraphs that attempt to explain results, make illustrations, and give intuition. Unfortunately, these are difficult to read. It would aid the reader greatly if these were, say, converted into cartoons (even if only in the SI), or made more accessible in some other way.

I agree with the importance of making the material accessible to readers in multiple ways. I have now added a figure with schematics in the SI titled ‘Illustrations of key concepts’ (new Fig. S2), which collects concepts that are relevant throughout the paper and might aid some readers.

The bulk of the supplementary is currently a collection of elementary mathematics results: to whit, pages 26 to 33 of the combined manuscript carry no more information than a quick description of the general model and the diagrams in Fig 1. Similarly, pages 34 to 39 (non-zero dilution rate), and pages 39 through 58 (response to permanent changes) each express a trivial mathematical point that is more than sufficiently made with one illustrative example.

I agree with the reviewer and have condensed these pages as suggested. I have added a pointer to the earlier version as containing further details for the readers who might prefer the explicit listing of these equations.

Overall, the paper appears to be a collection of numerical results obtained from different models, united by uncertain terminology that is not fully defined in this paper. The most promising aspects of the paper lie either in (a) combinatorially complete enumeration of all regulatory architectures, or (b) relating experimental manipulations in C. elegans to possible underlying regulatory architectures. Focusing on one or the other might improve the readability of the paper.

The two sections of the paper are complementary and when presented together help with the integration of concepts rather than the siloed pursuit of theory versus experimental analysis. When this work was presented at meetings before submission, it was clear that different researchers appreciated different aspects. This divergence is also apparent in the two reviews, with each reviewer appreciating different aspects. I have repeated the definitions and justifications from the earlier paper (Jose, J R Soc Interface, 2020) to provide a more fluid transition between the two complementary sections of the paper. Knowing both sides could aid in the development of models that are not only consistent with measurable quantities (e.g., anything that can be considered an entity) but are also logically constrained (e.g., entities matched with sensors while avoiding any entities that do not have a source of production – i.e., avoiding nodes with indegree = 0).

However, having said that many results of these types are well-known in models of regulatory networks, and it is unclear what precisely warrants the new framework that the author is proposing. Indeed, it would be good to understand in what way the framework here is novel, and how it is distinguished from prior studies of regulatory networks.

The key novelty of the work is the consideration of heritability for any regulation. With the explicit definition of the heritability for a regulatory architecture and the acknowledgement that there can be more than one notion of heredity, this paper now sets the foundation for examining many real networks in this light. I hope that the added justifications for the current framework in the revised paper strengthen these arguments. Future literature reviews on networks in general and how they address heritability or persistence will better define the prevalence of these considerations. Currently, most experimental biologists engaged in reductionist approaches and some systems biologists examining the function or prevalence of network motifs do not explicitly constrain their models for heritability or persistence. It is hoped that this work will raise awareness in both communities and lead to more constrained models that acknowledge incomplete knowledge of the network, which is always the case when analyzing living systems.

Reviewer #2 (Recommendations For The Authors):

Minor points/clarity

• page 1 line 57: "transgenerational waveforms that preserve form and function" is unclear.

This phrase was expanded upon in a previous paper (Jose, BioEssays, 2020). I have now added more explanation in this paper for completeness. The section now reads ‘For example, the localization and activity of many kinds of molecules are recreated in successive generations during comparable stages [1-3]. These recurring patterns can change throughout development such that following the levels and/or localizations of each kind of molecule over time traces waveforms that return in phase with the similarity of form and function across generations [2].’

• page 7 line 3-6: the sentence has an ambiguous structure.

I have now edited this long sentence to read as follows: ‘For systematic analysis, architectures that could persist for ~50 generations without even a transient loss of any entity/sensor were considered HRAs. Each HRA was perturbed (loss-of-function or gain-of-function) after five different time intervals since the start of the simulation (i.e., phases). The response of each HRA to such perturbations were compared with that of the unperturbed HRA.’

• page 9 lines 25-27: the sentence is convoluted: are you defining epigenetic inheritance?

I have simplified this sentence describing prior work by others (Karin et al., Cell Systems, 2023) and moved a clause to the subsequent sentence. This section now reads: ‘Recent considerations of competition for regulatory resources in populations of genes that are being silenced suggest explanations for some observations on RNA silencing in C. elegans [49]. Specifically, based on Little’s law of queueing, with a pool of M genes silenced for an average duration of T, new silenced genes arise at a rate  that is given by M = T’. I have also provided more context by preceding this section with: ‘Although the release of shared regulators upon loss of piRNA-mediated regulation in animals lacking PRG-1 could be adequate to explain enhanced HRDE-1-dependent transgenerational silencing initiated by dsRNA in prg-1(-) animals, such a competition model alone cannot explain the observed alternatives of susceptibility, recovery and resistance (Fig. 6A).’

• page 13 lines 51-53. This last sentence of the discussion is ambiguous/unclear.

I have now rephrased this sentence to read: ‘This pathway for increasing complexity through interactions since before the origin of life suggests that when making synthetic life, any form of high-density information storage that interacts with heritable regulatory architectures can act as the ‘genome’ analogous to DNA.’

• Figure 2: the letters in the nodes are hard to read; the difference between full and dotted lines in the graphs also.

I have enlarged the nodes and widened the gap in the dotted lines to make them clearer. I have also similarly edited Fig. 1 and Fig. S3 to Fig. S9.

eLife assessment

This useful manuscript explores conditions for epigenetic inheritance by studying the stability of simple network models to permanent and transient perturbations. A novel aspect of the study is that it unifies non-genetic inheritance phenomena across cell divisions of unicellular organisms and in the germline of multicellular organisms. However, the models studied are more a collection of vignettes of numerical studies than a systematic study, therefore the evidence presented remains incomplete. As a first step towards building a more systematic theoretical framework, this work will be of interest to colleagues in the field of epigenetic inheritance.

Reviewer #1 (Public Review):

The author studies a family of models for heritable epigenetic information, with a focus on enumerating and classifying different possible architectures. The key aspects of the paper are:

- Enumerate all 'heritable' architectures for up-to 4 constituents.<br /> - A study of whether permanent ("genetic") or transient ("epigenetic") perturbations lead to heritable changes<br /> - Enumerated the connectivity of the "sequence space" formed by these heritable architectures<br /> - Incorporating stochasticity, the authors explore stability to noise (transient perturbations)<br /> - A connection is made with experimental results on C elegans.

The study is timely, as there is a renewed interest in the last decade in non-genetic, heritable heterogeneity (e.g., from single-cell transcriptomics). Consequently, there is a need for a theoretical understanding of the constraints on such systems. There are some excellent aspects of this study: for instance, the attention paid to how one architecture "mutates" into another. Unfortunately, the manuscript as a whole does not succeed in formalising nor addressing any particular open questions in the field. Aside from issues in presentation and modelling choices (detailed below), it would benefit greatly from a more systematic approach rather than the vignettes presented.

## Terminology

The author introduces a terminology for networks of interacting species in terms of "entities" and "sensors" -- the former being nodes of a graph, and the latter being those nodes that receive inputs from other nodes. In the language of directed graphs, "entities" would seem to correspond to vertices, and "sensors" those vertices with positive indegree and outdegree. Unfortunately, the added benefit of redefining accepted terminology from the study of graphs and networks is not clear.

## Model

The model seems to suddenly change from Figure 4 onwards. While the results presented here have at least some attempt at classification or statistical rigour (i.e. Fig 4 D), there are suddenly three values associated with each entity ("property step, active fraction, and number"). Furthermore, the system suddenly appears to be stochastic. The reader is left unsure what has happened, especially after having made the effort to deduce the model as it was in Figs 1 through 3. No respite is to be found in the SI, either, where this new stochastic model should have been described in sufficient detail to allow one to reproduce the simulation.

## Perturbations

Inspired especially by experimental manipulations such as RNAi or mutagenesis, the author studies whether such perturbations can lead to a heritable change in network output. While this is naturally the case for permanent changes (such as mutagenesis), the author gives convincing examples of cases in which transient perturbations lead to heritable changes. Presumably, this is due the the underlying multistability of many networks, in which a perturbation can pop the system from one attractor to another.

Unfortunately, there appears to be no attempt at a systematic study of outcomes, nor a classification of when a particular behaviour is to be expected. Instead, there is a long and difficult-to-read description of numerical results that appear to have been sampled at random (in terms of both the architecture and parameter regime chosen). The main result here appears to be that "genetic" (permanent) and "epigenetic" (transient) perturbations can differ from each other -- and that architectures that share a response to genetic perturbation need not behave the same under an epigenetic one. This is neither surprising (in which case even illustrative evidence would have sufficed) nor is it explored with statistical or combinatorial rigour (e.g. how easy is it to mistake one architecture for another? What fraction share a response to a particular perturbation?)

As an additional comment, many of the results here are presented as depending on the topology of the network. However, each network is specified by many kinetic constants, and there is no attempt to consider the robustness of results to changes in parameters.

## DNA analogy

At two points, the author makes a comparison between genetic information (i.e. DNA) and epigenetic information as determined by these heritable regulatory architectures. The two claims the author makes are that (i) heritable architectures are capable of transmitting "more heritable information" than genetic sequences, and (ii) that, unlike DNA, the connectivity (in the sense of mutations) between heritable architectures is sparse and uneven (i.e. some architectures are better connected than others).

In both cases, the claim is somewhat tenuous -- in essence, it seems an unfair comparison to consider the basic epigenetic unit to be an "entity" (e.g., an entire transcription factor gene product, or an organelle), while the basic genetic unit is taken to be a single base-pair. The situation is somewhat different if the relevant comparison was the typical size of a gene (e.g., 1 kb).

Author Response

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

eLife assessment

This important study convincingly shows that the less common D-serine stereoisomer is transported in the kidney by the neutral amino acid transporter ASCT2 and that it is a noncanonical substrate for sodium-coupled monocarboxylate transporter SMCTs. With a multihierarchical approach, this important study further shows that Ischemia-Reperfusion Injury in the kidney causes a specific increment in renal reabsorption carried out, in part, by ASCT2.

Public Reviews:

Reviewer #1 (Public Review):

Most amino acids are stereoisomers in the L-enantiomer, but natural D-serine has also been detected in mammals and its levels shown to be connected to a number of different pathologies. Here, the authors convincingly show that D-serine is transported in the kidney by the neutral amino acid transporter ASCT2 and as a non-canonical substrate for the sodium-coupled monocarboxylate transporter SMCTs. Although both transport D-serine, this important study further shows in a mouse model for acute kidney injury that ASCT2 has the dominant role.

Strengths:

The paper combines proteomics, animal models, ex vivo transport analyses, and in vitro transport assays using purified components. The exhaustive methods employed provide compelling evidence that both transporters can translocate D-serine in the kidney.

Weakness:

In the model for acute kidney injury, the SMCTs proteins were not showing a significant change in expression levels and were rather analysed based on other, circumstantial evidence. Although its clear SMCTs can transport D-serine its physiological role is less obvious compared to ASCT2.

We greatly value the reviewer's efforts and feedback in reviewing our manuscript. We acknowledge the reviewer's observation that the changes indicated by our proteomic results are not markedly pronounced. To reinforce our findings, we have incorporated an analysis of gene alterations at the single-cell level (snRNA-seq) from the publicly accessible IRI mouse model data (Figure supplement 7). The snRNA-seq data align with our proteomic data in terms of the general trend of gene/protein alterations, but reveal more substantial changes in both ASCT2 and SMCTs. These discrepancies might stem from the different quantification methods used, suggesting a possible underestimation in our label-free proteomic quantification. The differences we see between the functional changes in transporters and their quantification in proteomics can be explained by the unique challenges posed by membrane proteins. Post-translational modifications and the complex nature of multiple transmembrane domains often impact the accurate measurement of these proteins in proteomic studies. This complexity can lead to a mismatch between the actual functional changes occurring in the transporters and their perceived abundance or alterations as detected by proteomic methods (Figure 4A) (Schey KL et al. Biochemistry 2015, doi: 10.1021/bi301604j). However, this label-free quantitative proteomics approach is well-suited for our study, given its screening efficiency, compatibility with animal models, and the absence of a labeling requirement. We may consider incorporating alternative quantitative proteomic methods in future for a more thorough comparison. We have included these considerations in lines 351-356 of the revised manuscript.

Manuscript lines 351-356

“When evaluating the extent of gene/protein alterations between the control and IRI conditions, we observed that the gene alterations of both Asct2 and Smcts, as revealed by snRNAsequencing, are more pronounced than the protein alteration ratios obtained from proteomics. This discrepancy may stem from difficulty in the quantification method, especially for membrane transport proteins in label-free quantitative proteomics.”

Regarding the roles of ASCT2 and SMCTs in renal D-serine transport, snRNA-seq showed that ASCT2 expression in the controls is less than 10% of the cell population. We suggest that ASCT2 contributes to D-serine reabsorption because of its high affinity and SMCTs (SMCT1 and SMCT2) would play a role in D-serine reabsorption in the cells without ASCT2 expression. In addition, we included other factors (the turnover rate and the presence of local canonical substrates) that may determine the capability of D-serine reabsorption. We have included this suggestion in the Discussion lines 386-404.

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of Dserine transport systems.”

Reviewer #2 (Public Review):

Summary:

The manuscript "A multi-hierarchical approach reveals D-1 serine as a hidden substrate of sodium-coupled monocarboxylate transporters" by Wiriyasermkul et al. is a resubmission of a manuscript, which focused first on the proteomic analysis of apical membrane isolated from mouse kidney with early Ischemia-Reperfusion Injury (IRI), a well-known acute kidney injury (AKI) model. In the second part, the transport of D-serine by Asct2, Smct1, and Smct2 has been characterized in detail in different model systems, such as transfected cells and proteoliposomes.

Strengths:

A major problem with the first submission was the explanation of the link between the two parts of the manuscript: it was not very clear why the focus on Asct2, Smct1, and Smct2 was a consequence of the proteomic analysis. In the present version of the manuscript, the authors have focused on the expression of membrane transporters in the proteome analysis, thus making the reason for studying Asct2, Smct1, and Smct2 transporters more clear. In addition, the authors used 2D-HPLC to measure plasma and urinary enantiomers of 20 amino acids in plasma and urine samples from sham and Ischemia-Reperfusion Injury (IRI) mice. The results of this analysis demonstrated the value of D-serine as a potential marker of renal injury. These changes have greatly improved the manuscript and made it more convincing.

We deeply appreciate the reviewer’s comments on the manuscript. We have responded to the recommendations one by one in the later section.

Reviewer #3 (Public Review):

Summary:

The main objective of this work has been to delve into the mechanisms underlying the increment of D-serine in serum, as a marker of renal injury.

Strengths:

With a multi-hierarchical approach, the work shows that Ischemia-Reperfusion Injury in the kidney causes a specific increment in renal reabsorption of D-serine that, at least in part, is due to the increased expression of the apical transporter ASCT2. In this way, the authors revealed that SMCT1 also transports D-serine.

The experimental approach and the identification of D-serine as a new substrate for SMCT1 merit publication in Elife.

The manuscript also supports that increased expression of ASCT2, even together with the parallel decreased expression of SMCT1, in renal proximal tubules underlies the increased reabsorption of D-serine responsible for the increment of this enantiomer in serum in a murine model of Ischemia-Reperfusion Injury.

Weaknesses:

Remains to be clarified whether ASCT2 has substantial stereospecificity in favor of D- versus L-serine to sustain a ~10-fold decrease in the ratio D-serine/L-serine in the urine of mice under Ischemia-Reperfusion Injury (IRI).

It is not clear how the increment in the expression of ASCT2, in parallel with the decreased expression of SMCT1, results in increased renal reabsorption of D-serine in IRI.

We thoughtfully appreciate the reviewer’s comment on the manuscript. Considering the alteration of D-/L-serine ratios, there are several factors including protein expression levels at both apical and basolateral sides, properties of the transporters (e.g. transport affinities, substrate stereoselectivities), and the expression of DAAO (D-amino acid oxidase) which selectively degrades D-amino acids. Moreover, the mechanism becomes more complicated when the transport systems of L- and D-enantiomers are different and have distinct stereoselectivities as in the case of serine. Future studies are required to complete the mechanism. However, we would like to explore the mechanism based on the current knowledge.

From this study, we identified ASCT2 and SMCTs (SMCT1 and SMCT2) as D-serine transport systems. We showed that SMCT1 prefers D-serine. Although we did not analyze ASCT2 stereoselectivity, based on the previous studies, ASCT2 recognizes both D- and Lserine with high affinities and slightly prefers L-enantiomer (Km of 18.4 µM for L-serine in oocyte expression system (Utsunomiya-Tate et al. J Biol Chem 1996) and 167 µM for Dserine in oocyte expression system (Foster et al. Plos ONE 2016), and the IC50 of 0.7 mM for L-serine and 4.9 mM for D-serine (in HEK293 expression systems, Foster et al. PLOS ONE 2016). The proteomics showed an increase of ASCT2 (1.6-fold increase) and a decrease of SMCTs (1.7-fold decrease in SMCT1, and 1.3-fold decrease in SMCT2) in IRI conditions. The table below summarizes D-serine transport by ASCT2 and SMCTs.

In the case of L-serine, ASCT2 and B0ATs (in particular B0AT3) have been revealed as L-serine transport systems in the kidneys (Bröer et al. Physiol Rev 2008; Singer et al. J Biol Chem 2009). Proteomics showed that B0ATs have higher expression levels than ASCT2 supporting the idea that B0ATs are the main L-serine transport system (Table S1: Abundance of B0AT1 = 1.34E+09, B0AT3 = 2.13E+08, ASCT2 = 1.46E+07). In IRI conditions, B0AT3 decreased 1.8 fold and B0AT1 decreased 1.1 fold. From these results, we included the contribution of B0ATs in L-serine transport in Author response table 1.

Author response table 1.

Taken together, we suggest that high ratios of D-/L-serine in IRI conditions are a combinational result of 1) increase of D-serine reabsorption by ASCT2 enhancement and SMCTs reduction and 2) decrease of L-serine reabsorption by B0ATs. We have included this suggestion in the Discussion lines 438-451.

Manuscript lines 438-451

“The enantiomeric profiles of serine revealed distinct plasma D/L-serine ratio, with low rations in the normal control but elevated ratios in IRI, despite the weak stereoselectivity of ASCT2 (Figure 1B). This observation suggested differential renal handling of D-serine compared to L-serine. While we identified SMCTs as a D-serine transport system, it has been reported that L-serine reabsorption is mediated by B0AT3 (Singer et al., 2009). We propose that the alterations in plasma and urinary D/L-serine ratios are the combined outcomes of: 1) transport systems for L-serine, and 2) transport systems for D-serine. In normal kidneys, the low plasma D/L-serine ratios could result from the efficient reabsorption of L-serine by B0AT3, coupled with the DAAO activity that degrades intracellular D-serine reabsorbed by SMCTs. In IRI conditions, our enantiomeric amino acid profiling revealed low plasma L-serine and high urinary L-serine (Figure supplements 1B, 2B). Additionally, the proteomic analysis indicated a reduction in B0AT3 levels (4h IRI/sham = 0.56 fold; 8h IRI/sham = 0.65 fold; Table S1). These observations suggest that the low L-serine reabsorption in IRI is a result of B0AT3 reduction.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

This is a thorough study that was reviewed previously under the old system. I think the authors have strengthened their findings and have no further suggestions.

We appreciate reviewer 1 for his/her effort and comments, which greatly contributed to improving this manuscript.

Reviewer #2 (Recommendations For The Authors):

The experiments seem to me to have been well performed and the data are readily available.

Weaknesses:

More than weakness I would speak of discussion points: I have a few suggestions that may help to make the paper more accessible to a general audience.

(1) In the Introduction, when the authors introduce the term "micromolecules", it would be beneficial to provide a precise definition or clarification of what they mean by this term. Adding a brief explanation may help the reader to better understand the context.

Following the reviewer’s comment, we have included the explanation of the micromolecule and membrane transport proteins in lines 41-43.

Manuscript lines 41-43

“Membrane transport proteins function to transport micromolecules such as nutrients, ions, and metabolites across membranes, thereby playing a pivotal role in the regulation of micromolecular homeostasis.”

(2) In line 91, I suggest specifying that this is a renal IRI model.

Following the reviewer’s comment, we have added the information that it is a renal IRI model of AKI (lines 90-92).

Manuscript lines 90-92

“We applied 2D-HPLC to quantify the plasma and urinary enantiomers of 20 amino acids of renal ischemia-reperfusion injury (IRI) mice, a model of AKI and AKI-to-CKD transition (Sasabe et al., 2014; Fu et al., 2018).”

(3) Lines 167-168 state that Asct2 is localised to the apical side of the renal proximal tubules. Is there any expression of Asct2 in other nephron segments?

To our knowledge, there is no report of ASCT2 expression in other nephron segments. Our immunofluorescent data of the ASCT2 staining in the whole kidney at the low magnification and another region of Figure 3 (below) as well as immunohistochemistry from Human Protein Atlas (update: Jun 9th, 2023) did not show a strong signal of ASCT2 expression in other regions besides the proximal tubules. Thus, we conclude that ASCT2 is mainly expressed in proximal tubules, but not in other nephron regions.

Author response image 1.

(4) Lines 225-226: Have the authors expressed the candidate genes in HEK293 cells with ASCT2 knockdown?

This experiment was done by expressing the candidate genes in the presence of endogenous ASCT2. We have added the information in lines 225-227 to emphasize this process.

Manuscript lines 225-227

“Based on this finding, we utilized cell growth determination assay as the screening method even in the presence of endogenous ASCT2 expression. HEK293 cells were transfected with human candidate genes without ASCT2 knockdown.”

(5) Lines 254-255: why was D-serine transport enhanced by ASCT2 knockdown in FlpInTRSMCT1 or 2 cells?

We appreciate the reviewer to point out this data. We apologize for causing the confusion in the text. The total amount of D-serine uptake in the cells did not enhance but the net uptake (uptake subtracted from the background) was increased. This enhancement is a result of the lower background by ASCT2 knockdown. We have revised the texts and explained this result in more detail (lines 256-258).

Manuscript lines 256-258

“In the cells with ASCT2 knockdown, the background level was lower, thereby enhancing the D-[3H]serine transport contributed by both SMCT1 and SMCT2 (the net uptake after subtracted with background) (Figure 5C).”

(6) Line 265: The low affinity of SMCT1 for D-serine alone makes it an unlikely transporter for urinary D-serine.

We admitted the reviewer’s concern about the low affinity of SMCT1. However, Km at mM range is widely accepted for several low-affinity amino acid transporters such as proton-coupled amino acid transporter PAT1 (Km = 2 – 5 mM; Miyauchi et al. Biochem J 2010), cationic amino acid transporter CAT2A (Km = 3 – 4 mM; Closs et al. Biochem 1997), and large-neutral amino acid transporter LAT4 (Km = 17 mM; Bodoy et al. J Biol Chem 2005). In the kidneys, many compounds are well-known to be reabsorbed by the low-affinity but high-capacity (high-expression) transporters. Similarly, D-serine was reported to be reabsorbed by the low-affinity transporter (Kragh-Hansen and Sheikh, J Physiol 1984; Shimomura et al. BBA 1988; Silbernagl et al. Am J Physiol Renal Physiol 1999). Moreover, amino acid profile showed urinary D-serine in the range of 100 – 200 µM (Figure supplement 2). This concentration range could drive SMCT1 function (Figure 5). Combined with the high and ubiquitous expression of SMCT1, we propose that SMCT1 is a low-affinity but highcapacity D-serine transporter in the kidneys.

snRNA-seq is a method that can directly compare the expression levels between different genes within the same cells. From Figure supplement 7, expression of SMCT1 is much more abundant than ASCT2. ASCT2 was presented in less than 10% of cell population. It is possible that 90% of the cells that do not express ASCT2 use SMCT1 to reabsorb Dserine.

We have revised the Discussion regarding this comment (lines 386-404).

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in the control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of Dserine transport systems.”

(7) Line 316: The authors state that there is a high tubular D-serine reabsorption in IRI and in line 424 there is an inactivation of DAAO during the pathology. This suggests that there is a reabsorption of D-serine mediated by a transport system in the basolateral membrane domain of proximal tubular cells. Do the authors have any information about this transporter?

We agree with the reviewer that transporters at the basolateral membrane are important to complete the D-serine reabsorption in the kidney, and have included this issue in the original manuscript. We stated that transport systems at the basolateral side are necessary to be analyzed in order to complete the picture of D-serine transport systems in the kidney (lines 481-483 of the revised manuscript). However, we did not have any strong candidates for basolateral D-serine transport systems. Because we analyzed the proteome of BBMV, which concentrates on the apical membrane proteins, the analysis did not detect several transporters at the basolateral side.

(8) In lines 462-463, the authors state: "It is suggested that PAT1 is less active at the apical membrane where the luminal pH is neutral". However, the pH of urine in the proximal tubules is normally acidic due to the high activity of NH3. I suggest rewording this sentence.

Thank you for your comment. Proximal tubule (PT) is the first and the main region to maintain acid-base homeostasis in the kidney. In PT cells, NH3 secretes H+ to titrate luminal HCO3- and creates CO2, which is absorbed into PT cells and produces "new intracellular HCO3-", which is subsequently reabsorbed into the blood. Although ion fluxes in PT is to maintain the pH homeostasis, the pH regulation in both luminal and intracellular PT cells is highly dynamic. We totally agree with the reviewer and to follow that, we have revised the text by emphasizing the pH around PT segments, rather than the final urine pH, and leaving the discussion open for the possibility of PAT1 function in PT of normal kidneys (lines 474481).

Manuscript lines 474-481

“PAT1, a low-affinity proton-coupled amino acid transporter (Km in mM range), has been found at both sub-apical membranes of the S1 segment and inside of the epithelia (The Human Protein Atlas: https://www.proteinatlas.org; updated on Dec 7th, 2022) (Sagné et al., 2001; Vanslambrouck et al., 2010). PAT1 exhibits optimum function at pH 5 - 6 but very low activity at pH 7 (Miyauchi et al., 2005; Bröer, 2008b). Future research is required to address the significance of PAT1 on D-serine transport in the proximal tubule segments where pH regulation is known to be highly dynamic (Boron, 2006; Nakanishi et al., 2012; Bouchard and Mehta, 2022; Imenez Silva and Mohebbi, 2022).”

Reviewer #3 (Recommendations For The Authors):

The authors proposed that the increased expression of ASCT2, even together with the decreased expression of SMCT1/2, causes the increased renal reabsorption of D-serine that occurs in IRI. In the discussion, the main argument to sustain this hypothesis is the higher apparent affinity for D-serine of ASCT2 (<200 uM Km) versus SMCT1 (3.4 mM Km). In the Discussion section (page 18- 1st complete paragraph), the authors indicate that the Mass Spec intensities of SMCT1 and 2 are two and one order of magnitude higher respectively than that of ASCT2. This suggests that SMCT1 is clearly more expressed than ASCT2 in control conditions. IRI increments ASCT2 protein expression in brush-border membrane vesicle from kidney 1.6 folds and decreases that of SMCT1 0.6 folds. How this fold changes, even taking into account the lower Km of ASCT2 versus SMCT1 would explain the dramatic changes in the D-/L-serine ratios in plasma and urine in IRI? The authors might discuss whether other transport characteristics, even unknown (e.g., a higher turnover rate of ASCT2 vs SMCT1), would also contribute to the higher D-serine reabsorption in IRI.

SMCT1 shows some enantiomer selectivity for D- vs L-serine. At 50 uM concentration the transport is almost double for D. vs L-serine, but is ASCT2 stereoselective between the two enantiomers of serine? Some of the authors of this manuscript showed in a previous paper that the basolateral transporter Asc1 also participates in the accumulation of D-serine in serum caused by renal tubular damage. (Serum D-serine accumulation after proximal renal tubular damage involves neutral amino acid transporter Asc-1. Suzuki M et al. Sci Rep. 2019 Nov 13;9(1):16705 (PMID: 31723194)). Asc1 shows no stereoselectivity between L- and D-serine. Can the authors discuss possible mechanisms resulting in increased renal reabsorption of Dserine than L-serine in IRI with the participation of transporters with modest stereoselectivity for D- vs L-serine?

We appreciate the reviewer’s comments on the degree of protein alteration in proteomics, the functional contributions of ASCT2 and SMCTs, and the alteration of D/L ratios. We have included the possibilities of the technical concerns and the discussion on the roles of ASCT2 and SMCTs as follows.

• Regarding the expression levels, proteomics and snRNA-seq showed the same tendency that ASCT2 increase and SMCTs decrease in IRI conditions. However, the degrees of alterations are more contrast in snRNA-seq. This may be due to the difference in quantification methods and probably points out the underestimated quantification of membrane transport proteins in label-free proteomics. The accuracy of protein quantifications in the label-free proteomics are often impacted by the presence of post-translational modifications and multiple trans-membrane domains like in the case of the membrane transport proteins (Schey KL et al. Biochemistry 2015, doi: 10.1021/bi301604j). Alternative methods of quantitative proteomics may be added in the future for a more thorough comparison. We have added this issue in lines 351-356 of the revised version.

Manuscript lines 351-356

“When evaluating the extent of gene/protein alterations between the control and IRI conditions, we observed that the gene alterations of both Asct2 and Smcts, as revealed by snRNA-sequencing, are more pronounced than the protein alteration ratios obtained from proteomics. This discrepancy may stem from difficulty in the quantification method, especially for membrane transport proteins in label-free quantitative proteomics.”

• For the functional contributions of ASCT2 and SMCTs in the kidney, we admitted the reviewer’s concern about the low affinity of SMCT1. Following the reviewer’s comment, we have included other factors besides transport affinities, e.g. expression levels and turnover rates of the transporters. From the results of both proteomics and snRNA-seq, ASCT2 expression is significantly lower than SMCTs in the normal conditions. snRNA-seq showed that ASCT2 was presented in less than 10% of the cell population (Figure supplement 7). We propose that most of the cells that do not express ASCT2 may use SMCT1 to reabsorb D-serine. This topic was included in the revised manuscript lines 386-404.

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in the control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of D-serine transport systems.”

• As for the dramatic alterations of D/L-serine ratios juxtaposed with minimal changes in ASCT2 and SMCTs expression level, we cautiously refrain from drawing a definitive conclusion regarding the entire mechanism. This caution is grounded in the scientific understanding of a comprehensive elucidation of both L-serine transport systems and D-serine transport systems at both apical and basolateral membranes. Nevertheless, we would like to suggest a mechanism at the apical membrane based on the current knowledge.

For D-serine transport systems, we found ASCT2 and SMCTs contributions in this study. Meanwhile, L-serine was previously reported to be mediated mainly by the neutral amino acid transporters B0AT3 (in particular B0AT3; Bröer et al. Physiol Rev 2008; Singer et al. J Biol Chem 2009). Hence, the mechanism behind the alterations of D/L-serine ratios should include B0AT3 functions as well. In IRI conditions, B0AT3 decreased 1.8 fold. We suggest that high ratios of D-/L-serine in IRI conditions are a combined outcome of 1) increase of D-serine reabsorption by ASCT2 enhancement and SMCTs reduction, and 2) decrease of L-serine reabsorption by B0AT3. We have included this suggestion in the Discussion lines 438-451.

Manuscript lines 438-451

“The enantiomeric profiles of serine revealed distinct plasma D/L-serine ratios, with low ratios in the normal control but elevated ratios in IRI, despite the weak stereoselectivity of ASCT2 (Figure 1B). This observation suggested the differential renal handling of D-serine compared to L-serine. While we identified SMCTs as a Dserine transport system, it has been reported that L-serine reabsorption is mediated by B0AT3 (Singer et al., 2009). We propose that the alterations in plasma and urinary D/Lserine ratios are the combined outcomes of: 1) transport systems for L-serine, and 2) transport systems for D-serine. In normal kidneys, the low plasma D/L-serine ratios could result from the efficient reabsorption of L-serine by B0AT3, coupled with the DAAO activity that degrades intracellular D-serine reabsorbed by SMCTs. In IRI conditions, our enantiomeric amino acid profiling revealed low plasma L-serine and high urinary L-serine (Figure supplements 1B, 2B). Additionally, the proteomics analysis indicated a reduction in B0AT3 levels (4h IRI/sham = 0.56 fold; 8h IRI/sham = 0.65 fold; Table S1). These observations suggest that the low L-serine reabsorption in IRI is a result of B0AT3 reduction.”

• In the case of Asc-1, it was reported to be a D-serine transporter in the brain (Rosenberg et al. J Neurosci 2013). Suzuki et al. 2019 showed the increase of Asc-1 in cisplatin-induced tubular injury. Notably, the mRNA of Asc-1 is predominantly found in Henle’s loop, distal tubules, and collecting ducts but not in proximal tubules, and its protein expression level is dramatically low in the kidney (Human Protein Atlas: update on Jun 19, 2023). Furthermore, in this study, Asc-1 expression was not detected in the brush border membrane proteome. Consequently, we have decided not to include Asc-1 in the Discussion of this study, which primarily focuses on the proximal tubules.

eLife assessment

This study shows compelling evidence that the less common D-serine stereoisomer is transported in the kidney by the neutral amino acid transporter ASCT2 and that it is a non-canonical substrate for sodium-coupled monocarboxylate transporter SMCTs. With a multi-hierarchical approach, this important study further shows that Ischemia-Reperfusion Injury in the kidney causes a specific increment in renal reabsorption carried out, in part, by ASCT2.

Reviewer #1 (Public Review):

Most amino acids are stereoisomers in the L-enantiomer, but natural D-serine has also been detected in mammals and its levels shown to be connected to a number of different pathologies. Here, the authors convincingly show that D-serine is transported in the kidney by the neutral amino acid transporter ASCT2 and as a non-canonical substrate for the sodium-coupled monocarboxylate transporter SMCTs. Although both transport D-serine, this important study further shows in a mouse model for acute kidney injury that ASCT2 has the dominant role.

Strengths:

The paper combines proteomics, animal models, ex vivo transport analyses and in vitro transport assays using purified components. The exhaustive methods employed provide compelling evidence that both transporters can translocate D-serine in the kidney.

Weakness:

In the model for acute kidney injury the SMCTs proteins were not showing a significant change in expression levels and were rather analysed based on other, circumstantial evidence. Although its clear SMCTs can transport D-serine its physiological role is less obvious compared to ASCT2.

Reviewer #2 (Public Review):

Summary:

The manuscript "A multi-hierarchical approach reveals D-1 serine as a hidden substrate of sodium-coupled monocarboxylate transporters" by Wiriyasermkul et al. is a resubmission of a manuscript, which focused first on the proteomic analysis of apical membrane isolated from mouse kidney with early ischemia- reperfusion injury (IRI), a well-known acute kidney injury (AKI) model. In a second part, the transport of D-serine by Asct2, Smct1, and Smct2 has been characterized in detail in different model systems, such as transfected cells and proteoliposomes.

Strengths:

A major problem with the first submission was the explanation of the link between the two parts of the manuscript: it was not very clear why the focus on Asct2, Smct1 and Smct2 was a consequence of the proteomic analysis. In the present version of the manuscript, the authors have focused on the expression of membrane transporters in the proteome analysis, thus making the reason for studying Asct2, Smct1 and Smct2 transporters more clear. In addition, the authors used 2D-HPLC to measure plasma and urinary enantiomers of 20 amino acids in plasma and urine samples from sham and ischaemia-reperfusion injury (IRI) mice. The results of this analysis demonstrated the value of D-serine as a potential marker of renal injury. These changes have greatly improved the manuscript and made it more convincing.

Weaknesses:

More than weakness I would speak of discussion points: I have a few suggestions that may help to make the paper more accessible to a general audience.<br /> (1) In the Introduction, when the authors introduce the term "micromolecules", it would be beneficial to provide a precise definition or clarification of what they mean by this term. Adding a brief explanation may help the reader to better understand the context.<br /> (2) In line 91, I suggest specifying that this is a renal IRI model.<br /> (3) Lines 167-168 state that Asct2 is localised to the apical side of the renal proximal tubules. Is there any expression of Asct2 in other nephron segments?<br /> (4) Lines 225-226: Have the authors expressed the candidate genes in HEK293 cells with ASCT2 knockdown?<br /> (5) lines 254-255: why was D-serine transport enhanced by ASCT2 knockdown in FlpInTR-SMCT1 or 2 cells?<br /> (6) line 265: The low affinity of SMCT1 for D-serine alone makes it an unlikely transporter for urinary D-serine.<br /> (7) line 316: The authors state that there is a high tubular D-serine reabsorption in IRI and in line 424 that there is an inactivation of DAAO during the pathology. This suggests that there is a reabsorption of D-serine mediated by a transport system in the basolateral membrane domain of proximal tubular cells. Do the authors have any information about this transporter?<br /> (8) in lines 462-463, the authors state: "It is suggested that PAT1 is less active at the apical membrane where the luminal pH is neutral". However, the pH of urine in the proximal tubules is normally acidic due to the high activity of NH3. I suggest rewording this sentence.

Reviewer #3 (Public Review):

Summary:

The main objective of this work has been to delve into the mechanisms underlying the increment of D-serine in serum, as a marker of renal injury.

Strengths:

With a multi-hierarchical approach, the work shows that Ischemia reperfusion injury in kidney causes a specific increment in renal reabsorption of D-serine that, at least in part, is due to the increased expression of the apical transporter ASCT2. In the way, the authors revealed that SMCT1 also transports D-serine.

The manuscript also supports that increased expression of ASCT2, even together with the parallel decreased expression of SMCT1, in renal proximal tubules underlies the increased reabsorption of D-serine responsible of the increment of this enantiomer in serum in a murine model of ischemia reperfusion injury.

Weaknesses:

Remains to be clarified whether ASCT2 has substantial stereospecificity in favor of D- versus L-serine to sustain a ~10-fold decreased in the ratio D-serine/L-serine in the urine of mouse under ischemia reperfusion injury (IRI).<br /> It is not clear how the increment in the expression of ASCT2, in parallel with the decreased expression of SMCT1, results in increased renal reabsorption of D-serine in IRI.

I am satisfied with the changes the authors have introduced in the text of the revised version of their manuscript.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

(1) More explanation/description of Fig 3C and 3D would be helpful for readers, including the color code of 3D and black lines shown in both panels.

We have added more description to the legend of Figure 3, and we have used the same color code as in Figure 2, which we now specifically note in the figure legend as well.

(2) Differences between cranial and trunk NCC could be experimentally shown or discussed. Fig 4C shows some differences between these two populations, but in situ, results using Dlc1/Sp5/Pak3 probes in the trunk region may be informative, like Fig 5 supplement 2 for cranial NCCs.

This is an important point. The focus of our study was on cranial neural crest cells, and the single cell sequencing data is therefore truly reflective of only cranial neural crest cells. We have not functionally tested for the roles of Dlc1/Sp5/Pak3 in trunk neural crest cells, however, based on the expression and loss-of-function phenotypes of Sp5 or Pak3 knockout mice, we predict they individually may not play a significant role. It remains plausible that Dlc1 could play an important role in the delamination of trunk neural crest cells, but we have not tested that definitively. Nonetheless, Sabbir et al 2010 showed in a gene trap mouse mutant that Dlc1 is expressed in trunk neural crest cells. Regarding the similarities and differences between cranial and trunk neural crest cells as noted by the reviewer with respect to Figure 4, it’s important to recognize the temporal differences illustrated in Figure 4. Neural crest cell delamination proceeds in a progressive wave from anterior to posterior, but also that the analysis was designed to quantify cell cycle status before and during neural crest cell delamination. We have compared cranial and trunk neural crest cells in more detail in the discussion and also speculate what might happen in the trunk based on what we know from other species.

(3) Discussion can be added about the potential functions of Dlc1 for NCC migration and/or differentiation based on available info from KO mice.

We have added specific details regarding the published Dlc1 knockout mouse phenotype to the discussion, particularly with respect to the craniofacial anomalies which included frontonasal prominence and pharyngeal arch hyperplasia, and defects in neural tube closure and heart development. Although the study didn’t investigate the mechanisms underpinning the Dlc1 knockout phenotype, the craniofacial morphological anomalies would be consistent with a deficit in neural crest cell delamination reducing the number of migrating neural crest cells, as we observed in our Dlc1 knockdown experiments.

Reviewer #2 (Recommendations For The Authors):

The authors used the (Tg(Wnt1-cre)11Rth Tg(Wnt1-GAL4)11Rth/J) line but work from the Bush lab (see Lewis et al., 2013) has demonstrated fully penetrant abnormal phenotypes that affect the midbrain neuroepithelium, increased CyclinD1 expression and overt cell proliferation as measured by BrdU incorporation. The authors should explain why they used this mouse line instead of the Wnt1-Cre2 mice (129S4-Tg(Wnt1-cre)1Sor/J) in the Jackson Laboratory (which lacks the phenotypic effects of the original Wnt1-Cre line), or a "Cre-only" control, or at a minimum explain the steps they took to ensure there were no confounding effects on their study, especially since cell proliferation was a major outcome measure.

This is an important point, and we thank the reviewer for raising it. Yes, it has been reported that the original Wnt1Cre mice exhibit a midbrain phenotype (Ace et al. 2013). However, it has also been noted that Wnt1Cre2 can exhibit recombination in the male germline leading to ubiquitous recombination (Dinsmore et al., 2022). Therefore, to avoid any potential for bias, we used an equal number of cells derived from the Wnt1 and F10N transgenic line embryos in our scRNA-seq, and this included multiple non-Cre embryos. Our scRNA-seq analysis was therefore not dependent upon Wnt1-Cre, but also because we used whole heads not fluorescence sorted cells. However, Wnt1-Cre lineage tracing was advantageous from a computational perspective to help define cells that were premigratory and migratory in concert with Mef2c-lacZ ¬based on their expression of YFP, LacZ or both. We note these specifics more clearly in the methods.

The Results section (line 122) states that scRNA-seq was performed on dissociated cranial tissues but the Methods section (lines 583-584) implies that whole E8.5 mouse embryos were dissociated. Which was dissociated, whole embryos or just cranial tissues? Obviously, the latter would be a better strategy to enrich for cranial neural crest, but the authors also examine the trunk neural crest. This should be clarified in the text.

We apologize that some of the details regarding the tissue isolation were confusing and we have clarified this in the methods and the text. For the record, after isolating E8.5 embryos, we then dissected the head from those embryos, and performed scRNA-seq on dissociated cranial tissues. As the reviewer correctly noted, this approach strategically enriches for cranial neural crest cells.

The authors do not justify why they chose a knockdown strategy, which has its limitations including its systemic injection into the amniotic cavity, its likely global and more variable effects, and its need to be conducted in culture. Why the authors did not instead use a Wnt1-Cre-mediated deletion of Dlc1, which would have been "cleaner" and more specific to the neural crest, is not clear (maybe so they could specifically target different Dcl1 isoforms?). Also, the authors use Sox10 as a marker to count neural crest cells, but Sox10 may only label a subset of neural crest cells and thus some unaffected lineages may not have been counted. The authors should mention what is known about the regulation of Dcl1 by Sox10 in the neural crest. Although the data are persuasive, a second marker for counting neural crest cells following knockdown would make the analysis more robust. Can the authors explain why they did not simply use the Mef2c-F10N-LacZ line and count LacZ-positive cells (if fluorescence signal was required for the quantification workflow, then could they have used an anti-beta Galactosidase antibody to label cells)?

We thank the reviewer for raising these important considerations. It has previously been noted that although Wnt1-Cre is the gold standard for conditional deletion analyses in neural crest cell development, especially migration and differentiation, it is not a good tool for functional studies of the specification and delamination of neural crest cells due to the timing of Wnt1 expression and Cre activation and excision (see Barriga et al., 2015). Therefore, we chose a knockdown strategy instead, and also because it allows us to more rapidly evaluate gene function. We agree that there are limitations to the approach with respect to variability, however, this is outweighed by the ability to repeatedly perform the knockdown at multiple and more relevant temporal stages such as E7.5 (which is prior to the onset of Wnt1-Cre activity), as well as target different isoforms, and also treat large numbers of embryos for quantitative analyses. The advantage of using Sox10 as a marker for counting neural crest cells is that at the time of analysis, cranial neural crest cells are still migrating towards the frontonasal prominences and pharyngeal arches, and the overwhelming majority of these cells are Sox10 positive. Moreover, we can therefore assay every Dlc1 knockdown embryo for Sox10 expression and count the number of migrating neural crest cells. The limitation of using the Mef2c-F10N-LacZ line is that this transgenic line is maintained as a heterozygote, and thus only half the embryos in a litter could reasonably be expected to be lacZ+. But combining Sox10 and Mef2c-F10N-LacZ fluorescent immunostaining for similar analyses in the future is a great idea.

Reviewer #3 (Recommendations For The Authors):

The putative intermediate cells differentially express mRNAs for genes involved in cell adhesion, polarity, and protrusion relative to bona fide premigratory cells (Fig. 2E). This is persuasive evidence, but only differentially expressed genes are shown. Discussing those markers that have not yet changed, e.g. Cdh1 or Zo1 (?), would be instructive and help to clarify the order of events.

We thank the author for this suggestion and we have provided more detail about adherens junction and tight junctions. Cdh1 is not expressed, and although Myh9 and Myh10 are expressed, we did not detect any significant changes. ZO1 is a tight junction protein encoded by the gene Tjp1, which along with other tight junctions protein encoding genes, is downregulated in intermediate NCCs as shown in the Figure 2E.

It is unclear whether the two putative intermediate state clusters differ other than their stage of the cell cycle. Based on the trajectory analysis in Fig. 3C-D, the authors state that these two populations form simultaneously and independently but then merge into a single population. However, without further differential expression, it seems more plausible that they represent a single population that is temporarily bifurcated due to cell cycle asynchrony.

We have addressed the cell cycle question in the discussion by noting that while it is possible the transition states represent a single population that is temporarily bifurcated due to cell cycle asynchrony, if this were true, then we should expect S phase inhibition to eliminate both transition state groups. Instead, our trajectory analyses suggest that the transition states are initially independent, and furthermore, S phase inhibition did not affect delamination of the other population of neural crest cells.

The authors do not present an in-depth comparison of these neural crest intermediate states to previously reported cancer intermediate states. This analysis would reveal how similar the signatures are and thus how extrapolatable these and future findings in delaminating neural crest are to different types of cancer.

We have also added more detail to the discussion to address the potential for similarities and differences in neural crest intermediate states compared to previously reported cancer intermediate states. The challenge, however, is that none of the cancer intermediate states have been characterized at a molecular level. Nonetheless, with the limited molecular markers available, we have not identified any similarities so far, but our datasets are now available for comparison with future cancer EMP datasets.

The reduction in SOX10+ cells may be in part or wholly attributable to inhibition of proliferation AFTER delamination. Showing that there are premigratory NCCs in G2/M at ~E8.0 would bolster the argument that this population is present from the earliest stages.

The presence of premigratory neural crest cells in G2/M is shown by the scRNA-seq data and cell cycle staining data in the neural plate border.

Lines 248-249: The pseudo-time analysis in Fig 3C/D does indicate that the two most mature cell clusters (pharyngeal arch and frontonasal mesenchyme) may arise from common or similar migratory progenitors. However, given the decades of controversy about fate restriction of neural crest cells, the statement that "EMT intermediate NCC and their immediate lineages are not fate restricted to any specific cranial NCC derivative at this timepoint" should be toned down so as to not give the impression that they have identified common progenitors of ectomesenchyme and neuro/glial/pigment derivatives.

We appreciate this comment, because as the reviewer noted, there has been considerable literature and debate about the fate restriction and plasticity of neural crest cells, and indeed we did not intend to imply we have identified common progenitors of ectomesenchyme and neuro/glial/pigment derivatives. That can only be truly functionally demonstrated by clonal lineage tracing analyses. Rather, we interpret our pseudo-time analyses to indicate that irrespective of cell cycle status at the time of delamination, these two populations come together with equivalent mesenchymal and migratory properties, but in the absence of fate determination in the collective of cells. This does not mean that individual cells are common progenitors of both ectomesenchyme and neuro/glial/pigment derivatives. The nuance is important, and we address this more carefully in the text.

Lines 320-321: "...this overlap in expression was notably not observed in older embryos in areas where EMT had concluded". It is unclear whether the markers no longer overlap in older embryos (i.e. segregate to distinct populations) or are simply no longer expressed.

The data in Figure 5 demonstrates the dynamic and overlapping expression of Dlc1, Sp5 and Pak3 in the different clusters of cells as they transition from being neuroepithelial to mesenchymal. In contrast to Sp5 and Pak3, Dlc1 is not expressed by premigratory neural crest cells but is expressed at high levels in all EMT intermediate stage neural crest cells. Later as Dlc1 continues to be expressed in migrating neural crest cells, Pak3 and Sp5 are downregulated. But the absence of overlapping expression in the dorsolateral neural plate at the conclusion of EMT coincides with their downregulation in that territory.

In the final results section on Dlc1, the previously published mutant mouse lines are referenced as having "craniofacial malformation phenotypes". The lack of detail given on what those malformations are (assuming descriptions are available) makes the argument that they may be related to insufficient delamination less persuasive. The degree of knockdown correlates so well with the percentage reduction in migratory neural crest (Fig. 6) that one would imagine a null mutant to have a very severe phenotype.

The inference from the reviewer is correct and indeed Dlc1 null mutant mice do have a severe phenotype. We have added more specific details regarding the craniofacial and other phenotypes of the Dlc1 mutant mice to the discussion. Of note the frontonasal prominences and the pharyngeal arches are hypoplastic in E10.5 Dlc1 mutant embryos, which would be consistent with a neural crest cell deficit. Although a deficit in neural crest cells can be caused my multiple distinct mechanisms, our Dlc1 knockdown analyses suggest that the phenotype is due to an effect on neural crest cell delamination which diminishes the number of migrating neural crest cells.

Use the same y-axis for Fig. 4C/D

This has been corrected.

Fig. 6C: Please note in the panel which gene is being measured by qPCR

This has been corrected to denoted Dlc1.

Lines 108-117: More concise language would be appropriate here.

As requested, we were more succinct in our language and have shortened this section.

The SABER-FISH images are very dim. I realize the importance of not saturating the pixels, but the colors are difficult to make out.

We thank the reviewer for pointing this out and have endeavored to make the SABER-FISH images brighter and easier to see.

eLife assessment

This fundamental study reports compelling findings that intermediate states exist in epithelial-mesenchymal transition (EMT) during natural development and differentiation of mammalian neural crest cells, similar to recent reports in cancer. The authors determined that there were at least two paths to delamination and migration - one that occurs during S-phase of cell cycle and another during G2/M phase, and that the process of delamination is not restricted to cell fate. Finally, the authors showed that expression of Dlc1 may be used to identify cells in an intermediate state of EMT as well as their spatial location in the mouse embryo. The work will be of interest to developmental biologists, neurobiologists and cancer researchers.

Reviewer #1 (Public Review):

Summary:

This describes the molecular identity of the intermediate status of cranial neural crest cells (NCCs) during the initial delamination process. Taking advantage of single-cell RNA seq, the authors identify new populations of cells during EMT characterized by a specific set of gene expressions, including Dlc1. Promigratory cranial NCCs differentiate through different trajectories depending on their cell cycle phases but converge into a common progenitor, then differentiate into mesenchymal cells expressing region-specific genes.

Strengths:

Single-cell RNA seq data convincingly support what the authors claim. This is the first time to identify intermediate states between premigratory and migratory cranial NCCs. Silencing one of the marker genes, Dlc1, reduces the migratory activity of cranial NCCs. These findings deepen our understanding of the mechanism of EMT in general.

Comments on revised version:

Weaknesses:

None after substantial revision.

Reviewer #2 (Public Review):

Zhao et al., focus on mechanisms through which cells convert from epithelium to mesenchyme and become migratory. This phenomenon of epithelial-to-mesenchymal transition (EMT) occurs during both embryonic development and cancer progression. During cancer progression, EMT seemingly includes cells at intermediate states as defined by the combinatorial expression of epithelial and mesenchymal markers. But the importance of these markers and the role of these intermediate states remains unclear. Moreover, whether EMT during development also involves equivalent intermediate cell states is not known. To address this gap in knowledge, the authors devise a strategy to identify and characterize changes that an embryonic population of cells called the cranial neural crest undergo as they delaminate from the neuroepithelium and become a highly migratory population of mesenchymal cells that ultimately give rise to a broad range of derivatives.

To isolate and study the neural crest, the authors use embryos collected at E8.5 from two transgenic mouse lines. Wnt1-Cre;RosaeYFP labels Wnt1-positive neuroepithelial cells in the dorsolateral neural plate, which includes pre-migratory neural crest that reside in the dorsal neuroectoderm and neural plate border before induction (as well as some other lineages). Mef2c-F10N-LacZ leverages a neural crest cell-specific enhancer of Mef2c to control LacZ expression in predominantly migratory neural crest. This dual genetic approach that allows the authors to distinguish and compare pre-migratory and migratory neural crest cells is a strength of the work.

To assay for the differential expression of genes involved in the EMT and migration of cranial neural crest, the authors perform single cell RNA sequencing (scRNA-seq) using current methods. A strength is a large sample size per mouse line, and relatively high numbers of single cells analyzed. The authors identify six major cell/tissue types present in mouse E8.5 cranial tissues using known markers, which they then segregate into a cranial neural crest cluster using a well-reasoned bioinformatic strategy. The cranial neural crest cluster contains pre-migratory and migratory cells that they partition further into five subclusters and then characterize using the differential expression and combinatorial patterns of neural crest specifier genes, markers of pre-migratory neural crest, markers of early versus late migratory neural crest, markers of undifferentiated versus differentiated neural crest, tissue-specific markers, and region-specific markers. One weakness is that there is little attempt to map potential novel genes and/or pathways that also distinguish these clusters.

The authors then go on to subdivide the five cranial neural crest subclusters into almost two dozen smaller subclusters, again using the combinatorial expression of known markers (e.g., neural crest genes, cell junction genes, and cell cycle genes). A weakness is that the marker analysis and accompanying interpretation of the results relies heavily on the purported roles of different genes as described in the published work of others, which potentially introduces some untested assumptions and a bit of hand-waving into the study. Moreover, the limited correlation between mRNA and protein abundance for cell cycle markers is well documented in the literature but the authors rely heavily on gene expression to determine cell cycle status. Even though the authors add a compelling Edu/pHH3 double-labeling experiment and cell cycle inhibition studies, the work would be strengthened by including some analysis of protein expression to see if the cell cycle correlations hold up. Nonetheless, the subcluster and cell cycle analyses lead the authors to conclude that there are a series of intermediate cell states between neural crest EMT and delamination, and that cell cycle regulation is a defining feature and necessary component of those states. These novel findings are generally well supported by the data.

To test if there are spatiotemporal differences in the localization of neural crest cells during EMT in vivo, the authors apply a cutting-edge technique called signal amplification by exchange reaction for multiplexed fluorescent in situ hybridization (SABER-FISH), which they validate using standard in situ hybridization. The authors select specific marker genes that seem justified based on their scRNA-seq dataset, and they generate a series of convincing images and quantitative data that add valuable depth to the story.

As a functional test of their hypothesis that one of the genes indicative of an EMT intermediate stage (i.e., Dlc1) is essential for neural crest migration, the authors use a lentivirus-mediated knockdown strategy. A strength is that the authors include appropriate scramble and cell death controls as part of their experimental design.

The authors use Sox10 as a marker to count neural crest cells, but Sox10 may only label a subset of neural crest cells and thus some unaffected lineages may not have been counted. Although the data are persuasive, a second marker for counting neural crest cells following knockdown would make the analysis more robust.

Overall, this is a first-rate study with many more strengths than weaknesses. The authors generate high quality data, and their interpretations are reasonable and balanced. Another strength is the writing, which is clear and well organized, and the figures (including supplemental), which are excellent and provide unambiguous visualization of some very complex data sets. The methods are state-of the art and are effectively executed, and they will be useful to the broader cell and developmental biology community. The work contains well-substantiated findings and supports the conclusion that EMT is a highly dynamic, multi-step process, which was previously thought to be more-or-less binary. Such findings will alter the way the field thinks about EMT in neural crest and the work will likely serve as an important example alongside cancer metastasis.

Reviewer #3 (Public Review):

Summary:

Zhao et al. address the question of whether intermediate states of the epithelial-to-mesenchymal transition (EMT) exist in a natural developmental context as well as in cancer cells. This is important not only for our understanding of these developmental systems but also for their development as resources for new anti-cancer approaches. Guided by single-cell RNA sequencing analysis of delaminating mouse cranial neural crest cells, they identify two distinct populations with transcriptional signatures intermediate between neuroepithelial progenitors and migrating crest. Both clusters are also spatially intermediate and are actively cycling, with one in S-phase and one in G2/M. They show that blocking progression through S phase prior to the onset of delamination and knockdown of intermediate state marker Dlc1 both reduce the number of migratory cells that have completed EMT. Overall, the work provides a modern take and new insights into the classical developmental process of neural crest delamination.

Strengths:

• Deep analysis of the scRNAseq dataset revealed previously unappreciated cell populations intermediate between premigratory and migratory crest.<br /> • The observation that delaminating/intermediate neural crest cells appear to be in S or G2/M phase is interesting and worth reporting, though the ultimate significance remains unclear, given that they do not make distinct derivatives depending on their cycle state.<br /> • The authors employ new methods for multiplex spatial imaging to more accurately define their populations of interest and their relative positions.<br /> • The authors present evidence that intermediate state gene Dlc1 (a Rho GAP) is not just a marker but functionally required for neural crest delamination in mouse, as previously shown in chicken.

Weaknesses:

• Similar experiments involving blockade of cell cycle progression and Dlc1 dose manipulation were previously performed in chick models, as noted in the discussion. The newly-defined intermediate states give added context to the results, but they are not entirely novel.

eLife assessment

This valuable study further discloses the function of LRRK2 in BDNF-dependent synaptic processes in identifying postsynaptic actin cytoskeleton as a convergent site of LRRK2 pathophysiological activity. Multiple approaches in different cellular models provide mostly solid (but at times preliminary) evidence to support (many) of the conclusions, overall consistent with bioinformatics analyses covering previously published work. While an exciting start that should be pursued, examples are suggested by reviewers to add in additional experimentation to better support the expansive interpretation. The identification of mechanisms of LRRK2 action at the synapse is considered highly significant, as better knowledge in this regard may provide insight into why dopaminergic cells die with over-active LRRK2.

Reviewer #1 (Public Review):

Summary:

LRRK2 protein is familially linked to Parkinson's disease by the presence of several gene variants that all confer a gain-of-function effect on LRRK2 kinase activity.

The authors examine the effects of BDNF stimulation in immortalized neuron-like cells, cultured mouse primary neurons, hIPSC-derived neurons, and synaptosome preparations from the brain. They examine an LRRK2 regulatory phosphorylation residue, LRRK2 binding relationships, and measures of synaptic structure and function.

Strengths:

The study addresses an important research question: how does a PD-linked protein interact with other proteins, and contribute to responses to a well-characterized neuronal signalling pathway involved in the regulation of synaptic function and cell health?

They employ a range of good models and techniques to fairly convincingly demonstrate that BDNF stimulation alters LRRK2 phosphorylation and binding to many proteins. Some effects of BDNF stimulation appear impaired in (some of the) LRRK2 knock-out scenarios (but not all). A phosphoproteomic analysis of PD mutant Knock-in mouse brain synaptosomes is included.

Weaknesses:

The data sets are disjointed, conclusions are sweeping, and not always in line with what the data is showing. Validation of 'omics' data is very light. Some inconsistencies with the major conclusions are ignored. Several of the assays employed (western blotting especially) are likely underpowered, findings key to their interpretation are addressed in only one or other of the several models employed, and supporting observations are lacking.

As examples to aid reader interpretation:

(a) pS935 LRRK2 seems to go up at 5 minutes but goes down below pre-stimulation levels after (at times when BDNF-induced phosphorylation of other known targets remains very high). This is ignored in favour of discussion/investigation of initial increases, and the fact that BDNF does many things (which might indirectly contribute to initial but unsustained changes to pLRRK2) is not addressed.

(b) Drebrin coIP itself looks like a very strong result, as does the increase after BDNF, but this was only demonstrated with a GFP over-expression construct despite several mouse and neuron models being employed elsewhere and available for copIP of endogenous LRRK2. Also, the coIP is only demonstrated in one direction. Similarly, the decrease in drebrin levels in mice is not assessed in the other model systems, coIP wasn't done, and mRNA transcripts are not quantified (even though others were). Drebrin phosphorylation state is not examined.

(c) The large differences in the CRISPR KO cells in terms of BDNF responses are not seen in the primary neurons of KO mice, suggesting that other differences between the two might be responsible, rather than the lack of LRRK2 protein.

(d) No validation of hits in the G2019S mutant phosphoproteomics, and no other assays related to the rest of the paper/conclusions. Drebrin phosphorylation is different but unvalidated, or related to previous data sets beyond some discussion. The fact that LRRK2 binding occurs, and increases with BDNF stimulation, should be compared to its phosphorylation status and the effects of the G2019S mutation.

Reviewer #2 (Public Review):

Taken as a whole, the data in the manuscript show that BDNF can regulate PD-associated kinase LRRK2 and that LRRK2 modifies the BDNF response. The chief strength is that the data provide a potential focal point for multiple observations across many labs. Since LRRK2 has emerged as a protein that is likely to be part of the pathology in both sporadic and LRRK2 PD, the findings will be of broad interest. At the same time, the data used to imply a causal throughline from BDNF to LRRK2 to synaptic function and actin cytoskeleton (as in the title) are mostly correlative and the presentation often extends beyond the data. This introduces unnecessary confusion. There are also many methodological details that are lacking or difficult to find. These issues can be addressed.

(1) The writing/interpretation gets ahead of the data in places and this was confusing. For example, the abstract highlights prior work showing that Ser935 LRRK2 phosphorylation changes LRRK2 localization, and Figure 1 shows that BDNF rapidly increases LRRK2 phosphorylation at this site. Subsequent figures highlight effects at synapses or with synaptic proteins. So is the assumption that LRRK2 is recruited to (or away from) synapses in response to BDNF? Figure 2H shows that LRRK2-drebrin interactions are enhanced in response to BDNF in retinoic acid-treated SH-SY5Y cells, but are synapses generated in these preps? How similar are these preps to the mouse and human cortical or mouse striatal neurons discussed in other parts of the paper (would it be anticipated that BDNF act similarly?) and how valid are SH-SY5Y cells as a model for identifying synaptic proteins? Is drebrin localization to synapses (or its presence in synaptosomes) modified by BDNF treatment +/- LRRK2? Or do LRRK2 levels in synaptosomes change in response to BDNF? The presentation requires re-writing to stay within the constraints of the data or additional data should be added to more completely back up the logic.

(2) The experiments make use of multiple different kinds of preps. This makes it difficult at times to follow and interpret some of the experiments, and it would be of great benefit to more assertively insert "mouse" or "human" and cell type (cortical, glutamatergic, striatal, gabaergic) etc.

(3) Although BDNF induces quantitatively lower levels of ERK or Akt phosphorylation in LRRK2KO preps based on the graphs (Figure 4B, D), the western blot data in Figure 4C make clear that BDNF does not need LRRK2 to mediate either ERK or Akt activation in mouse cortical neurons and in 4A, ERK in SH-SY5Y cells. The presentation of the data in the results (and echoed in the discussion) writes of a "remarkably weaker response". The data in the blots demand more nuance. It seems that LRRK2 may potentiate a response to BDNF that in neurons is independent of LRRK2 kinase activity (as noted). This is more of a point of interpretation, but the words do not match the images.

(4) Figure 4F/G shows an increase in PSD95 puncta per unit length in response to BDNF in mouse cortical neurons. The data do not show spine induction/dendritic spine density/or spine morphogenesis as suggested in the accompanying text (page 8). Since the neurons are filled/express gfp, spine density could be added or spines having PSD95 puncta. However, the data as reported would be expected to reflect spine and shaft PSDs and could also include some nonsynaptic sites.

(5) Experimental details are missing that are needed to fully interpret the data. There are no electron microscopy methods outside of the figure legend. And for this and most other microscopy-based data, there are few to no descriptions of what cells/sites were sampled, how many sites were sampled, and how regions/cells were chosen. For some experiments (like Figure 5D), some detail is provided in the legend (20 segments from each mouse), but it is not clear how many neurons this represents, where in the striatum these neurons reside, etc. For confocal z-stacks, how thick are the optical sections and how thick is the stack? The methods suggest that data were analyzed as collapsed projections, but they cite Imaris, which usually uses volumes, so this is confusing. The guide (sgRNA) sequences that were used should be included. There is no mention of sex as a biological variable.

(6) For Figures 1F, G, and E, how many experimental replicates are represented by blots that are shown? Graphs/statistics could be added to the supplement. For 1C and 1I, the ANOVA p-value should be added in the legend (in addition to the post hoc value provided).

(7) Why choose 15 minutes of BDNF exposure for the mass spec experiments when the kinetics in Figure 1 show a peak at 5 mins?

(8) The schematic in Figure 6A suggests that iPSCs were plated, differentiated, and cultured until about day 70 when they were used for recordings. But the methods suggest they were differentiated and then cryopreserved at day 30, and then replated and cultured for 40 more days. Please clarify if day 70 reflects time after re-plating (30+70) or total time in culture (70). If the latter, please add some notes about re-differentiation, etc.

(9) When Figures 6B and 6C are compared it appears that mEPSC frequency may increase earlier in the LRRK2KO preps than in the WT preps since the values appear to be similar to WT + BDNF. In this light, BDNF treatment may have reached a ceiling in the LRRK2KO neurons.

(10) Schematic data in Figures 5A and C and Figures 5B and E are too small to read/see the data.

eLife assessment

This is a valuable study of the spatial organization of innate granulomas following Chromobacterium violaceum infection and the expression of CC and CXC chemokines in the granuloma at several time points following infection. There is a wealth of information to be gained from this study. However, the analysis of these granulomas is incomplete, with room for orthogonal validation of some of the key findings with additional animals (using ISH or IHC), in addition to a more quantitative analysis of some of the currently more qualitative conclusions.

Reviewer #1 (Public Review):

Amason et al. investigated the formation of granulomas in response to Chromobacterium violaceum infection, aiming to uncover the cellular mechanisms governing the granuloma response. They identify spatiotemporal gene expression of chemokines and receptors associated with the formation and clearance of granulomas, with a specific focus on those involved in immune trafficking. By analyzing the presence or absence of chemokine/receptor RNA expression, they infer the importance of immune cells in resolving infection. Despite observing increased expression of neutrophil-recruiting chemokines, treatment with reparixin (an inhibitor of CXCR1 and CXCR2) did not inhibit neutrophil recruitment during infection. Focusing on monocyte trafficking, they found that CCR2 knockout mice infected with C. violaceum were unable to form granulomas, ultimately succumbing to infection.

The spatial transcriptomics data presented in the figures could be considered a valuable resource if shared, with the potential for improved and clarified analyses. The primary conclusion of the paper, that C. violaceum infection in the liver cannot be contained without macrophages, would benefit from clarification.

While the spatial transcriptomic data generated in the figures are interesting and valuable, they could benefit from additional information. The manual selection of regions of granulomas for analysis could use additional context - was the rest of the liver not sequenced, or excluded for other reasons? Including a healthy liver in the analysis could serve as a control for any lasting effects at the final time point of 21 days. Providing more context for the scalebars throughout the spatial analyses, such as whether the data are raw counts or normalized based on the number of reads per spatial spot, would be helpful for interpretation, as changes in expression could signal changes in the numbers of cells or changes in the gene expression of cells.

In Figure 4, qualitative measurements are valuable, but having an idea of the raw data for a few of the pursued chemokines/receptors would aid interpretation. It would also be beneficial to clarify whether the reported values are across all clusters and consider focusing on clusters with the greatest change in expression. Figures 5E and F would benefit from clarification regarding the x-axis units and whether the expression levels are summed across all clusters for each time point. Additionally, information on the sequencing depth of the samples would be helpful, particularly as shallow sequencing of RNA can result in poor capture of low-expression transcripts.

Regarding the conclusion of the essentiality of macrophages in granuloma formation, it may be prudent to further investigate the role of macrophages versus CCR2. Analyzing total cell counts in the liver after infection could provide insight into whether the decrease in the fraction of macrophages is due to decreased numbers or infiltration of other cell types. Consideration of experiments deleting macrophages directly, instead of CCR2, could provide more definitive evidence of the necessity of macrophage migration in containing infections.

Reviewer #2 (Public Review):

Summary:

In this study, Amason et al employ spatial transcriptomics and intervention studies to probe the spatial and temporal dynamics of chemokines and their receptors and their influence on cellular dynamics in C. violaceum granulomas. As a result of their spatial transcriptomic analysis, the authors narrow in on the contribution of neutrophil- and monocyte-recruiting pathways to host response. This results in the observation that monocyte recruitment is critical for granuloma formation and infection control, while neutrophil recruitment via CXCR2 may be dispensable.

Strengths:

Since C. violaceum is a self-limiting granulomatous infection, it makes an excellent case study for 'successful' granulomatous inflammation. This stands in contrast to chronic, unproductive granulomas that can occur during M. tuberculosis infection, sarcoidosis, and other granulomatous conditions, infectious or otherwise. Given the short duration of C. violaceum infection, this study specifically highlights the importance of innate immune responses in granulomas.

Another strength of this study is the temporal analysis. This proves to be important when considering the spatial distribution and timing of cellular recruitment. For example, the authors observe that the intensity and distribution of neutrophil- and monocyte-recruiting chemokines vary substantially across infection time and correlate well with their previous study of cellular dynamics in C. violaceum granulomas.

The intervention studies done in the last part of the paper bolster the relevance of the authors' focus on chemokines. The authors provide important negative data demonstrating the null effect of CXCR1/2 inhibition on neutrophil recruitment during C. violaceum infection. That said, the authors' difficulty with solubilizing reparixin in PBS is an important technical consideration given the negative result. On the other hand, monocyte recruitment via CCR2 proves to be indispensable for granuloma formation and infection control. I would hesitate to agree with the authors' interpretation that their data proves macrophages are serving as a physical barrier from the uninvolved liver. It is possible and likely that they are contributing to bacterial control through direct immunological activity and not simply as a structural barrier.

Weaknesses:

There are several shortcomings that limit the impact of this study. The first is that the cohort size is very limited. While the transcriptomic data is rich, the authors analyze just one tissue from one animal per time point. This assumes that the selected individual will have a representative lesion and prevents any analysis of inter-individual variability. Granulomas in other infectious diseases, such as schistosomiasis and tuberculosis, are very heterogeneous, both between and within individuals. It will be difficult to assert how broadly generalizable the transcriptomic features are to other C. violaceum granulomas. Furthermore, this undermines any opportunity for statistical testing of features between time points, limiting the potential value of the temporal data.

Another caveat to these data is the limited or incompletely informative data analysis. The authors use Visium in a more targeted manner to interrogate certain chemokines and cytokines. While this is a great biological avenue, it would be beneficial to see more general analyses considering Visum captures the entire transcriptome. Some important questions that are left unanswered from this study are:

What major genes defined each spatial cluster?

What were the top differentially expressed genes across time points of infection?

Did the authors choose to focus on chemokines/receptors purely from a hypothesis perspective or did chemokines represent a major signature in the transcriptomic differences across time points?

In addition to the absence of deep characterization of the spatial transcriptomic data, the study lacks sufficient quantitative analysis to back up the authors' qualitative assessments. Furthermore, the authors are underutilizing the spatial information provided by Visium with no spatial analysis conducted to quantify the patterning of expression patterns or spatial correlation between factors.

Impact:

The author's analysis helps highlight the chemokine profiles of protective, yet host protective granulomas. As the authors comment on in their discussion, these findings have important similarities and differences with other notable granulomatous conditions, such as tuberculosis. Beyond the relevance to C. violaceum infection, these data can help inform studies of other types of granulomas and hone candidate strategies for host-directed therapy strategies.

eLife assessment

This valuable study by Bartas and colleagues examined how patterns of monosynaptic input to specific cell types in the ventral tegmental area are altered by drugs of abuse. The authors applied a dimensionality reduction approach (principal component analysis) and showed that various drugs of abuse, and somewhat surprisingly the anesthesia alone (ketamine/xylasin), caused changes in the distribution of inputs labeled by the transsynaptic rabies virus. While there are some issues to be addressed, the evidence supporting the conclusions is overall solid, and provides information that is of value to the field, as well as a cautionary note on the interpretation of rabies virus-based tracing experiments.

Reviewer #1 (Public Review):

Summary:

In this study, the authors distinguished afferent inputs to different cell populations in the VTA using dimensionality reduction approaches and found significantly distinct patterns between normal and drug treatment conditions. They also demonstrated negative correlations of the inputs induced by drugs with gene expression of ion channels or proteins involved in synaptic transmission and demonstrated the knockdown of one of the voltage-gated calcium ion channels caused decreased inputs.

Weaknesses:

(1) For quantifications of brain regions in this study, boundaries were based on the Franklin-Paxinos (FP) atlas according to previous studies (Beier KT et al 2015, Beier KT et al 2019). It has been reported significant discrepancies exist between the anatomical labels on the FP atlas and the Allen Brain Atlas (ref: Chon U et al., Nat Commun 2019). Although a summary of conversion is provided as a sheet, the authors need to describe how consistent or different the brain boundaries they defined in the manuscript with Allen Brain Atlas by adding histology images. Also, I wonder how reliable the annotations were for over a hundred of animals with manual quantification. The authors should briefly explain it rather than citing previous studies in the Material and Methods Section.

(2) Regarding the ellipsoids in the PC, although it's written in the manuscript that "Ellipsoids were centered at the average coordinate of a condition and stretched one standard deviation along the primary and secondary axes", it's intuitively hard to understand in some figures such as Figure 2O, P and Figure S1. The authors need to make their data analysis methods more accessible by providing source code to the public.

(3) In histology images (Figure 1B and 3K), the authors need to add dashed lines or arrows to guide the reader's attention.

(4) In Figure 2A and G, apparently there are significant differences in other brain regions such as NAcMed or PBN. If they are also statistically significant, the authors should note them as well and draw asterisks(*).

(5) In Figure 2N about the spatial distribution of starter cells, the authors need to add histology images for each experimental condition (i.e. saline, fluoxetine, cocaine, methamphetamine, amphetamine, nicotine, and morphine) as supplement figures.

(6) In the manuscript, it is necessary to explain why Cacna1e was selected among other calcium ion channels.

Reviewer #2 (Public Review):

The application of rabies virus (RabV)-mediated transsynaptic tracing has been widely utilized for mapping cell-type-specific neural connectivities and examining potential modifications in response to biological phenomena or pharmacological interventions. Despite the predominant focus of studies on quantifying and analyzing labeling patterns within individual brain regions based on labeling abundance, such an approach may inadvertently overlook systemic alterations. There exists a considerable opportunity to integrate RabV tracing data with the global connectivity patterns and the transcriptomic signatures of labeled brain regions. In the present study, the authors take an important step towards achieving these objectives.

Specifically, the authors conducted an intensive reanalysis of a previously generated large dataset of RabV tracing to the ventral tegmental area (VTA) using dimension reduction methods such as PCA and UMPA. This reaffirmed the authors's earlier conclusion that different cell types in the VTA, namely dopamine neurons (DA) and GABAergic neurons, exhibit quantitatively distinct input patterns, and a single dose of addictive drugs, such as cocaine and morphine, induced altered labeling patterns. Additionally, the authors illustrate that distinct axes of PCA can discriminate experimental variations, such as minor differences in the injection site of viral tracers, from bona fide alternations in labeling patterns caused by drugs of abuse. While the specific mechanisms underlying altered labeling in most brain regions remain unclear, whether involving synaptic strength, synaptic numbers, pre-synaptic activities, or other factors, the present study underscores the efficacy of an informatics approach in extracting more comprehensive information from the RabV-based circuit mapping data.

Moreover, the authors showcased the utility of their previously devised bulk gene expression patterns inferred by the Allen Gene Expression Atlas (AGEA) and "projection portrait" derived from bulk axon mapping data sourced from the Allen Mouse Brain Connectivity Atlas. The utilization of such bulk data rests upon several limitations. For instance, the collection of axon mapping data involves an arbitrary selection of both cell type-specific and non-specific data, which might overlook crucial presynaptic partners, and often includes contamination from neighboring undesired brain regions. Concerns arise regarding the quantitativeness of AGEA, which may also include the potential oversight of key presynaptic partners. Nevertheless, the authors conscientiously acknowledged these potential limitations associated with the dataset.

Notably, building on the observation of a positive correlation between the basal expression levels of Ca2+ channels and the extent of drug-induced changes in RabV labeling patterns, the authors conducted a CRISPRi-based knockdown of a single Ca2+ channel gene. This intervention resulted in a reduction of RabV labeling, supporting that the observed gene expression patterns have causality in RabV labeling efficiency. While a more nuanced discussion is necessary for interpreting this result (see below), overall I commend the authors for their efforts to leverage the existing dataset in a more meaningful way. This endeavor has the potential to contribute significantly to our understanding of the mechanisms underlying alterations in RabV labeling induced by drugs of abuse.

Finally, drawing upon the aforementioned reanalysis of previous data, the authors underscored that a single administration of ketamine/xylazine anesthesia could induce enduring modifications in RabV labeling patterns for VTA DA neurons, specifically those projecting to the nucleus accumbens and amygdala. Given the potential impact of such alterations on motivational behaviors at a broader level, I fully agree that prudent consideration is warranted when employing ketamine/xylazine for the investigation of motivational behaviors in mice.

Specific Points:

(1) Beyond advancements in bioinformatics, readers may find it insightful to explore whether the PCA/UMPA-based approach yields novel biological insights. For example, the authors are encouraged to discuss more functional implications of PBN and LH in the context of drugs of abuse, as their labeling abundance could elucidate the PC2 axis in Fig. 2M.

2) While I appreciate the experimental data on Cacna1e knockdown, I am unclear about the rationale behind specifically focusing on Cacna1e. The logic behind the statement, "This means that expression of this gene is not inhibitory towards RABV transmission," is also unclear. Loss-of-function experiments only signify the necessity or permissive functions of a gene. In this context, Cacna1e expression levels are required for efficient RabV labeling, but this neither supports nor excludes the possibility that this gene expression instructively suppresses RabV labeling/transmission, which could be assessed through gain-of-function experiments.

Reviewer #3 (Public Review):

Summary:

Authors mapped monosynaptic inputs to dopamine, GABA, and glutamate neurons in VTA under different anesthesia methods, and under drugs (cocaine, morphine, methamphetamine, amphetamine, nicotine, fluoxetine). They found that input patterns under different conditions are separated, and identified some key brain areas to contribute to such separation. They also searched a database for gene expression patterns that are common across input brain areas with some changes by anesthesia or drug administration.

Strengths:

The whole-brain approach to address drug effects is appealing and their conclusion is clear. The methodology and motivation are clearly explained.

Weaknesses:

While gene expression analyses may not be related to their findings on the anatomical effects of drugs, this will be a nice starting point for follow-up studies.

Author Response

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

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

To resolve and further test the claim that TBI did not induce cell proliferation:

How many brains did they analyse? Sample sizes must be provided in Figure S1.

As per reviewer’s suggestion, we removed one of the unsupported claims shown in Figure S1. The original Figure S1 is shown below with the sample number added.

Author response image 1.

The authors could either improve the TBI method or the detection of cells in S-phase, mitosis or cycling. They could use PCNA-GFP or BrdU, EdU or FUCCI instead and at least provide evidence that they can detect cells in S-phase in intact brains. Timing is critical (ie cell cycle is longer than in larvae) so multiple time points should be tested. Or they could use pH3 but test more time points and rather large sample sizes. If they are not able to provide any evidence, then their lack of evidence is no evidence. The authors should consider removing pH3 and PCNA-GFP related claims instead.

We have removed pH3 and PCNA-GFP related results and claims.

Other unsupported claims:

Figure 2A-C is not very clear what they are showing, but it is not evidence of astrocyte hypertrophy. It does not have cellular resolution and does not show the cell size, membranes, nor number

(1) We have avoided the term “hypertrophy” and changed the description throughout the text to “astrocyte swelling”.

(2) Images in the resolution of Figure 2E and 2F were able to show the enlarged soma of astrocytes, suggesting swelling.

What is the point of using RedStinger in Figure 2?

We used RedStinger to label the astrocyte nuclei.

Figure S5 is not convincing, as anti-Pvr does not look localised to specific cells. Instead, it looks like uniform background. If they really think the antibody is localised, they should do double stainings with cell type specific markers. If the antibody does not work, then remove the data and the claim. They could test with RNAi knock-down in specific cell types and qRT-PCR which cells express pvr instead.

We have removed the claim that “Pvr is predominantly expressed in astrocytes” and changed the description to “Immunostainings using the anti-Pvr antibodies revealed that endogenous Pvr expression is low in the control brains, yet significantly enhanced upon TBI. Reducing Pvr expression, but not Pvr overexpression, in astrocytes blocked the TBI-induced increase of Pvr expression (Figure S5)”.

Figure S6: it is unclear what they are trying to show, but these data do not demonstrate that astrocytes do not engulf debris after TBI, as there isn't sufficient cellular resolution to make such claim. Firstly, they analyse one single cell per treatment. Secondly, the cell projections are not visible in these images, and therefore engulfment cannot be seen. The authors could remove the claim or visualise whether astrocytes phagocytose debris or not either using clones or with TEM.

We agree with the reviewer that our images do not have the resolution to make this claim. We have removed Figure S6 and corresponding text description.

On statistics:

The statistical analysis needs revising as it is wrong in multiple places, eg Fig.1F,G,H; Figure 2D. They only use Student t-tests. These can only be used when data are continuous, distributed uniformly and only two samples are compared; if more than 2 samples, distributed uniformly, then use One-Way ANOVA and multiple comparisons tests. If data are categorical, use Chi-Square.

We have double checked and compared the experimental group to the control separately using the Student t-tests throughout the study.

Other points for improvement:

Figure 2E,F: what are GFP puncta and how are they counted?

I. Each GFP puncta looks like a little circle, likely representing a functional or dysfunctional structure. The biology of the GFP puncta is currently unkonwn.

II. We used the ImageJ to quantify the GFP puncta:

(1) Image- type-8 bits

(2) Process-subtract background (Rolling ball radio:10)

(3) Image-Adjust-Threshold-Apply

(4) Analyze-Measure-set measurements-choose “area” “limit to threshold”-OK

(5) Count the puncta number in the choosing area.

(6) Get the number of puncta per square micron.

All genotypes must be provided (including for MARCM clones), currently they are not.

We have shown the full genotype in the corresponding legend.

Figure 7O,P indicate on figure that these are RNAi

We have revised the labels to RNAi in Figure 7O,P.

Reviewer #2 (Recommendations For The Authors):

Several typos are present in the text.

We have read the manuscript carefully and corrected typos throughout.

eLife assessment

This study represents a valuable finding on the neuron-glia communication and glial responses to traumatic brain injury (TBI). The data supporting the authors' conclusions on TBI analysis, RNA-seq on FACS sorted astrocytes, genetic analyses on Pvr-JNK/MMP1 are solid. However, cellular aspects of the response to TBI, statistical analysis, and molecular links between Pvr-AP1 are incomplete, which could be further strengthened in the future by more rigorous analyses.

Reviewer #1 (Public Review):

Li et al report that upon traumatic brain injury (TBI), Pvr signalling in astrocytes activates the JNK pathway and up-regulates the expression of the well-known JNK target MMP1. The FACS sort astrocytes, and carry out RNAseq analysis, which identifies pvr as well as genes of the JNK pathway as particularly up-regulated after TBI. They use conventional genetics loss of function, gain of function and epistasis analysis with and without TBI to verify the involvement of the JNK-MMP1 signalling pathway downstream of PVR. They also show that blocking endocytosis prolongs the involvement of this pathway in the TBI response.

The strengths are that multiple experiments are used to demonstrate that TBI in their hands damaged the BBB, induced apoptosis and increased MMP1 levels. The RNAseq analysis on FACS sorted astrocytes is nice and will be valuable to scientists beyond the confines of this paper. The functional genetic analysis is conventional, yet sound, and supports claims of JNK and MMP1 functioning downstream of Pvr in the TBI context.

For this revised version the authors have removed all the unsupported claims. This renders their remaining claims more solid. However, it has resulted in the loss of important cellular aspects of the response to TBI, limiting the scope and value of the work.

The main weakness is that novelty and insight are both rather limited. Others had previously published that both JNK signalling and MMP1 were activated upon injury, in multiple contexts (as well as the articles cited by the authors, they should also see Losada-Perez et al 2021). That Pvr can regulate JNK signalling was also known (Ishimaru et al 2004). The authors claim that the novelty was investigating injury responses in astrocytes in Drosophila. However, others had investigated injury responses by astrocytes in Drosophila before. It had been previously shown that astrocytes - defined as the Prospero+ neuropile glia, and also sharing evolutionary features with mammalian NG2 glia - respond to injury both in larval ventral nerve cords and in adult brains, where they proliferate regenerating glia and induce a neurogenic response (Kato et al 2011; Losada-Perez et al 2016; Harrison et al 2021; Simoes et al 2022). The authors argue that the novelty of the work is the investigation of the response of astrocytes to TBI. However, this is of somewhat limited scope. The authors mention that MMP1 regulates tissue remodelling, the inflammatory process and cancer. Exploring these functions further would have been an interesting addition, but the authors did not investigate what consequences the up-regulation of MMP1 after injury has in repair or regeneration processes.

The statistical analysis is incorrect in places, and this could affect the validity of some claims.

Altogether, this is an interesting and valuable addition to the repertoire of articles investigating neuron-glia communication and glial responses to injury in the Drosophila central nervous system (CNS). It is good and important to see this research area in Drosophila grow. This community together is building a compelling case for using Drosophila and its unparalleled powerful genetics to investigate nervous system injury, regeneration and repair, with important implications. Thus, this paper will be of interest to scientists investigating injury responses in the CNS using Drosophila, other model organisms (eg mice, fish) and humans.

Reviewer #3 (Public Review):

In this study, authors used the Drosophila model to characterize molecular details underlying traumatic brain injury (TBI). Authors used the transcriptomic analysis of astrocytes collected by FACS sorting of cells derived from Drosophila heads following brain injury and identified upregulation of multiple genes, such as Pvr receptor, Jun, Fos, and MMP1. Additional studies identified that Pvr positively activates AP-1 transciption factor (TF) complex consisting of Jun and Fos, of which activation leads to the induction of MMP1. Finally, authors found that disruption of endocytosis and endocytotic trafficking facilitates Pvr signaling and subsequently leads to induction of AP-1 and MMP1.

Overall, this study provides important clues to understanding molecular mechanisms underlying TBI. The identified molecules linked to TBI in astrocytes could be potential targets for developing effective therapeutics. The obtained data from transcriptional profiling of astrocytes will be useful for future follow-up studies. The manuscript is well-organized and easy to read.

However, the connection suggested by the authors between Pvr and AP-1, potentially mediated through the JNK pathway, lacks strong experimental support in my view. It's important to recognize that AP-1 activity is influenced by multiple upstream signaling pathways, not just the JNK pathway, which is the most well-characterized among them. Therefore, assuming that AP-1 transcriptional activity solely reflects the activity of the JNK pathway without additional direct evidence is unwarranted. To strengthen their argument, the study could benefit from direct evidence implicating the JNK pathway in linking Pvr to AP-1. This could be achieved through genetic studies involving mutants or transgenes targeting key components of the JNK pathway, such as Bsk and Hep, the Drosophila homologues of JNK and JNKK, respectively. Alternatively, employing p-JNK antibody-based techniques like Western blotting, while considering the potential challenges associated with p-JNK immunohistochemistry, could provide further validation. This important criticism regarding the molecular link between Pvr and AP-1 has been overlooked.

Author Response

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

eLife assessment

This study presents valuable findings on the roles of the axon growth regulator Sema7a in the formation of peripheral sensory circuits in the lateral line system of zebrafish. The evidence supporting the claims of the authors is solid, although further work directly testing the roles of different sema7a isoforms would strengthen the analysis. The work will be of interest to developmental neuroscientists studying circuit formation.

Public Reviews:

Reviewer #1 (Public Review):

In this work, Dasguta et al. have dissected the role of Sema7a in fine tuning of a sensory microcircuit in the posterior lateral line organ of zebrafish. They attempt to also outline the different roles of a secreted verses membrane-bound form of Sema7a in this process. Using genetic perturbations and axonal network analysis, the authors show that loss of both Sema7a isoforms causes abnormal axon terminal structure with more bare terminals and fewer loops in contact with presynaptic sensory hair cells. Further, they show that loss of Sema7a causes decreased number and size of both the pre- and post-synapse. Finally, they show that overexpression of the secreted form of Sema7a specifically can elicit axon terminal outgrowth to an ectopic Sema7a expressing cell. Together, the analysis of Sema7a loss of function and overexpression on axon arbor structure is fairly thorough and revealed a novel role for Sema7a in axon terminal structure. However, the connection between different isoforms of Sema7a and the axon arborization needs to be substantiated. Furthermore, an autocrine role for Sema7a on the presynaptic cell is not ruled out as a contributing factor to the synaptic and axon structure phenotypes.

Finally, critical controls are absent from the overexpression paradigm.

Comments: Thank you for your valuable comments. We have analyzed the hair cell scRNA transcriptome data of zebrafish neuromasts from published works and have not identified known expression of receptors of the Sema7A protein, particularly PlexinC1 and Integrin β1 molecules (reference 4 and 15) in hair cells. This result suggests that the Sema7A protein molecule, either secreted or membrane-bound, does not possess its cognate receptor to elicit an autocrine function on the hair cells. Moreover, the GPI-anchored Sema7A lacks a cytosolic domain. So it is unlikely that Sema7A signaling directly induces the formation of presynaptic ribbons. We propose that the decrease in average number and area of synaptic aggregates likely reflects decreased stability of the synaptic structures owing to lack of contact between the sensory axons and the hair cells, which has been identified in zebrafish neuromasts (reference 38).

Thank you for pointing missing critical control experiments. Additional control experiments (lines 333-346) with a new figure (Figure 5) have been added.

These issues weaken the claims made by the authors including the statement that they have identified differential roles for the GPI-anchored verses secreted forms of Sema7a on synapse formation and as a chemoattractant for axon arborization respectively.

Comments: We have rephrased our statement and argue in lines 428-430 that our experiments “suggest a potential mechanism for hair cell innervation in which a local Sema7Asec diffusive cue likely consolidates the sensory arbors at the hair cell cluster and the membrane-anchored Sema7A-GPI molecule guides microcircuit topology and synapse assembly.”

The manuscript itself would benefit from the inclusion of details in the text to help the reader interpret the figures, tools, data, and analysis.

Comments: We have made significant revisions to the text and figures to improve clarity and consistency of the manuscript.

Reviewer #2 (Public Review):

In this work, Dasgupta et al. investigates the role of Sema7a in the formation of peripheral sensory circuit in the lateral line system of zebrafish. They show that Sema7a protein is present during neuromast maturation and localized, in part, to the base of hair cells (HCs). This would be consistent with pre-synaptic Sema7a mediating formation and/or stabilization of the synapse. They use sema7a loss-of-function strain to show that lateral line sensory terminals display abnormal arborization. They provide highly quantitative analysis of the lateral line terminal arborization to show that a number of specific topological parameters are affected in mutants. Next, they ectopically express a secreted form of Sema7a to show that lateral line terminals can be ectopically attracted to the source. Finally, they also demonstrate that the synaptic assembly is impaired in the sema7a mutant. Overall, the data are of high quality and properly controlled. The availability of Sema7a antibody is a big plus, as it allows to address the endogenous protein localization as well to show the signal absence in the sema7a mutant. The quantification of the arbor topology should be useful to people in the field who are looking at the lateral line as well as other axonal terminals. I think some results are overinterpreted though. The authors state: "Our findings demonstrate that Sema7A functions both as a juxtracrine and as a secreted cue to pattern neural circuitry during sensory organ development." However, they have not actually demonstrated which isoform functions in HCs (also see comments below).

Comments: Thank you for making this point. To investigate the presence of both sema7a transcripts in the hair cells of the lateral-line neuromasts, we used the Tg(myo6b:actb1EGFP) transgenic fish to capture the labeled hair cells by fluorescence-activated cell sorting (FACS) and isolated total RNA. Using transcript specific DNA oligonucleotide primers, we have identified the presence of both sema7a transcript variants in the hair cell of the neuromast. Even though we have not developed transcript specific knockout animals, we speculate that the presence of both transcript variants in the hair cell implies that they function in distinct fashion. We have changed our interpretation in lines 32-34 to “Our findings propose that Sema7A likely functions both as a juxtracrine and as a secreted cue to pattern neural circuitry during sensory organ development.”

In future we will utilize the CRISPR/Cas9 technique to target the unique C-terminal domain of the GPI-anchored sema7a transcript variant. We believe that this will only perturb the formation of the full-length Sema7A protein and help us determine the role of the membrane-bound Sema7AGPI molecule as well as the Sema7Asec in sensory arborization and synaptic assembly.

In addition, they have to be careful in interpreting their topology analysis, as they cannot separate individual axons. Thus, such analysis can generate artifacts. They can perform additional experiments to address these issues or adjust their interpretations.

Comments: Thank you for this insightful comment. In a previous eLife publication from our laboratory, we utilized the serial blockface scanning electron micrograph (SBFSEM) technique to characterize the connectome of the neuromast microcircuit where patterns of innervation of all the individual axons can be delineated in five-days-old larvae (reference 8). However, the collective behavior of all the sensory axons that build the innervation network remained enigmatic, especially in a living animal during development. In this paper we addressed how the sensory-axon collective behaves around the clustered hair cells and build the innervation network in living animals during diverse developmental stages. Our analyses have not only identified how the axons associates with the hair cell cluster as the organ matures, but also discovered distinct topological features in the arbor network that emerges during organ maturation, which may influence assembly of postsynaptic aggregates (lines 384-403, Figure 6G-I). We believe that our quantitative approach to capture collective axonal behaviors and their topological attributes during circuit formation have highlighted the importance of understanding network assembly during sensory organ development.

Reviewer #3 (Public Review):

Summary:

This study demonstrates that the axon guidance molecule Sema7a patterns the innervation of hair cells in the neuromasts of the zebrafish lateral line, as revealed by quantifying gain- and loss-of function effects on the three-dimensional topology of sensory axon arbors over developmental time. Alternative splicing can produce either a diffusible or membrane-bound form of Sema7a, which is increasingly localized to the basolateral pole of hair cells as they develop (Figure 1). In sema7a mutant zebrafish, sensory axon arbors still grow to the neuromast, but they do not form the same arborization patterns as in controls, with many arbors overextending, curving less, and forming fewer loops even as they lengthen (Figure 2,3). These phenotypes only become significant later in development, indicating that Sema7a functions to pattern local microcircuitry, not the gross wiring pattern. Further, upon ectopic expression of the diffusible form of Sema7a, sensory axons grow towards the Sema7a source (Figure 4). The data also show changes in the synapses that form when mutant terminals contact hair cells, evidenced by significantly smaller pre- and post-synaptic punctae (Figure 5). Finally, by replotting single cell RNA-sequencing data (Figure 6), the authors show that several other potential cues are also produced by hair cells and might explain why the sema7a phenotype does not reflect a change in growth towards the neuromast. In summary, the data strongly indicate that Sema7a plays a role in shaping connectivity within the neuromast.

Strengths:

The main strength of this study is the sophisticated analysis that was used to demonstrate fine-level effects on connectivity. Rather than asking "did the axon reach its target?", the authors asked "how does the axon behave within the target?". This type of deep analysis is much more powerful than what is typical for the field and should be done more often. The breadth of analysis is also impressive, in that axon arborization patterns and synaptic connectivity were examined at 3 stages of development and in three-dimensions.

Weaknesses:

The main weakness is that the data do not cleanly distinguish between activities for the secreted and membrane-bound forms of Sema7a, which the authors speculate may influence axon growth and synapse formation respectively. The authors do not overstate the claims, but it would have been nice to see some additional experimentation along these lines, such as the effects of overexpressing the membrane-bound form,

Comments: We have accepted this useful suggestion. In lines 333-346 and in Figure 5 we have demonstrated the impact of overexpressing the membrane-bound transcript variant on arborization pattern of the sensory axons.

Some analysis of the distance over which the "diffusible" form of Sema7a might act (many secreted ligands are not in fact all that diffusible), or

Comments: We have reported this in lines 311-317 and in Figure 4F,G.

Some live-imaging of axons before they reach the target (predicted to be the same in control and mutants) and then within the target (predicted to be different).

Comments: We have accepted this useful suggestion. We demonstrate the dynamics of the sensory arbors that are attracted to an ectopic Sema7Asec source in lines 325-332, Figure 4I,J; Figure 4—figure supplement 2A, and Videos 13-16.

Clearly, although the gain-of-function studies show that Sema7a can act at a distance, other cues are sufficient. Although the lack of a phenotype could be due to compensation, it is also possible that Sema7a does not actually act in a diffusible manner within its natural context. Overall, the data support the authors' carefully worded conclusions. While certain ideas are put forward as possibilities, the authors recognize that more work is needed. The main shortcoming is that the study does not actually distinguish between the effects of the two forms of Sema7a, which are predicted but not actually shown to be either diffusible or membrane linked (the membrane linkage can be cleaved). Although the study starts by presenting the splice forms, there is no description of when and where each splice form is transcribed.

Comments: We have utilized the HCR™ RNA-FISH Technology to generate transcript specific probes. To generate transcript-specific HCR probes to distinctly detect the sema7aGPI (NM_001328508) and the sema7asec (NM_001114885) transcripts, Molecular Instruments could design only 11 probes against the sema7aGPI transcript and only one probe against the sema7asec transcript (personal correspondence with Mike Liu, PhD, Head of Operations and Product Development Lead Molecular Instruments, Inc.). The HCR probe against the sema7aGPI transcript showed a very faint signal. Unfortunately, the HCR probe against the sema7asec transcript failed to detect the presence of any transcript. For robust detection of transcripts, the protocol demands a minimum of 20 probes. We believe that the very low number of probes against our transcripts is the primary reason for the absence of a signal.

We therefore utilized fluorescence-activated cell sorting (FACS) to capture the labeled hair cells and isolated total RNA to perform RT-PCR using transcript specific DNA oligonucleotide primers. We identified the presence of both the secreted and the membrane-bound transcripts at four-days-old neuromasts (lines 80-84, Figure 1B-D).

Additionally, since the mutants are predicted to disrupt both forms, it is a bit difficult to disentangle the synaptic phenotype from the earlier changes in circuit topology - perhaps the change at the level of the synapse is secondary to the change in topology.

Comments: Thank you for the insightful suggestion. We have analyzed the relationship between the sensory arbor network topology and the distribution of postsynaptic structures (lines 384-403, Figure 6G-I). We identified that the distribution of the postsynaptic aggregates is closely associated with the topological attributes of the sensory circuit. We further clarify the potential origin of disrupted synaptic assemblies in sema7a-/- mutants in lines 380-382 and lines 417-420.

Further, the authors do not provide any data supporting the idea that the membrane bound form of Sema7a acts only locally. Without these kinds of data, the authors are unable to attribute activities to either form.

Comments: We have accepted this useful suggestion and have prepared the Figure 5 with the necessary details.

The main impact on the field will be the nature of the analysis. The field of axon guidance benefits from this kind of robust quantification of growing axon trajectories, versus their ability to actually reach a target. This study highlights the value of more careful analysis and as a result, makes the point that circuit assembly is not just a matter of painting out paths using chemoattractants and repellants, but is also about how axons respond to local cues. The study also points to the likely importance of alternative splice forms and to the complex functions that can be achieved using different forms of the same ligand.

Reviewer #4 (Public Review):

Summary:

The work by Dasgupta et al identifies Sema7a as a novel guidance molecule in hair cell sensory systems. The authors use the both genetic and imaging power of the zebrafish lateralline system for their research. Based on expression data and immunohistochemistry experiments, the authors demonstrate that Sema7a is present in lateral line hair cells. The authors then examine a sema7a mutant. In this mutant, Sema7a proteins levels are nearly eliminated. Importantly, the authors show that when Sema7a is absent, afferent terminals show aberrant projections and fewer contacts with hair cells. Lastly the authors show that ectopic expression of the secreted form of Sema7a is sufficient to recruit aberrant terminals to non-hair cell targets. The sema7a innervation defects are well quantified. Overall, the paper is extremely well written and easy to follow.

Strengths:

(1) The axon guidance phenotypes in sema7a mutants are novel, striking and thoroughly quantified.

(2) By combining both loss of function sema7a mutants and ectopic expression of the secreted form of Sema7a the authors demonstrate the Sema7a is both necessary and sufficient to guide sensory axons

Weaknesses:

(1) Control. There should be an uninjected heatshock control to ensure that heatshock itself does not cause sensory afferents to form aberrant arbors. This control would help support the hypothesis that exogenously expressed Sema7a (via a heatshock driven promoter) is sufficient to attract afferent arbors.

Comments: Thank you for the suggestion. We have added the uninjected heatshock control experiment in Figure 5 and described experimental details in the text, lines 343-345.

(2) Synapse labeling. The numbers obtained for postsynaptic labeling in controls do not match up with the published literature - they are quite low. Although there are clear differences in postsynaptic counts between sema7a mutants and controls, it is worrying that the numbers are so low in controls. In addition, the authors do not stain for complete synapses (pre- and post-synapses together). This staining is critical to understand how Sema7a impacts synapse formation.

Comments: Thank you for raising this issue. We believe the low average numbers of the postsynaptic punctae in control neuromasts arise from lack of formation of postsynaptic aggregates beneath the immature hair cells, which are abundant in early stages of neuromast maturation. We have performed exhaustive analysis on the formation of pre- and postsynaptic structures and have identified how their distribution changes along neuromast development in control larvae. We have further analyzed how such distribution is perturbed in the sema7a-/- mutants. We do not think analyzing the complete synapse structure will add much to our understanding of how Sema7A influence synapse formation and maintenance.

(3) Hair cell counts. The authors need to provide quantification of hair cell counts per neuromast in mutant and control animals. If the counts are different, certain quantification may need to be normalized.

Comments: We have added the raw data with the hair cell counts in both control and sema7a-/- mutants across developmental stages. The homozygous sema7a-/- mutants have slightly less hair cells and we have normalized all our topological analyses by the corresponding hair cell numbers for each neuromast in each experiment (lines 669-675).

(4) Developmental delay. It is possible that loss of Sema7a simply delays development. The latest stage examined was 4 dpf, an age that is not quite mature in control animals. The authors could look at a later age, such as 6 dpf to see if the phenotypes persist or recover.

Comments: The homozygous sema7a-/- mutants are unviable and die at 6 dpf. We therefore restricted our analysis till 4 dpf. The association of the sensory arbors with the clustered hair cells gradually decreases as the neuromasts mature from 2 dpf to 4dpf in the sema7a-/- mutants (lines 174-176, Figure 2I). Moreover, in the sema7a-/- mutants the sensory axons throw long projections that keep getting farther away from the clustered hair cells as the neuromast matures from 2 dpf to 4 dpf (lines 166-168, Figure 2H; Figure 2—figure supplement 1K,L). These observations suggest that if the phenotypes in the sema7a-/- mutants were due to developmental delays, then we should have seen a recovery of disrupted arborization patterns over time. But instead, we observe a further deterioration of the arborization patterns and other architectural assemblies. These findings confirm that the observed phenotypes in the sema7a-/- mutants are not due to delayed development of the larvae, but a specific outcome for the loss of Sema7A protein.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Major concerns:

Issue 1: One of the most interesting conclusions in this manuscript is the function of the GPIanchored vs. secreted form of Sema7a in axon structure and synapse formation. In lines 357360 of the discussion (for example) the authors state that they have shown that the GPIanchored form of Sema7a is responsible for contact-mediated synapse formation while the secreted form functions as a chemoattractant for axon arbor structure. "We have discovered dual modes of Sema7A function in vivo: the chemoattractive diffusible form is sufficient to guide the sensory arbors toward their target, whereas the membrane-attached form likely participates in sculpting accurate neural circuitry to facilitate contact-mediated formation and maintenance of synapses." However, the data do not support this conclusion. Specifically, no analysis is done showing unique expression of either isoform in hair cells and no functional analysis is done to conclusively determine which isoform is important for either phenotype.

Comments: We have shown that both sema7a transcripts are expressed in the hair cells of four-day-old neuromasts (lines 78-84, Figure 1C,D). Ectopic expression of the sema7asec transcript variant robustly attracts the lateral-line sensory arbors toward itself, whereas ectopic expression of the sema7aGPI variant fails to impart sensory guidance from a distance, suggesting that the membrane-bound form likely participates in contact-mediated neural guidance. These experiments decisively show, for the first time in zebrafish, the dual modes of Sema7A function in vivo. However, we agree that the sema7aGPI transcript-specific knockout animal would be essential to conclusively prove that the membrane-attached form is primarily involved in forming accurate neural circuitry and contact-mediated formation and maintenance of synapses. Hence, we have very carefully stated in lines 427-428 that “the membrane-attached form likely participates in sculpting accurate neural circuitry to facilitate contact-mediated formation and maintenance of synapses”. We will follow up on this suggestion in our upcoming manuscript that will incorporate transcript-specific genetic ablations.

Though the authors present RT-PCR analysis of sema7a isoforms, it is not interpretable. The second reverse primer will also recognize the full-length transcript (from what I can gather) so it does not simply show the presence of the secreted form. Is there a unique 3'UTR for the short transcript that can be used? Additionally, for the GPI-anchored version can you use a forward primer that is not present in the short isoform? This would shed some light on the respective levels of both transcripts.

Comments: The C-termini of the two transcript variants are distinct and we have designed distinct primers that will selectively bind to each transcript (lines 503-511). Since, we have not performed quantitative polymerase chain reaction (qPCR), relative levels of each transcript are hard to determine.

Alternatively, and perhaps of more use, in situ hybridization using unique probes for each isoform would allow you to determine which are actually present in hair cells.

Comments: We have tried this approach and explained the point earlier (refer to lines 203212 of this response letter).

To decisively state that these isoforms have unique functions in axon terminal structure and synapse formation, other experiments are also essential. For example, RNA-mediated rescue analyses using both isoforms would tell you which can rescue the axonal structure and synapse size/number phenotypes. Overexpression of the GPI-anchored form, like the secreted form in Figure 4, would allow you to determine if only the secreted form can cause abnormal axon extension phenotypes. Expression of both forms in hair cells (using a myo6b promotor for example) would allow assessment of their role in presynapse formation.

Comments: We have ectopically expressed the sema7aGPI transcript variant near the sensory arbor network and observed that Sema7A-GPI fails to impart sensory axon guidance from a distance.

Thank you for suggesting the rescue experiments. We are in the process of generating CRISPR/Cas9-mediated transcript-specific knockout animals. We are currently preparing another manuscript that incorporates the above-mentioned rescue experiments to dissect the role of each transcript in regulating arbor topology and synapse formation.

For the overexpression experiments, expression of mKate alone (with and without heat shock) is also a critical control to include.

Comments: We have incorporated two control experiments: (1) larvae injected with hsp70:sema7asec-mKate2 plasmid that were not heat shocked and (2) Uninjected larvae that were heatshocked. We think these two controls are sufficient to demonstrate that the abnormal arborization patterns are not artifacts generated due to plasmid injection and heatshocking.

Issue 2: A second concern is the lack of data showing support cell and hair cell formation and function is unaffected. Analysis of support and hair cell number with loss of Sema7a as well as simple analyses of mechanotransduction (FM4-64) would help alleviate concerns that phenotypes are due to disrupted neuromast formation and basic hair cell function rather than a specific role for Sema7a in this process.

Comments: We have measured the hair cell numbers in both control and sema7a-/- mutants across developmental stages. We have added this to our submitted raw data.

We have utilized the styryl fluorophore FM4-64 to test the mechanotransduction function of the hair cells in sema7a-/- mutants. We have detailed our finding in lines 137141 and in Figure 2—figure supplement 1C,D.

Expression analysis of Sema7a receptors would also help strengthen the argument for a specific effect on lateral line afferent axons.

Comments: Thank you for this suggestion. Currently, we do not possess an RNA transcriptome dataset for the lateral line ganglion. This deficit limits a systematic screen for lateral-line sensory neuronal gene expressions either through antibody stains or via HCRmediated in situ techniques. In future we plan to develop an RNA transcriptome for the lateral-line ganglion and identify potential binding partners for Sema7A.

Issue 3: The manuscript could also be improved to include more detail in some areas and less in others. In general, each section has a fairly long lead up but lacks important experimental details that would help the reader interpret the data. For example:

Figure 1: What is the label for the lateral line axons? Is it a specific transgenic? The legend states that 3 asterisks indicate p<0.0001. What about the other asterisk combinations?

Comments: We have clarified these issues in lines 118-121 and in lines 906-907.

Figure 2: For the network analysis, are the traces for all axons that branch to innervate the neuromast?

Comments: Yes, we have traced the entire arbor containing all the axons that branched from the lateral line nerve and extended toward the clustered hair cells. The three-dimensional traces depict a skeletonized representation of the arbor network.

Can the tracing method distinguish individual axons?

Comments: No, our goal is to understand how the axon-collective behave around the clustered hair cells during development.

How do you know where an end is versus continued looping?

Comments: We have categorically defined the topological attributes in lines 187-191 and in Figure 3A.

Also, are all neuromasts similarly affected or is there a divergence based on which organ you are imaging? What neuromast was imaged in this and other figures?

Comments: Yes, all the neuromasts in the trunk and tail regions were affected similarly by the sema7a mutation. We did not observe any region-specific phenotypic outcome. We consistently imaged the trunk neuromasts, particularly the second, third, and fourth neuromasts.

Discussion: The short discussion failed to put these findings into context or to discuss how this unique topological arrangement of axon terminals impacts function.

Comments: We have added a new segment, lines 432-448, in the discussion section which mentions the potential role of the topological features in arranging the distribution pattern of the postsynaptic densities and thereby potentially influencing the network’s ability to gather sensory inputs through properly placed postsynaptic aggregates.

Can you speculate on how the looping structure may alter number of synaptic contacts per axon for instance? For this, it would be useful to know if normally the synapses form on loops versus bare terminals.

Comments: Thank you for this insightful suggestion. We have performed detailed analysis, as mentioned in lines 384-397, to characterize the distribution of the postsynaptic densities between the two topological attributes.

Does this looping facilitate single axons contacting more hair cells of the same polarity? Would that be beneficial?

Comments: Looping behaviors indeed facilitate the contact between the axons and the hair cells. As we have observed, the primary topological attribute that the sensory arbor network underneath the clustered hair cells adopts is a loop. The bare terminals are predominantly projected transverse to the clustered hair cells and lack contact with them. Whether a single axon, being part of a loop, preferentially contacts hair cells of same polarity is yet to be determined. We can address this question by mosaic labeling a single axon in the arbor network and determine its association with the hair cells. We intend to do these experiments in our upcoming manuscript.

Minor concerns:

(1) For the stacked charts quantifying topological features, I found interpreting them challenging. Is it possible to put these into overlapping histograms or line graphs to better compare wild type to mutant directly?

Comments: Thank you for your suggestion. We tried several ways to represent our data and found that the stacked charts optimally signify our analysis and depict the characteristic phenological differences between the control and the sema7a-/- mutants.

(2) There are numerous strong statements throughout not directly supported by the data, e.g. lines 110-113; 206-208; 357-360 and others. These should be tempered.

Comments: For lines 110-113, we have updated this section with new experiments and the new segment is represented in lines 115-126.

For lines 206-208, we have updated the statement to “This result suggests that the stereotypical circuit topology observed in the mature organ may emerge through transition of individual arbors from forming bare terminals to forming closed loops encircling topological holes” in lines 225-227.

Reviewer #2 (Recommendations For The Authors):

The authors should be careful about making any assumptions which form of sema7a is active in NMs. Their RT-PCR demonstrates presence of both isoforms in a whole animal; however, whether they are similarly present in HCs is not investigated here.

Comments: We have addressed this concern and have updated the manuscript with new experiments, detailed in lines 78-84.

Also, there is an issue of translation and trafficking to the membrane with subsequent secretion. An important experiment that would address this question is expressing two sema7a isoforms in mutant HCs and asking whether this can suppress the mutant phenotype.

Comments: Thank you for suggesting the rescue experiments. We are in the process of generating CRISPR/Cas9-mediated transcript-specific knockout animals. We are currently preparing another manuscript that incorporates the above-mentioned rescue experiments to dissect the role of each transcript in regulating arbor topology and synapse formation.

Presumably, sema7a is trafficked to the membrane during HC maturation. This is consistent with the authors' observation that sema7a localization is changing as NM mature. However, actin-sema7a co-labeling does not actually show whether sema7a is on the membrane. Labeling HCs with a membrane marker (transgene) would be much more convincing. Alternatively, can the authors show sema7a localization actually correlates with the presence of sensory axon terminals? They already have immunos that label both. Thus, this should be pretty straightforward.

Comments: Thank you for these suggestions. We have addressed these issues in lines 112114, and in lines 119-126.

Figure 2 should have a control panel, so the reduced sema7a staining can be compared to the control side-by-side.

Comments: We have depicted Sema7A staining in control neuromasts in multiple images, including Figure 1E, Figure 1H, and in Figure 2—figure supplement 1B. We have kept the control panel in the supplementary figure due to space restrictions in Figure 2.

Arborization topology: While I appreciate the very careful characterization of the topology for wild-type and mutant NMs, I think it would be much more informative to mark individual axons and then analyze their topology. The main reason is that the authors cannot really distinguish whether some aspects of topology they describe are really due to the densely packed overlapping terminals of multiple axons or these are really characteristic, higher order organization of individual axons. Because of this, they cannot be certain what is really happening with sema7a mutant terminals. Related to the point above. While it is clear that the overall topology is abnormal in the mutant, the authors should be careful in concluding that sema7a regulates specific aspects of it. The overall structure is probably highly interconnected perturbing one parameter would likely affect all the others.

Comments: Thank you for this comment. In a previous eLife publication from our laboratory, we utilized the serial blockface scanning electron micrograph (SBFSEM) technique to characterize the connectome of the neuromast microcircuit where patterns of innervation of all the individual axons can be delineated in five-days-old larvae (reference number 8). However, the collective behavior of all the sensory axons that build the innervation network remained enigmatic, especially in a living animal during development. In this paper we addressed how the sensory axon-collective behave around the clustered hair cells and build the innervation network in living animals during diverse developmental stages. Our analyses have not only identified how the axon-collective associates itself with the hair cell cluster as the organ matures, but also discovered distinct topological features in the arbor network that emerges during organ maturation, which may influence assembly of postsynaptic aggregates (lines 384-403, Figure 6G-I). We believe that our quantitative approach to capture collective axonal behaviors and their topological attributes during circuit formation have highlighted the importance of understanding network assembly during sensory organ development.

Experiments with the secreted sema7a isoform would be much more informative if they were compared/contrasted to the GPI anchored isoform.

Comments: We added a new section, lines 338-351, and a new Figure 5 to address this issue.

The phenotype of ectopic projections in sema7a overexpression experiments is pretty dramatic, especially given the fact that these were performed in wild-type animals. Does this mean that the phenotype would be even more dramatic in sema7a mutants, as they have more bare axon terminals according to the authors' analysis. Have the authors attempted this type of experiments?

Comments: That is an interesting suggestion. We have not tested that yet. Our guess is that in the sema7a-/- mutants, the abundant bare terminals will be far more sensitive to an ectopic source of Sema7A. But even in the sema7a-/- mutants, other chemotropic cues are still functional, which may impart certain restrictions on how many bare terminals are allowed to leave the neuromast region.

Reviewer #3 (Recommendations For The Authors):

(1) No raw data are shown, such that it is difficult to assess variability across animals or within animals, just the overall trends within the whole dataset. Raw data need to be shown for every measurement, at least in supplemental figures. It would also be useful to reliably show control next to mutant in the same plot, as it is a bit hard to compare across panels, which occurs in several figures.

Comments: We have uploaded all the raw data related to each experiment.

(2) Given the focus on the two possible forms of Sema7a, the authors should use HCR or another form of reliable in situ hybridization to show the spatiotemporal pattern of expression of each isoform.

Comments: We have utilized the HCR™ RNA-FISH Technology to generate transcript specific probes. To generate transcript-specific HCR probes to distinctly detect the sema7aGPI (NM_001328508) and the sema7asec (NM_001114885) transcripts, Molecular Instruments could design only 11 probes against the sema7aGPI transcript and only one probe against the sema7asec transcript (personal correspondence with Mike Liu, PhD, Head of Operations and Product Development Lead Molecular Instruments, Inc.). The HCR probe against the sema7aGPI transcript showed a very faint signal. Unfortunately, the HCR probe against the sema7asec transcript failed to detect the presence of any transcript. For robust detection of transcripts, the protocol demands a minimum of 20 probes. We believe that the very low number of probes against our transcripts is the primary reason for the lack of a signal.

(3) The authors should explain the criteria used to select the 22 embryos used to analyze the effects of expressing diffusible Sema7a.

Comments: We have explained this in lines 291-292. We identified 22 mosaic sema7asecmKate2 integration events, in which a single mosaic ectopic integration had occurred near the network of sensory arbors, from a total of almost 100 integrations. We rejected events where the sema7asec-mKate2 integration occurred either farther away from the sensory arbor network or had happened in multiple neighboring cells.

(4) Although arbors were imaged in live embryos, time is never presented as a variable, so I cannot tell whether axon topology was changing as the images were collected. This needs to be clarified.

Comments: We imaged the trunk neuromasts of both control and sema7a-/- mutant live zebrsfish larvae at 2, 3, and 4 dpf. We imaged the control and the sema7a-/- mutants of each developmental stage in parallel, within a span of two hours, and repeated these experiments multiple times to gather almost a hundred larvae from each genotype. Even though the sensory arbor network is dynamic, we believe imaging both the genotypes in parallel and within a span of two hours, and averaging almost a hundred larvae from each genotype minimize the temporal variability observed in the arbor architecture.

(5) Ideally, the authors should use CRISPR/cas-9 to create a mutation in the C-terminus that would prevent production of the GPI-anchored form and not of the diffusible form. I understand if this is too much work to do in a short time, and would be satisfied with another experiment that could distinguish roles for at least one isoform more clearly. For instance, it would be interesting to see an analysis of how far an axon can be from a source to detect diffusible Sema7a (live imaging would be ideal for this) and then to show that the effect is different when the membrane bound form is expressed.

Comments: Thank you for this comment. We are currently working in generating transcript specific knockout animals.

We have added live timelapse video microscopy data in lines 330-337, Figure 4H-J, Figure 4—figure supplement 2, Video15,16.

We have added a new segment analyzing the membrane-bound transcript variant in lines 338-351.

Reviewer #4 (Recommendations For The Authors):

Feedback to authors

Overall, this is a very important and novel study. Currently the manuscript does need revision.

Major concerns:

(1) Controls. For the ectoptic expression of Sema7a, injection of a construct expressing Sema7a under a heatshock promoter is used to drive ectopic expression. No heatshock (injected) animal are used as a control. In many systems heatshock can impact neuron morphology. And heatshock proteins are required for normal neurite and synapse formation. Please examine sensory axons in uninjected wildtype animals with heatshock.

Comments: We have added this control experiment in a new segment, explained in detail in lines 348-350 and Figure 5.

(2) Synapse staining - regarding Figure 5 and related supplement

Understanding whether guidance defects ultimately impact synapse formation is an important aspect of this paper. Therefore, is necessary to have accurate measurements of the number of complete synapses, and the overall numbers of pre- and postsynaptic components. Currently the data plotted in Figure 5 is extensive, but the way the data is laid out, the relevant comparisons are challenging to make. Perhaps include this quantification in the supplement, and move the data from the supplement to the main figure? The quantifications in the supplement are easier to follow and easier to compare between genotypes.

Comments: We have performed exhaustive analysis on the formation of pre- and postsynaptic structures and have identified how their distribution changes along neuromast development in control larvae. We have further analyzed how such distribution is perturbed in the sema7a-/- mutants. We believe that showing only the average numbers will not reveal the changes in the distribution of the synaptic structures during development and across genotypes.

Looking at the data itself, there seems to be some discrepancies with the synaptic counts compared to published work. While the CTBP numbers seem in order, the Maguk numbers do not. In both mutant and control there are many hair cells without any Maguk puncta/aggregates-leading to 0.75-1 postsynapses per hair cell (Figure 5 supplement H-I). Typically, the numbers should be more comparable to what was obtained for CTBP, 3-4 puncta per cells (Figure 5 supplement B-C), especially by 3-4 dpf. 3-4 CTPB or Maguk puncta per cell is based on previously published immunostaining and EM work.

The Maguk immunostaining, especially at early stages (2-3 dpf) is challenging. To compound a challenging immunostain, around 2019 Neuromab began to outsource the purification of their Maguk antibody. After this outsourcing our lab was no longer able to get reliable label with the Maguk antibody from Neuromab.

Millipore sells the same monoclonal antibody and it works well: https://www.emdmillipore.com/US/en/product/Anti-pan-MAGUK-Antibody-clone-K2886,MM_NF-MABN72

I would recommend this source.

Comments: Thank you for suggesting the new MAGUK antibody. We have utilized this new MAGUK antibody from Millipore and added a new segment in lines 389-408. In future publication we will utilize this antibody to capture the postsynaptic densities in the sensory arbors.

The discrepancies in the postsynaptic punctae number in our control larvae may arise due to the reliability of the Neuromab MAGUK antibody. We have utilized this same antibody to stain the sema7a-/- mutants and have observed a significant decrease in MAGUK punctae number and area. On grounds of keeping parity between the control and the sema7a-/- mutants, we have decided to keep our experimental results in the manuscript.

In addition to a more accurate Maguk label, a combined pre- and post-synaptic label is essential to understand whether synapses pair properly in the sema7a mutants. This can be accomplished using subtype specific antibodies using goat anti-mouse IgG1/Maguk and goat anti-mouse IgG2a/CTBP secondaries.

Comments: Thank you for suggesting this. We are preparing another manuscript in which we will utilize this technique along with other suggestions to tease apart the role of distinct transcript variants in regulating neural guidance and synapse formation.

(3) Does sema7a lesion impact the number of hair cells per neuromast? If hair cell numbers are reduced several of the quantifications could be impacted.

Comments: We have added the raw data with the hair cell counts in both control and sema7a-/- mutants across developmental stages. The homozygous sema7a-/- mutants have slightly less hair cells and we have normalized all our topological analyses by the corresponding hair cell numbers for each neuromast in each experiment (lines 669-675).

(4) Could innervation just be developmentally delayed in sema7a mutants? At 4 dpf the sensory system is just starting to come online and could still be in the process of refinement. Did you look at slightly older ages, after the sensory system is functional behaviorally, for example, 6 dpf? Do the cores phenotypes (synapse defects and excess arbors) persist at 6 dpf in the sema7a mutants?

Comments: The homozygous sema7a-/- mutants are unviable and start to die at 6 dpf. We therefore restricted our analysis until 4 dpf. The association of the sensory arbors with the clustered hair cells gradually decreases as the neuromasts mature from 2 dpf to 4dpf in the sema7a-/- mutants (lines 174-176, Figure 2I). Moreover, in the sema7a-/- mutants the sensory axons throw long projections that keep getting farther away from the clustered hair cells as the neuromast matures from 2 dpf to 4 dpf (lines 166-168, Figure 2H; Figure 2—figure supplement 1K,L). These observations suggests that if the phenotypes in the sema7a-/- mutants were due to developmental delays, then we should have seen a recovery of disrupted arborization patterns over time. But instead, we observe a further deterioration of the arborization patterns and other architectural assemblies. These findings confirm that the observed phenotypes in the sema7a-/- mutants are not due to delayed development of the larvae, but a specific outcome for the loss of Sema7A protein.

Minor comments to address:

Results

Page 4 lines 89-91. For the readers, explain why you examined levels in Sema7a in rostral and caudal hair cells. Also, this sentence is, in general, a little bit misleading-initially reading that there is no difference in Sema7a at 1.5-4 dpf.

Comments: In lines 44-48, we explain that the hair cells in the neuromast contain mechanoreceptive hair cells of opposing polarities that help them detect water currents from opposing directions. In lines 93-106, we tested whether the Sema7A level varies between the two polarities. We observed that the Sema7A level is similar between the two polarities of hair cells, but the average Sema7A intensity increases significantly over the developmental period of 2 dpf to 4 dpf in both rostrally and caudally polarized hair cells.

Page 10-11 Lines 263-270. What was the frequency of these 2 outcomes- out of the 22 cases with ectopic expression?

Comments: We have explained this in lines 291-292. We identified 22 mosaic sema7asecmKate2 integration events, in which a single mosaic ectopic integration had occurred near the network of sensory arbors, from a total of almost 100 integrations. We rejected events where the sema7asec-mKate2 integration occurred either farther away from the sensory arbor network or had happened in multiple neighboring cells.

Discussion

Page 14 Lines 359-360. There is not enough evidence provided in this work to suggest that the membrane attached form of Sema7a is playing a role. Both the secreted and membrane form are gone in the sema7a mutants. If the membrane attached form was specifically lesioned, and resulted in a phenotype, then there would be sufficient evidence. Currently there is strong evidence for a distinct role for the secreted form. Although the authors qualify the outlined statement with the word 'likely', stating this possibility in the discussion take-home is misleading.

Comments: In future we will utilize the CRISPR/Cas9 technique to target the unique Cterminal domain of the GPI-anchored sema7a transcript variant. We believe that this will only perturb the formation of the full-length Sema7A protein and help us differentiate between the roles of the membrane-bound Sema7AGPI molecule and the secreted Sema7Asec in sensory arborization and synaptic assembly.

It might be interesting in either the intro or discussion to reference the role Sema3F in axon guidance in the mouse auditory epithelium. https://elifesciences.org/articles/07830

Comments: We have added this reference in lines 61-64.

Figures

Please indicate on one of your Figures where the mutation is (roughly) in the sema7a mutant (in addition to stating it in the results).

Comments: We have added this information in Figure 2—figure supplement 1A.

Either state or indicate in a Figure where the epitope used to make the Sema7a antibody-to show that the antibody is predicted to recognize both isoforms.

Comments: We have stated the details of the epitope in lines 528-529.

Figure 2-S1 what is the scale in panel A, is it different between mutant and wildtype?

Comments: We have updated the images. New images are depicted in Figure 2—figure supplement 1A.

Methods

What were the methods used to quantify synapse number and area?

Comments: We have added a new section in lines 702-708 to explain the measurement techniques.

eLife assessment

Dasgupta and colleagues make a valuable contribution to the understanding how the guidance factor Sema7a promotes connections between mechanosensory hair cells and afferent neurons of the zebrafish lateral line system. The authors provide solid evidence that loss of Sema7a function results in fewer contacts between hair cells and afferents through comprehensive quantitative analysis. Additional work is needed to distinguish the effects of different isoforms of Sema7a to determine whether there are specific roles of secreted and membrane bound forms.

Reviewer #1 (Public Review):

Dasguta et al. have dissected the role of Sema7a in fine tuning of a sensory microcircuit in the posterior lateral line organ of zebrafish. They attempt to also outline the different roles of a secreted verses membrane-bound form of Sema7a in this process. Using genetic perturbations and axonal network analysis, the authors show that loss of both Sema7a isoforms causes abnormal axon terminal structure with more bare terminals and fewer loops in contact with presynaptic sensory hair cells. Further, they show that loss of Sema7a causes decreased number and size of both the pre- and post-synapse. Finally, they show that overexpression of the secreted form of Sema7a specifically can elicit axon terminal outgrowth to an ectopic Sema7a expressing cell. Together, the analysis of Sema7a loss of function and overexpression on axon arbor structure is fairly thorough and revealed a novel role for Sema7a in axon terminal structure. However, the connection between different isoforms of Sema7a and the axon arborization needs to be substantiated. Furthermore, the effect of loss of Sema7a on the presynaptic cell is not ruled out as a contributing factor to the synaptic and axon structure phenotypes. These issues weaken the claims made by the authors including the statement that they have identified dual roles for the GPI-anchored verses secreted forms of Sema7a on synapse formation and as a chemoattractant for axon arborization respectively.

Reviewer #2 (Public Review):

In this work, Dasgupta et al. investigates the role of Sema7a in the formation of peripheral sensory circuit in the lateral line system of zebrafish. They show that Sema7a protein is present during neuromast maturation and localized, in part, to the base of hair cells (HCs). This would be consistent with pre-synaptic Sema7a mediating formation and/or stabilization of the synapse. They use sema7a loss-of-function strain to show that lateral line sensory terminals display abnormal arborization. They provide highly quantitative analysis of the lateral line terminal arborization to show that a number of specific topological parameters are affected in mutants. Next, they ectopically express a secreted form of Sema7a to show that lateral line terminals can be ectopically attracted to the source. Finally, they also demonstrate that the synaptic assembly is impaired in the sema7a mutant. Overall, the data are of high quality and properly controlled. The availability of Sema7a antibody is a big plus, as it allows to address the endogenous protein localization as well to show the signal absence in the sema7a mutant. The quantification of the arbor topology should be useful to people in the field who are looking at the lateral line as well as other axonal terminals. I think some results are overinterpreted though. The authors state: "Our findings demonstrate that Sema7A functions both as a juxtracrine and as a secreted cue to pattern neural circuitry during sensory organ development." However, they have not actually demonstrated which isoform functions in HCs (also see comments below). In addition, they have to be careful in interpreting their topology analysis, as they cannot separate individual axons. Thus, such analysis can generate artifacts. They can perform additional experiments to address these issues or adjust their interpretations.

Reviewer #3 (Public Review):

The data reported here demonstrate that Sema7a defines the local behavior of growing axons in the developing zebrafish lateral line. The analysis is sophisticated and convincingly demonstrates effects on axon growth and synapse architecture. Collectively, the findings point to the idea that the diffusible form of sema7a may influence how axons grow within the neuromast and that the GPI-linked form of sema7a may subsequently impact how synapses form, though additional work is needed to strongly link each form to its' proposed effect on circuit assembly.

Comments on revised submission:

The revised manuscript is significantly improved. The authors comprehensively and appropriately addressed most of the reviewers' concerns. In particular, they added evidence that hair cells express both Sema7A isoforms, showed that membrane bound Sema7A does not have long range effects on guidance, demonstrated how axons behave close to ectopic Sema7A, and analyzed other features of the hair cells that revealed no strong phenotypes. The authors also softened the language in many, but not all places. Overall, I am satisfied with the study as a whole.

Reviewer #4 (Public Review):

This study provides direct evidence showing that Sema7a plays a role in the axon growth during the formation of peripheral sensory circuits in the lateral-line system of zebrafish. This is a valuable finding because the molecules for axon growth in hair-cell sensory systems are not well understood. The majority of the experimental evidence is convincing, and the analysis is rigorous. The evidence supporting Sema7a's juxtracrine vs. secreted role and involvement in synapse formation in hair cells is less conclusive. The study will be of interest to cell, molecular and developmental biologists, and sensory neuroscientists.

eLife assessment

This study makes a valuable contribution to spatial transcriptomics by rigorously benchmarking cell-type deconvolution methods, assessing their performance across diverse datasets with a focus on biologically relevant, previously unconsidered aspects. The authors demonstrate the strengths of RCTD, cell2location, and SpatialDWLS for their performance, while also revealing the limitations of many methods when compared to simpler baselines. By implementing a full Nextflow pipeline, Docker containers, and a rigorous assessment of the simulator, this work offers robust insights that elevate the standards for future evaluations and provides a resource for those seeking to improve or develop new deconvolution methods. The thorough comparison and analysis of methods, coupled with a strong emphasis on reproducibility, provide solid support for the findings.

Author Response

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

Reviewer #1 (Recommendation for the authors)

I only have one comment for improvement of this study and it has to do with the comparison of simulators that they conducted. There are many other simulators around now, including scDesign3, spaSim, SPIDER, SRTSIM, etc. Are any of those methods worth including in the comparison?

Indeed, many of the mentioned simulators did not exist when we initially developed synthspot, and upon closer examination, they are not directly comparable to our tool.

• scDesign3: The runtime of scDesign3 is quite long as a result of its generative model. The example provided in its tutorial only simulates 183 genes and takes over seven minutes when using four cores on a system with Intel Xeon E5-2640 CPUs running at 2.5GHz. In a small downsampling analysis, we simulated 10, 50, 100, and 150 genes with scDesign3 and observed runtimes of 30, 130, 245, and 360 seconds, respectively. This seems to indicate a linear relationship between the number of genes and the runtime, therefore rendering it unsuitable for simulating whole-transcriptome datasets for deconvolution.

• spaSim: spaSim focuses on modelling cell locations in different tissue structures but does not provide gene expression data. It is designed for testing cell colocalization capabilities rather than simulating gene expression.

• SPIDER: Although SPIDER appears to have some overlap with our work, it seems to be in the early stages of development. The GitHub repository contains only two scripts without any documentation, and the preprint does not provide instructions on how to use the tool.

• SRTSim: SRTSim explicitly states in its publication that it is not suitable for evaluating cell type deconvolution, as its focus is on simulating gene expression data without modelling cell type composition.

• scMultiSim: scMultiSim, like scDesign3, is limited in its capability to model the entire transcriptome.

Nonetheless, the inherent modularity of our Nextflow framework makes it possible for users to simply run the deconvolution methods on data that has been simulated by other simulators if need be.

Additionally, we have added the following rationale for why we developed synthspot in “Synthspot allows simulation of artificial tissue patterns”:

“On the other hand, general-purpose simulators are typically more focused on other inference tasks, such as spatial clustering and cell-cell communication, and are unsuitable for deconvolution. For instance, generative models and kinetic models like those of scDesign3 and scMultiSim are computationally intensive and unable to model entire transcriptomes. SRTSim focuses on modeling gene expression trends and does not explicitly model tissue composition, while spaSim only models tissue composition without gene expression.”

The other aspect of the simulation comparison that I'm missing is some kind of spatial metric. There are metrics about feature correlation, sample-sample correlation, library size, etc. But, what about spatial correlation (e.g., Moran's I or similar). Perhaps comparing the distribution of Moran's I across genes in a simulated and real dataset would be a good first start.

We would like to clarify that synthspot does not actually simulate the spatial location of spots, but synthetic regions where spots from the same region share similar compositions. Hence, incorporating a spatial metric in the comparison is not feasible. However, as RCTD is the only method that explicitly uses spot locations in its model (Supplementary Table 2, "Location information"), we believe that generating synthetic datasets with actual coordinates would not significantly impact the conclusions of the study.

Reviewer #2 (Public Review)

On the other hand, the authors state that in silver standard datasets one half of the scRNA-seq data was used for simulation and the other half was used as a reference for the algorithms, but the method of splitting the data, i.e., at random or proportionally by cell type, was not specified.

The data was split proportionally by cell type. To clarify this, we have included an additional sentence in the main text under the first paragraph of “Cell2location and RCTD perform well in synthetic data”, as well as in Figure S2.

Reviewer #2 (Recommendation for the authors)

Figure legends in Figures 3, 4 and across most Supplementary material are almost illegible. Please consider increasing font size for better readability.

Thank you for bringing this to our attention. The font size has been increased for all main and supplementary figures. Additionally, the supplementary figures have also been exported in higher resolution.

Supplementary Notes Figure 2c reads "... total count per sampled multiplied by..."

This has been adapted, as well as the captions of Supplementary Notes Figure 3c and 4c which had the same typo.

Review #3 (Public review)

The simulation setup has a significant weakness in the selection of reference single-cell RNAseq datasets used for generating synthetic spots. It is unclear why a mix of mouse and human scRNA-seq datasets were chosen, as this does not reflect a realistic biological scenario. This could call into question the findings of the "detecting rare cell types remains challenging even for top-performing methods" section of the paper, as the true "rare cell types" would not be as distinct as human skin cells in a mouse brain setting as simulated here.

We appreciate the reviewer’s concern and would like to clarify that within one simulated dataset, we never mix mouse and human scRNA-seq data together. The synthetic spots generated for the silver standards are always sampled from a single scRNA-seq or snRNA-seq dataset. Specifically, for each of the seven public scRNA-seq datasets, we generate synthetic datasets with one of nine abundance patterns, resulting in a total of 63 synthetic datasets. These abundance patterns only affect the sampling priors that are used—the spots are still created with combinations of cells sampled from the same dataset.

Furthermore, it is unclear why the authors developed Synthspot when other similar frameworks, such as SRTsim, exist. Have the authors explored other simulation frameworks?

While there are other simulation frameworks available now, synthspot was designed to specifically address the requirements of our study, offering unique capabilities that make it suitable for deconvolution evaluation. Moreover, many of the simulators did not exist when we initially developed our tool. We have added the following rationale for why we developed synthspot in “Synthspot allows simulation of artificial tissue patterns”:

“On the other hand, general-purpose simulators are typically more focused on other inference tasks, such as spatial clustering and cell-cell communication, and are unsuitable for deconvolution. For instance, generative models and kinetic models like those of scDesign3 and scMultiSim are computationally intensive and unable to model entire transcriptomes. SRTSim focuses on modeling gene expression trends and does not explicitly model tissue composition, while spaSim only models tissue composition without gene expression.”

In our response to Reviewer 1 copied below, we also outline specific reasons why other simulators were not suitable for our benchmark:

• scDesign3: The runtime of scDesign3 is quite long as a result of its generative model. The example provided in its tutorial only simulates 183 genes and takes over seven minutes when using four cores on a system with Intel Xeon E5-2640 CPUs running at 2.5GHz. In a small downsampling analysis, we simulated 10, 50, 100, and 150 genes with scDesign3 and observed runtimes of 30, 130, 245, and 360 seconds, respectively. This seems to indicate a linear relationship between the number of genes and the runtime, therefore rendering it unsuitable for simulating whole-transcriptome datasets for deconvolution.

• spaSim: spaSim focuses on modelling cell locations in different tissue structures but does not provide gene expression data. It is designed for testing cell colocalization capabilities rather than simulating gene expression.

• SPIDER: Although SPIDER appears to have some overlap with our work, it seems to be in the early stages of development. The GitHub repository contains only two scripts without any documentation, and the preprint does not provide instructions on how to use the tool.

• SRTSim: SRTSim explicitly states in its publication that it is not suitable for evaluating cell type deconvolution, as its focus is on simulating gene expression data without modelling cell type composition.

• scMultiSim: scMultiSim, like scDesign3, is limited in its capability to model the entire transcriptome.

Finally, we would have appreciated the inclusion of tissue samples with more complex structures, such as those from tumors, where there may be more intricate mixing between cell types and spot types.

We acknowledge the reviewer's suggestion and have incorporated a melanoma dataset from Karras et al. (2022) in response to this suggestion. This study profiled melanoma tumors by using both scRNA-seq and spatial technologies. The scRNA-seq consists of eight immune cell types and seven melanoma cell states. We have included this study as an additional silver standard and case study, the latter of which is presented in a separate section following the liver analysis (and a corresponding section in Methods).

We found that method performances on synthetic datasets generated from this melanoma dataset follow previous trends (Figure S3-S5). However, the inclusion of the case study led to the following changes in the overall rankings: cell2location and RCTD are now tied for first place (previously RCTD ranked first), and Seurat and SPOTlight have swapped places. Despite these changes, the core messages and conclusions of our paper remain unchanged. All relevant figures (Figures 1a, 2, 3a, 4a, 6b, 7a, S3-S6, S9) have been updated to incorporate these new analyses and results.

Review #3 (Recommendation for the authors)

To maintain consistency in the results, it is recommended to exclude the human scRNAseq set when generating synthetic spots. Furthermore, addressing the other significant weaknesses mentioned earlier would be beneficial.

Please refer to our response to the public review where we address the same remark.

It is essential to differentiate this work from previous benchmarking and simulation frameworks.

In addition to the rationale on why we developed our own framework (see response to the public review), we have included the following text in the discussion that highlights our versatile approach when using a real spatial dataset for evaluation:

“In the case studies, we demonstrated two approaches for evaluating deconvolution methods in datasets without an absolute ground truth. These approaches include using proportions derived from another sequencing or spatial technology as a proxy, and leveraging spot annotations, e.g., zonation or blood vessel annotations, that typically have already been generated for a separate analysis.”

Furthermore, we conducted an extra analysis in the liver case study, generating synthetic datasets with one experimental protocol and using the remaining two as separate references (Figure S13). This further illustrates the usefulness of our simulation framework, which we mentioned by appending this sentence in the discussion:

“As in our silver standards, users can select the abundance pattern most resembling the real tissue to generate the synthetic spatial dataset, as we have also demonstrated in the liver case study.”

Reviewer #1 (Public Review):

Cell type deconvolution is one of the early and critical steps in the analysis and integration of spatial omic and single cell gene expression datasets, and there are already many approaches proposed for the analysis. Sang-aram et al. provide an up-to-date benchmark of computational methods for cell type deconvolution.

In doing so, they provide some (perhaps subtle) additional elements that I would say are above the average for a benchmarking study: i) a full Nextflow pipeline to reproduce their analyses; ii) methods implemented in Docker containers (which can be used by others to run their datasets); iii) a fairly decent assessment of their simulator compared to other spatial omics simulators. A key aspect of their results is that they are generally very concordant between real and synthetic datasets. And, it is important that they authors include an appropriate "simpler" baseline method to compare against and surprisingly, several methods performed below this baseline. Overall, this study also has the potential to also set the standard of benchmarks higher, because of these mentioned elements.

The only weakness of this study that I can readily see is that this is a very active area of research and we may see other types of data start to dominate (CosMx, Xenium) and new computational approaches will surely arrive. The Nextflow pipeline will make the the prospect of including new reference datasets and new computational methods easier.

Reviewer #2 (Public Review):

In this manuscript Sangaram et al provide a systematic methodology and pipeline for benchmarking cell type deconvolution algorithms for spatial transcriptomic data analysis in a reproducible manner. They developed a tissue pattern simulator that starts from single-cell RNA-seq data to create silver standards and used spatial aggregation strategies from real in situ-based spatial technologies to obtain gold standards. By using several established metrics combined with different deconvolution challenges they systematically scored and ranked 12 different methods and assessed both functional and usability criteria. Altogether, they present a reusable and extendable platform and reach very similar conclusions to other deconvolution benchmarking paper, including that RCTD, SpatialDWLS and Cell2location typically provide the best results. Major strengths of the simulation engine include the ability to downsample and recapitulate several cell and tissue organization patterns.

More specifically, the authors of this study sought to construct a methodology for benchmarking cell type deconvolution algorithms for spatial transcriptomic data analysis in a reproducible manner. The authors leveraged publicly available scRNA-seq, seqFISH, and STARMap datasets to create synthetic spatial datasets modeled after that of the Visium platform. It should be noted that the underlying experimental techniques of seqFISH and STARMap (in situ hybridization) do not parallel that of Visium (sequencing), which could potentially bias simulated data. Furthermore, to generate the ground truth datasets cells and their corresponding count matrix are represented by simple centroids. Although this simplifies the analysis it might not necessarily accurately reflect Visium spots where cells could lie on a boundary and affect deconvolution results.

The authors thoroughly and rigorously compare methods while addressing situational discrepancies in model performance, indicative of a strong analysis. The authors make a point to address both inter- and intra- dataset reference handling, which has a significant impact on performance, as the authors note in the text and conclusions. Indeed, supplying optimal reference data is - potentially most - important to achieve the best performance and hence it's important to understand that experimental design or sample matching is at least equally important to selecting the ideal deconvolution tool.

Similarly, the authors conclude that many methods are still outperformed by bulk deconvolution methods (e.g. Music or NNLS), however, it needs to be noted that these 'bulk' methods are also among the most sensitive when using an external (inter) dataset (S10), which likely resembles the more realistic scenario for most labs.

As the authors also discuss it's important to realize that deconvolution approaches are typically part of larger exploratory data analysis (EDA) efforts and require users to change parameters and input data multiple times. Thus, running time, computing needs, and scalability are probably key factors that researchers would like to consider when looking to deconvolve their datasets.

The authors achieve their aim to benchmark different deconvolution methods and the results from their thorough analysis support the conclusions that creating cell type deconvolution algorithms that can handle both cell abundance and rarity throughout a given tissue sample are challenging.

The reproducibility of the methods described will have significant utility for researchers looking to develop cell type deconvolution algorithms, as this platform will allow simultaneous replication of the described analysis and comparison to new methods.

Author Response

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

eLife assessment

This article presents important results describing how the gathering, integration, and broadcasting of information in the brain changes when consciousness is lost either through anesthesia or injury. They provide convincing evidence to support their conclusions, although the paper relies on a single analysis tool (partial information decomposition) and could benefit from a clearer explication of its conceptual basis, methodology, and results. The work will be of interest to both neuroscientists and clinicians interested in fundamental and clinical aspects of consciousness.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this paper, Luppi et al., apply the recently developed integrated information decomposition to the question how the architecture of information processing changes when consciousness is lost. They explore fMRI data from two different populations: healthy volunteers undergoing reversible anesthesia, as well as from patients who have long-term disorders of consciousness. They show that, in both populations, synergistic integration of information is disrupted in common ways. These results are interpreted in the context of the SAPHIRE model (recently proposed by this same group), that describes information processing in the brain as being composed of several distinct steps: 1) gatekeeping (where gateway regions introduce sensory information to the global synergistic workspace where 2) it is integrated or "processed" before 3) by broadcast back to to the brain.

I think that this paper is an excellent addition to the literature on information theory in neuroscience, and consciousness science specifically. The writing is clear, the figures are informative, and the authors do a good job of engaging with existing literature. While I do have some questions about the interpretations of the various information-theoretic measures, all in all, I think this is a significant piece of science that I am glad to see added to the literature.

One specific question I have is that I am still a little unsure about what "synergy" really is in this context. From the methods, it is defined as that part of the joint mutual information that is greater than the maximum marginal mutual information. While this is a perfectly fine mathematical measure, it is not clear to me what that means for a squishy organ like the brain. What should these results mean to a neuro-biologist or clinician?

Right now the discussion is very high level, equating synergy to "information processing" or "integrated information", but it might be helpful for readers not steeped in multivariate information theory to have some kind of toy model that gets worked out in detail. On page 15, the logical XOR is presented in the context of the single-target PID, but 1) the XOR is discrete, while the data analyzed here are continuous BOLD signals w/ Gaussian assumptions and 2) the XOR gate is a single-target system, while the power of the Phi-ID approach is the multi-target generality. Is there a Gaussian analog of the single-target XOR gate that could be presented? Or some multi-target, Gaussian toy model with enough synergy to be interesting? I think this would go a long way to making this work more accessible to the kind of interdisciplinary readership that this kind of article with inevitably attract.

We appreciate this observation. We now clarify that:

“redundancy between two units occurs when their future spontaneous evolution is predicted equally well by the past of either unit. Synergy instead occurs when considering the two units together increases the mutual information between the units’ past and their future – suggesting that the future of each is shaped by its interactions with the other. At the microscale (e.g., for spiking neurons) this phenomenon has been suggested as reflecting “information modification” 36,40,47. Synergy can also be viewed as reflecting the joint contribution of parts of the system to the whole, that is not driven by common input48.”

In the Methods, we have also added the following example to provide additional intuition about synergy in the case of continuous rather than discrete variables:

“As another example for the case of Gaussian variables (as employed here), consider a 2-node coupled autoregressive process with two parameters: a noise correlation c and a coupling parameter a. As c increases, the system is flooded by “common noise”, making the system increasingly redundant because the common noise “swamps” the signal of each node. As a increases, each node has a stronger influence both on the other and on the system as a whole, and we expect synergy to increase. Therefore, synergy reflects the joint contribution of parts of the system to the whole that is not driven by common noise. This has been demonstrated through computational modelling (Mediano et al 2019 Entropy).”

See below for the relevant parts of Figures 1 and 2 from Mediano et al (2019 Entropy), where Psi refers to the total synergy in the system.

Author response image 1.

Strengths

The authors have a very strong collection of datasets with which to explore their topic of interest. By comparing fMRI scans from patients with disorders of consciousness, healthy resting state, and various stages of propofol anesthesia, the authors have a very robust sample of the various ways consciousness can be perturbed, or lost. Consequently, it is difficult to imagine that the observed effects are merely a quirk of some biophysical effect of propofol specifically, or a particular consequence of long-term brain injury, but do in fact reflect some global property related to consciousness. The data and analyses themselves are well-described, have been previously validated, and are generally strong. I have no reason to doubt the technical validity of the presented results.

The discussion and interpretation of these results is also very nice, bringing together ideas from the two leading neurocognitive theories of consciousness (Global Workspace and Integrated Information Theory) in a way that feels natural. The SAPHIRE model seems plausible and amenable to future research. The authors discuss this in the paper, but I think that future work on less radical interventions (e.g. movie watching, cognitive tasks, etc) could be very helpful in refining the SAPHIRE approach.

Finally, the analogy between the PID terms and the information provided by each eye redundantly, uniquely, and synergistically is superb. I will definitely be referencing this intuition pump in future discussions of multivariate information sharing.

We are very grateful for these positive comments, and for the feedback on our eye metaphor.

Weaknesses

I have some concerns about the way "information processing" is used in this study. The data analyzed, fMRI BOLD data is extremely coarse, both in spatial and temporal terms. I am not sure I am convinced that this is the natural scale at which to talk about information "processing" or "integration" in the brain. In contrast to measures like sample entropy or Lempel-Ziv complexity (which just describe the statistics of BOLD activity), synergy and Phi are presented here as quasi-causal measures: as if they "cause" or "represent" phenomenological consciousness. While the theoretical arguments linking integration to consciousness are compelling, is this is right data set to explore them in? For example, the work by Newman, Beggs, and Sherril (nee Faber), synergy is associated with "computation" performed in individual neurons: the information about the future state of a target neuron that is only accessible when knowing both inputs (analogous to the synergy in computing the sum of two dice). Whether one thinks that this is a good approach neural computation or not, it fits within the commonly accepted causal model of neural spiking activity: neurons receive inputs from multiple upstream neurons, integrate those inputs and change their firing behavior accordingly.

In contrast, here, we are looking at BOLD data, which is a proxy measure for gross-scale regional neural activity, which itself is a coarse-graining of millions of individual neurons to a uni-dimensional spectrum that runs from "inactive to active." It feels as though a lot of inferences are being made from very coarse data.

We appreciate the opportunity to clarify this point. It is not our intention to claim that Phi-R and synergy, as measured at the level of regional BOLD signals, represent a direct cause of consciousness, or are identical to it. Rather, our work is intended to use these measures similarly to the use of sample entropy and LZC for BOLD signals: as theoretically grounded macroscale indicators, whose empirical relationship to consciousness may reveal the relevant underlying phenomena. In other words, while our results do show that BOLD-derived Phi-R tracks the loss and recovery of consciousness, we do not claim that they are the cause of it: only that an empirical relationship exists, which is in line with what we might expect on theoretical grounds. We have now clarified this in the Limitations section of our revised manuscript, as well as revising our language accordingly in the rest of the manuscript.

We also clarify that the meaning of “information processing” that we adopt pertains to “intrinsic” information that is present in the system’s spontaneous dynamics, rather than extrinsic information about a task:

“Information decomposition can be applied to neural data from different scales, from electrophysiology to functional MRI, with or without reference to behaviour 34. When behavioural data are taken into account, information decomposition can shed light on the processing of “extrinsic” information, understood as the translation of sensory signals into behavioural choices across neurons or regions 41,43,45,47. However, information decomposition can also be applied to investigate the “intrinsic” information that is present in the brain’s spontaneous dynamics in the absence of any tasks, in the same vein as resting-state “functional connectivity” and methods from statistical causal inference such as Granger causality 49. In this context, information processing should be understood in terms of the dynamics of information: where and how information is stored, transferred, and modified 34.”

References:

(1) Newman, E. L., Varley, T. F., Parakkattu, V. K., Sherrill, S. P. & Beggs, J. M. Revealing the Dynamics of Neural Information Processing with Multivariate Information Decomposition. Entropy 24, 930 (2022).

Reviewer #2 (Public Review):

The authors analysed functional MRI recordings of brain activity at rest, using state-of-the-art methods that reveal the diverse ways in which the information can be integrated in the brain. In this way, they found brain areas that act as (synergistic) gateways for the 'global workspace', where conscious access to information or cognition would occur, and brain areas that serve as (redundant) broadcasters from the global workspace to the rest of the brain. The results are compelling and consisting with the already assumed role of several networks and areas within the Global Neuronal Workspace framework. Thus, in a way, this work comes to stress the role of synergy and redundancy as complementary information processing modes, which fulfill different roles in the big context of information integration.

In addition, to prove that the identified high-order interactions are relevant to the phenomenon of consciousness, the same analysis was performed in subjects under anesthesia or with disorders of consciousness (DOC), showing that indeed the loss of consciousness is associated with a deficient integration of information within the gateway regions.

However, there is something confusing in the redundancy and synergy matrices shown in Figure 2. These are pair-wise matrices, where the PID was applied to identify high-order interactions between pairs of brain regions. I understand that synergy and redundancy are assessed in the way the brain areas integrate information in time, but it is still a little contradictory to speak about high-order in pairs of areas. When talking about a "synergistic core", one expects that all or most of the areas belonging to that core are simultaneously involved in some (synergistic) information processing, and I do not see this being assessed with the currently presented methodology. Similarly, if redundancy is assessed only in pairs of areas, it may be due to simple correlations between them, so it is not a high-order interaction. Perhaps it is a matter of language, or about the expectations that the word 'synergy' evokes, so a clarification about this issue is needed. Moreover, as the rest of the work is based on these 'pair-wise' redundancy and synergy matrices, it becomes a significative issue.

We are grateful for the opportunity to clarify this point. We should highlight that PhiID is in fact assessing four variables: the past of region X, the past of region B, the future of region X, and the future of region Y. Since X and Y each feature both in the past and in the future, we can re-conceptualise the PhiID outputs as reflecting the temporal evolution of how X and Y jointly convey information: the persistent redundancy that we consider corresponds to information that is always present in both X and Y; whereas the persistent synergy is information that X and Y always convey synergistically. In contrast, information transfer would correspond to the phenomenon whereby information was conveyed by one variable in the past, and by the other in the future (see Luppi et al., 2024 TICS; and Mediano et al., 2021 arXiv for more thorough discussions on this point). We have now added this clarification in our Introduction and Results, as well as adding the new Figure 2 to clarify the meaning of PhiID terms.

We would also like to clarify that all the edges that we identify as significantly changing are indeed simultaneously involved in the difference between consciousness and unconsciousness. This is because the Network-Based Statistic differs from other ways of identifying edges that are significantly different between two groups or conditions, because it does not consider edges in isolation, but only as part of a single connected component.

Reviewer #3 (Public Review):

The work proposes a model of neural information processing based on a 'synergistic global workspace,' which processes information in three principal steps: a gatekeeping step (information gathering), an information integration step, and finally, a broadcasting step. The authors determined the synergistic global workspace based on previous work and extended the role of its elements using 100 fMRI recordings of the resting state of healthy participants of the HCP. The authors then applied network analysis and two different measures of information integration to examine changes in reduced states of consciousness (such as anesthesia and after-coma disorders of consciousness). They provided an interpretation of the results in terms of the proposed model of brain information processing, which could be helpful to be implemented in other states of consciousness and related to perturbative approaches. Overall, I found the manuscript to be well-organized, and the results are interesting and could be informative for a broad range of literature, suggesting interesting new ideas for the field to explore. However, there are some points that the authors could clarify to strengthen the paper. Key points include:

(1) The work strongly relies on the identification of the regions belonging to the synergistic global workspace, which was primarily proposed and computed in a previous paper by the authors. It would be great if this computation could be included in a more explicit way in this manuscript to make it self-contained. Maybe include some table or figure being explicit in the Gradient of redundancy-to-synergy relative importance results and procedure.

We have now added the new Supplementary Figure 1 to clarify how the synergistic workspace is identified, as per Luppi et al (2022 Nature Neuroscience).

(2) It would be beneficial if the authors could provide further explanation regarding the differences in the procedure for selecting the workspace and its role within the proposed architecture. For instance, why does one case uses the strength of the nodes while the other case uses the participation coefficient? It would be interesting to explore what would happen if the workspace was defined directly using the participation coefficient instead of the strength. Additionally, what impact would it have on the procedure if a different selection of modules was used? For example, instead of using the RSN, other criteria, such as modularity algorithms, PCA, Hidden Markov Models, Variational Autoencoders, etc., could be considered. The main point of my question is that, probably, the RSN are quite redundant networks and other methods, as PCA generates independent networks. It would be helpful if the authors could offer some comments on their intuition regarding these points without necessarily requiring additional computations.

We appreciate the opportunity to clarify this point. Our rationale for the procedure used to identify the workspace is to find regions where synergy is especially prominent. This is due to the close mathematical relationship between synergistic information and integration of information (see also Luppi et al., 2024 TICS), which we view as the core function of the global workspace. This identification is based on the strength ranking, as per Luppi et al (2022 Nature Neuroscience), which demonstrated that regions where synergy predominates (i.e., our proposed workspace) are also involved with high-level cognitive functions and anatomically coincide with transmodal association cortices at the confluence of multiple information streams. This is what we should expect of a global workspace, which is why we use the strength of synergistic interactions to identify it, rather than the participation coefficient. Subsequently, to discern broadcasters from gateways within the synergistic workspace, we seek to encapsulate the meaning of a “broadcaster” in information terms. We argue that this corresponds with making the same information available to multiple modules. Sameness of information corresponds to redundancy, and multiplicity of modules can be reflected in the network-theoretic notion of participation coefficient. Thus, a broadcaster is a region in the synergistic workspace (i.e., a region with strong synergistic interactions) that in addition has a high participation coefficient for its redundant interactions.

Pertaining specifically to the use of resting-state networks as modules, indeed our own (Luppi et al., 2022 Nature Neuroscience) and others’ research has shown that each RSN entertains primarily redundant interactions among its constituent regions. This is not surprising, since RSNs are functionally defined: their constituent elements need to process the same information (e.g., pertaining to a visual task in case of the visual network). We used the RSNs as our definition of modules, because they are widely understood to reflect the intrinsic organisation of brain activity into functional units; for example, Smith et al., (2009 PNAS) and Cole et al (2014 Neuron) both showed that RSNs reflect task-related co-activation of regions, whether directly quantified from fMRI in individuals performing multiple tasks, or inferred from meta-analysis of the neuroimaging literature. This is the aspect of a “module” that matters from the global workspace perspective: modules are units with distinct function, and RSNs capture this well. This is therefore why we use the RSNs as modules when defining the participation coefficient: they provide an a-priori division into units with functionally distinct roles.

Nonetheless, we also note that RSN organisation is robustly recovered using many different methods, including seed-based correlation from specific regions-of-interest, or Independent Components Analysis, or community detection on the network of inter-regional correlations - demonstrating that they are not merely a function of the specific method used to identify them. In fact, we show significant correlation between participation coefficient defined in terms of RSNs, and in terms of modules identified in a purely data-driven manner from Louvain consensus clustering (Figure S4).

(3) The authors acknowledged the potential relevance of perturbative approaches in terms of PCI and quantification of consciousness. It would be valuable if the authors could also discuss perturbative approaches in relation to inducing transitions between brain states. In other words, since the authors investigate disorders of consciousness where interventions could provide insights into treatment, as suggested by computational and experimental works, it would be interesting to explore the relationship between the synergistic workspace and its modifications from this perspective as well.

We thank the Reviewer for bringing this up: we now cite several studies that in recent years have applied perturbative approaches to induce transitions between states of consciousness.

“The PCI is used as a means of assessing the brain’s current state, but stimulation protocols can also be adopted to directly induce transitions between states of consciousness. In rodents, carbachol administration to frontal cortex awakens rats from sevoflurane anaesthesia120, and optogenetic stimulation was used to identify a role of central thalamus neurons in controlling transitions between states of responsiveness121,122. Additionally, several studies in non-human primates have now shown that electrical stimulation of the central thalamus can reliably induce awakening from anaesthesia, accompanied by the reversal of electrophysiological and fMRI markers of anaesthesia 123–128. Finally, in human patients suffering from disorders of consciousness, stimulation of intra-laminar central thalamic nuclei was reported to induce behavioural improvement 129, and ultrasonic stimulation 130,131 and deep-brain stimulation are among potential therapies being considered for DOC patients 132,133. It will be of considerable interest to determine whether our corrected measure of integrated information and topography of the synergistic workspace also restored by these causal interventions.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

I would appreciate it if the authors could revisit the figures and make sure that:

(1) All fonts are large enough to be readable for people with visual impairments (for ex. the ranges on the colorbars in Fig. 2 are unreadably small).

Thank you: we have increased font sizes.

(2) The colormaps are scaled to show meaningful differences (Fig. 2A)

We have changed the color scale in Figure 2A and 2B.

Also, the authors may want to revisit the references section: some of the papers that were pre-prints at one point have now been published and should be updated.

Thank you: we have updated our references.

Minor comments:

• In Eqs. 2 and 3, the unique information term uses the bar notation ( | ) that is typically indicative of "conditioned on." Perhaps the authors could use a slash notation (e.g. Unq(X ; Z / Y)) to avoid this ambiguity? My understanding of the Unique information is that it is not necessarily "conditioned on", so much as it is "in the context of".

Indeed, the “|” sign of “conditioning” could be misleading; however, the “/” sign could also be misleading, if interpreted as division. Therefore, we have opted for the “\” sign of “set difference”, in Eq 2 and 3, which is conceptually more appropriate in this context.

• The font on the figures is a little bit small - for readers with poor eyes, it might be helpful to increase the wording size.

We have increased font sizes in the figures where relevant.

• I don't quite understand what is happening in Fig. 2A - perhaps it is a colormap issue, but it seems as though it's just a bit white square? It looks like redundancy is broadly correlated with FC (just based on the look of the adjacency matrices), but I have no real sense of what the synergistic matrix looks like, other than "flat."

We have now changed the color scale in Figure 2.

Reviewer #2 (Recommendations For The Authors):

Besides the issues mentioned in the Public review, I have the following suggestions to improve the manuscript:

• At the end of the introduction, a few lines could be added explaining why the study of DOC patients and subjects under anesthesia will be informative in the context of this work.

By comparing functional brain scans from transient anaesthetic-induced unconsciousness and from the persistent unconsciousness of DOC patients, which arises from brain injury, we can search for common brain changes associated with loss of consciousness – thereby disambiguating what is specific to loss of consciousness.

• On page and in general the first part of Results, it is not evident that you are working with functional connectivity. Many times the word 'connection' is used and sometimes I was wondering whether they were structural or functional. Please clarify. Also, the meaning of 'synergistic connection' or 'redundant connection' could be explained in lay terms.

Thank you for bringing this up. We have now replaced the word “connection” with “interaction” to disambiguate this issue, further adding “functional” where appropriate. We have also provided, in the Introduction, an intuitive explanation of what synergy and redundancy mean int he context of spontaneous fMRI signals.

• Figure 2 needs a lot of improvement. The matrix of synergistic interactions looks completely yellow-ish with some vague areas of white. So everything is above 2. What does it mean?? Pretty uninformative. The matrix of redundant connections looks a lot of black, with some red here and there. So everything is below 0.6. Also, what are the meaning and units of the colorbars?.

We agree: we have increased font sizes, added labels, and changed the color scale in Figure 2. We hope that the new version of Figure 2 will be clearer.

• Caption of Figure 2 mentions "... brain regions identified as belonging to the synergistic global workspace". I didn't get it clear how do you define these areas. Are they just the sum of gateways and broadcasters, or is there another criterion?

Regions belonging to the synergistic workspace are indeed the set comprising gateways and broadcasters; they are the regions that are synergy-dominated, as defined in Luppi et al., 2022 Nature Neuroscience. We have now clarified this in the figure caption.

• In the first lines of page 7, it is said that data from DOC and anesthesia was parcellated in 400 + 54 regions. However, it was said in a manner that made me think it was a different parcellation than the other data. Please make it clear that the parcellation is the same (if it is).

We have now clarified that the 400 cortical regions are from the Schaefer atlas, and 54 subcortical regions from the Tian atlas, as for the other analysis. The only other parcellation that we use is the Schaefer-232, for the robustness analysis. This is also reported in the Methods.

• Figure 3: the labels in the colorbars cannot be read, please make them bigger. Also, the colorbars and colorscales should be centered in white, to make it clear that red is positive and blue is negative. O at least maintain consistency across the panels (I can't tell because of the small numbers).

Thank you: we have increased font sizes, added labels, indicated that white refers to zero (so that red is always an increase, and blue is always a decrease), and changed the color scale in Figure 2.

• The legend of Figure 4 is written in a different style, interpreting the figure rather than describing it. Please describe the figure in the caption, in order to let the read know what they are looking at.

We have endeavoured to rewrite the legend of Figure 4 in a style that is more consistent with the other figures.

• In several parts the 'whole-minus-sum' phi measure is mentioned and it is said that it did not decrease during loss of consciousness. However, I did not see any figure about that nor any conspicuous reference to that in Results text. Where is it?

We apologise for the confusion: this is Figure S3A, in the Supplementary. We have now clarified this in the text.

Reviewer #3 (Recommendations For The Authors):

(1) In the same direction, regarding Fig. 2, in my opinion, it does not effectively aid in understanding the selection of regions as more synergistic or redundant. In panels A) and B), the color scales could be improved to better distinguish regions in the matrices (panel A) is saturated at the upper limit, while panel B) is saturated at the lower limit). Additionally, I suggest indicating in the panels what is being measured with the color scales.

Thank you: we have increased font sizes, added labels, and changed the color scale in Figure 2.

(2) When investigating the synergistic core of human consciousness and interpreting the results of changes in information integration measures in terms of the proposed framework, did the authors consider the synergistic workspace computed in HCP data? If the answer is positive, it would be helpful for the authors to be more explicit about it and elaborate on any differences that may be found, as well as the potential impact on interpretation.

This is correct: the synergistic workspace, including gateways and broadcasters, are identified from the Human Connectome Project dataset. We now clarify this in the manuscript.

Minors:

(1) I would suggest improving the readability of figures 2 and 3, considering font size (letters and numbers) and color bars (numbers and indicate what is measured with this scale). In Figure 1, the caption defines steps instead stages that are indicated in the figure.

Thank you: we have increased font sizes, added labels, and replaced steps with “stages” in Figure 1.

Reviewer #2 (Public Review):

The authors analysed functional MRI recordings of brain activity at rest, using state-of-the-art methods that reveal the diverse ways in which the information can be integrated in the brain. In this way, they found brain areas that act as (synergistic) gateways for the 'global workspace', where conscious access to information or cognition would occur, and brain areas that serve as (redundant) broadcasters from the global workspace to the rest of the brain. The results are compelling and consisting with the already assumed role of several networks and areas within the Global Neuronal Workspace framework. Thus, in a way, this work comes to stress the role of synergy and redundancy as complementary information processing modes, which fulfill different roles in the big context of information integration.<br /> In addition, to prove that the identified high-order interactions are relevant to the phenomenon of consciousness, the same analysis was performed in subjects under anesthesia or with disorders of consciousness (DOC), showing that indeed the loss of consciousness is associated with a deficient integration of information within the gateway regions.

However, there is still a standing issue that could be the basis for an improved analysis: the concepts of gateways and broadcasters allude to a directionality in the information flow. In fact, Figure 1 depicts Stage (i) and Stage (iii) as one-way processes. However, the identification of gateway and broadcaster regions relies on matrices that are symmetrical, i.e. they are not directed. Would it be possible to assess the gateway or broadcaster nature of a region taking into account the directionality of the information flow? In other words, if region X is a gateway, I would expect a synergistic relationship between the past of X,Y and present of Y (Y not being a gateway) towards the present of X; but not necessarily the other way around (i.e. the present of Y being less dependent on the past/present of X). A similar reasoning can be made for broadcasters.

Although regional differences in haemodynamics complicate attempts to map directed information flow from fMRI recordings, perhaps the IID framework could be leveraged to extract directed data (i.e., there are many atoms that are explicitly directed). As an avenue for future research, it would be interesting to discuss the feasibility or limitations of such analysis.

Also, there is something confusing in Figure 4B-C and its description. Awake should be similar to recovery (they are both awake, aren't they? Not much info is given, anyway); thus it seems counterintuitive that anesthesia minus awake looks so different than anesthesia minus recovery. The first is mostly blue-ish and the second is mostly red. Is it possible that Figure 4C is actually recovery minus anesthesia? That would make much more sense, also for Figure 4D. Please correct me if I am wrong.

Reviewer #3 (Public Review):

The work proposes a model of neural information processing based on a 'synergistic global workspace,' which processes information in three principal steps: a gatekeeping step (information gathering), an information integration step, and finally, a broadcasting step. They provided an interpretation of the reduced human consciousness states in terms of the proposed model of brain information processing, which could be helpful to be implemented in other states of consciousness. The manuscript is well-organized, and the results are important and could be interesting for a broad range of literature, suggesting interesting new ideas for the field to explore.

Comments on revised version:

The authors have addressed all my comments made in the previous revision.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Qin et al. set out to investigate the role of mechanosensory feedback during swallowing and identify neural circuits that generate ingestion rhythms. They use Drosophila melanogaster swallowing as a model system, focusing their study on the neural mechanisms that control cibarium filling and emptying in vivo. They find that pump frequency is decreased in mutants of three mechanotransduction genes (nompC, piezo, and Tmc), and conclude that mechanosensation mainly contributes to the emptying phase of swallowing. Furthermore, they find that double mutants of nompC and Tmc have more pronounced cibarium pumping defects than either single mutants or Tmc/piezo double mutants. They discover that the expression patterns of nompC and Tmc overlap in two classes of neurons, md-C and md-L neurons. The dendrites of md-C neurons warp the cibarium and project their axons to the subesophageal zone of the brain. Silencing neurons that express both nompC and Tmc leads to severe ingestion defects, with decreased cibarium emptying. Optogenetic activation of the same population of neurons inhibited filling of the cibarium and accelerated cibarium emptying. In the brain, the axons of nompC∩Tmc cell types respond during ingestion of sugar but do not respond when the entire fly head is passively exposed to sucrose. Finally, the authors show that nompC∩Tmc cell types arborize close to the dendrites of motor neurons that are required for swallowing, and that swallowing motor neurons respond to the activation of the entire Tmc-GAL4 pattern.

Strengths:

• The authors rigorously quantify ingestion behavior to convincingly demonstrate the importance of mechanosensory genes in the control of swallowing rhythms and cibarium filling and emptying

• The authors demonstrate that a small population of neurons that express both nompC and Tmc oppositely regulate cibarium emptying and filling when inhibited or activated, respectively

• They provide evidence that the action of multiple mechanotransduction genes may converge in common cell types

Thank you for your insightful and detailed assessment of our work. Your constructive feedback will help to improve our manuscript.

Weaknesses:

• A major weakness of the paper is that the authors use reagents that are expressed in both md-C and md-L but describe the results as though only md-C is manipulated-Severing the labellum will not prevent optogenetic activation of md-L from triggering neural responses downstream of md-L. Optogenetic activation is strong enough to trigger action potentials in the remaining axons. Therefore, Qin et al. do not present convincing evidence that the defects they see in pumping can be specifically attributed to md-C.

Thank you for your comments. This is important point that we did not adequately address in the original preprint. We have obtained imaging and behavioral results that strongly suggest md-C, rather than md-L, are essential for swallowing behavior.

36 hours after the ablation of the labellum, the signals of md-L were hardly observable when GFP expression was driven by the intersection between Tmc-GAL4 & nompC-QF (see F Figure 3—figure supplement 1A). This observation indicates that the axons of md-L likely degenerated after 36 hours, and were unlikely to influence swallowing. Moreover, the projecting pattern of Tmc-GAL4 & nompC-QF>>GFP exhibited no significant changes in the brain post labellum ablation.

Furthermore, even after labellum ablation for 36 hours, flies exhibited responses to light stimulation (see Figure 3—figure supplement 1B-C, Video 5) when ReaChR was expressed in md-C. We thus reasoned that md-C but not md-L, plays a crucial role in the swallowing process.

• GRASP is known to be non-specific and prone to false positives when neurons are in close proximity but not synaptically connected. A positive GRASP signal supports but does not confirm direct synaptic connectivity between md-C/md-L axons and MN11/MN12.

In this study, we employed the nSyb-GRASP, wherein the GRASP is expressed at the presynaptic terminals by fusion with the synaptic marker nSyb. This method demonstrates an enhanced specificity compared to the original GRASP approach.

Additionally, we utilized +/ UAS-nSyb-spGFP1-10, lexAop-CD4-spGFP11 ; + / MN-LexA fruit flies as a negative control to mitigate potential false signals originating from the tool itself (Author response image 1, scale bar = 50μm). Beside the genotype Tmc-Gal4, Tub(FRT. Gal80) / UAS-nSyb-spGFP1-10, lexAop-CD4-spGFP11 ; nompC-QF, QUAS-FLP / MN-LexA fruit flies discussed in this manuscript, we also incorporated genotype Tmc-Gal4, Tub(FRT. Gal80) / lexAop-nSyb-spGFP1-10, UAS-CD4-spGFP11 ; nompC-QF, QUAS-FLP / MN-LexA fruit flies as a reverse control (Author response image 2). Unexpectedly, similar positive signals were observed, indicating that, positive signals may emerge due to close proximity between neurons even with nSyb-GRASP.

Author response image 1

It should be noted that the existence of synaptic projections from motor neurons (MN) to md-C cannot be definitively confirmed at this juncture. At present, we can only posit the potential for synaptic connections between md-C and motor neurons. A more conclusive conclusion may be attainable with the utilization of comprehensive whole-brain connectome data in future studies.

Author response image 2

• As seen in Figure 2—figure supplement 1, the expression pattern of Tmc-GAL4 is broader than md-C alone. Therefore, the functional connectivity the authors observe between Tmc expressing neurons and MN11 and 12 cannot be traced to md-C alone

It is true that the expression pattern of Tmc-GAL4 is broader than that of md-C alone. Our experiments, including those flies expressing TNT in Tmc+ neurons, demonstrated difficulties in emptying (Figure 2A, 2D). Notably, we encountered challenges in finding fly stocks bearing UAS>FRT-STOP-P2X2. Consequently, we opted to utilize Tmc-GAL4 to drive UAS-P2X2 instead. We believe that the results further support our hypothesis on the role of md-C in the observed behavioral change in emptying.

Overall, this work convincingly shows that swallowing and swallowing rhythms are dependent on several mechanosensory genes. Qin et al. also characterize a candidate neuron, md-C, that is likely to provide mechanosensory feedback to pumping motor neurons, but the results they present here are not sufficient to assign this function to md-C alone. This work will have a positive impact on the field by demonstrating the importance of mechanosensory feedback to swallowing rhythms and providing a potential entry point for future investigation of the identity and mechanisms of swallowing central pattern generators.

Reviewer #2 (Public Review):

In this manuscript, the authors describe the role of cibarial mechanosensory neurons in fly ingestion. They demonstrate that pumping of the cibarium is subtly disrupted in mutants for piezo, TMC, and nomp-C. Evidence is presented that these three genes are co-expressed in a set of cibarial mechanosensory neurons named md-C. Silencing of md-C neurons results in disrupted cibarial emptying, while activation promotes faster pumping and/or difficulty filling. GRASP and chemogenetic activation of the md-C neurons is used to argue that they may be directly connected to motor neurons that control cibarial emptying.

The manuscript makes several convincing and useful contributions. First, identifying the md-C neurons and demonstrating their essential role for cibarium emptying provides reagents for further studying this circuit and also demonstrates the important of mechanosensation in driving pumping rhythms in the pharynx. Second, the suggestion that these mechanosensory neurons are directly connected to motor neurons controlling pumping stands in contrast to other sensory circuits identified in fly feeding and is an interesting idea that can be more rigorously tested in the future.

At the same time, there are several shortcomings that limit the scope of the paper and the confidence in some claims. These include:

a) the MN-LexA lines used for GRASP experiments are not characterized in any other way to demonstrate specificity. These were generated for this study using Phack methods, and their expression should be shown to be specific for MN11 and MN12 in order to interpret the GRASP experiments.

Thanks for the suggestion. We have checked the expression pattern of MN-LexA, which is similar to MN-GAL4 used in previous work (Manzo et al., PNAS., 2012, PMID:22474379) . Here is the expression pattern:

Author response image 3

b) There is also insufficient detail for the P2X2 experiment to evaluate its results. Is this an in vivo or ex vivo prep? Is ATP added to the brain, or ingested? If it is ingested, how is ATP coming into contact with md-C neuron if it is not a chemosensory neuron and therefore not exposed to the contents of the cibarium?

The P2X2 experimental preparation was done ex vivo. We immersed the fly in the imaging buffer, as described in the Methods section under Functional Imaging. Following dissection and identification of the subesophageal zone (SEZ) area under fluorescent microscopy, we introduced ATP slowly into the buffer, positioned at a distance from the brain

c) In Figure 3C, the authors claim that ablating the labellum will remove the optogenetic stimulation of the md-L neuron (mechanosensory neuron of the labellum), but this manipulation would presumably leave an intact md-L axon that would still be capable of being optogenetically activated by Chrimson.

Please refer to the corresponding answers for reviewer 1 and Figure 3—figure supplement 1.

d) Average GCaMP traces are not shown for md-C during ingestion, and therefore it is impossible to gauge the dynamics of md-C neuron activation during swallowing. Seeing activation with a similar frequency to pumping would support the suggested role for these neurons, although GCaMP6s may be too slow for these purposes.

Profiling the dynamics of md-C neuron activation during swallowing is crucial for unraveling the operational model of md-C and validating our proposed hypothesis. Unfortunately, our assay faces challenges in detecting probable 6Hz fluorescent changes with GCaMP6s.

In general, we observed an increase of fluorescent signals during swallowing, but movement of alive flies during swallowing influenced the imaging recording, so we could not depict a decent tracing for calcium imaging for md-C neurons. To enhance the robustness of our findings, patching the md-C neurons would be a more convincing approach. As illustrated in Figure 2, the somata of md-C neurons are situated in the cibarium rather than the brain. patching of the md-C neuron somata in flies during ingestion is difficult.

e) The negative result in Figure 4K that is meant to rule out taste stimulation of md-C is not useful without a positive control for pharyngeal taste neuron activation in this same preparation.

We followed methods used in the previous work (Chen et al., Cell Rep., 2019, PMID:31644916), which we believe could confirm that md-C do not respond to sugars.

In addition to the experimental limitations described above, the manuscript could be organized in a way that is easier to read (for example, not jumping back and forth in figure order).

Thanks for your suggestion and the manuscript has been reorganized.

Reviewer #3 (Public Review):

Swallowing is an essential daily activity for survival, and pharyngo-laryngeal sensory function is critical for safe swallowing. In Drosophila, it has been reported that the mechanical property of food (e.g. Viscosity) can modulate swallowing. However, how mechanical expansion of the pharynx or fluid content sense and control swallowing was elusive. Qin et al. showed that a group of pharyngeal mechanosensory neurons, as well as mechanosensory channels (nompC, Tmc, and Piezo), respond to these mechanical forces for regulation of swallowing in Drosophila melanogaster.

Strengths:

There are many reports on the effect of chemical properties of foods on feeding in fruit flies, but only limited studies reported how physical properties of food affect feeding especially pharyngeal mechanosensory neurons. First, they found that mechanosensory mutants, including nompC, Tmc, and Piezo, showed impaired swallowing, mainly the emptying process. Next, they identified cibarium multidendritic mechanosensory neurons (md-C) are responsible for controlling swallowing by regulating motor neuron (MN) 12 and 11, which control filling and emptying, respectively.

Weaknesses:

While the involvement of md-C and mechanosensory channels in controlling swallowing is convincing, it is not yet clear which stimuli activate md-C. Can it be an expansion of cibarium or food viscosity, or both? In addition, if rhythmic and coordinated contraction of muscles 11 and 12 is essential for swallowing, how can simultaneous activation of MN 11 and 12 by md-C achieve this? Finally, previous reports showed that food viscosity mainly affects the filling rather than the emptying process, which seems different from their finding.

We have confirmed that swallowing sucrose water solution activated md-C neurons, while sucrose water solution alone could not (Figure 4J-K). We hypothesized that the viscosity of the food might influence this expansion process.

While we were unable to delineate the activation dynamics of md-C neurons, our proposal posits that these neurons could be activated in a single pump cycle, sequentially stimulating MN12 and MN11. Another possibility is that the activation of md-C neurons acts as a switch, altering the oscillation pattern of the swallowing central pattern generator (CPG) from a resting state to a working state.

In the experiments with w1118 flies fed with MC (methylcellulose) water, we observed that viscosity predominantly affects the filling process rather than the emptying process, consistent with previous findings. This raises an intriguing question. Our investigation into the mutation of mechanosensitive ion channels revealed a significant impact on the emptying process. We believe this is due to the loss of mechanosensation affecting the vibration of swallowing circuits, thereby influencing both the emptying and filling processes. In contrast, viscosity appears to make it more challenging for the fly to fill the cibarium with food, primarily attributable to the inherent properties of the food itself.

Reviewer #4 (Public Review):

A combination of optogenetic behavioral experiments and functional imaging are employed to identify the role of mechanosensory neurons in food swallowing in adult Drosophila. While some of the findings are intriguing and the overall goal of mapping a sensory to motor circuit for this rhythmic movement are admirable, the data presented could be improved.

The circuit proposed (and supported by GRASP contact data) shows these multi-dendritic neurons connecting to pharyngeal motor neurons. This is pretty direct - there is no evidence that they affect the hypothetical central pattern generator - just the execution of its rhythm. The optogenetic activation and inhibition experiments are constitutive, not patterned light, and they seem to disrupt the timing of pumping, not impose a new one. A slight slowing of the rhythm is not consistent with the proposed function.

Motor neurons implicated in patterned motions can be considered effectors of Central Pattern Generators (CPGs)(Marder et al., Curr Biol., 2001, PMID: 11728329; Hurkey et al., Nature., 2023, PMID:37225999). Given our observation of the connection between md-C neurons and motor neurons, it is reasonable to speculate that md-C neurons influence CPGs. Compared to the patterned light (0.1s light on and 0.1s light off) used in our optogenetic experiments, we noted no significant changes in their responses to continuous light stimulation. We think that optogenetic methods may lead to overstimulation of md-C neurons, failing to accurately mimic the expansion of the cibarium during feeding.

Dysfunction in mechanosensitive ion channels or mechanosensory neurons not only disrupts the timing of pumping but also results in decreased intake efficiency (Figure 1E). The water-swallowing rhythm is generally stable in flies, and swallowing is a vital process that may involve redundant ion channels to ensure its stability.

The mechanosensory channel mutants nompC, piezo, and TMC have a range of defects. The role of these channels in swallowing may not be sufficiently specific to support the interpretation presented. Their other defects are not described here and their overall locomotor function is not measured. If the flies have trouble consuming sufficient food throughout their development, how healthy are they at the time of assay? The level of starvation or water deprivation can affect different properties of feeding - meal size and frequency. There is no description of how starvation state was standardized or measured in these experiments.

Defects in mechanosensory channel mutants nompC, piezo, and TMC, have been extensively investigated (Hehlert et al., Trends Neurosci., 2021, PMID:332570000). Mutations in these channels exhibit multifaceted effects, as illustrated in our RNAi experiments (see Figure 2E). Deprivation of water and food was performed in empty fly vials. It's important to note that the duration of starvation determines the fly's willingness to feed but not the pump frequency (Manzo et al., PNAS., 2012, PMID:22474379).

In most cases, female flies were deprived water and food in empty vials for 24 hours because after that most flies would be willing to drink water. The deprivation time is 12 hours for flies with nompC and Tmc mutated or flies with Kir2.1 expressed in md-C neurons, as some of these flies cannot survive 24h deprivation.

The brain is likely to move considerably during swallow, so the GCaMP signal change may be a motion artifact. Sometimes this can be calculated by comparing GCaMP signal to that of a co-expressed fluorescent protein, but there is no mention that this is done here. Therefore, the GCaMP data cannot be interpreted.

We did not co-express a fluorescent protein with GCaMP for md-C. The head of the fly was mounted onto a glass slide, and we did not observe significant signal changes before feeding.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

.>Abstract: I disagree that swallow is the first step of ingestion. The first paragraph also mentions the final checkpoint before food ingestion. Perhaps sufficient to say that swallow is a critical step of ingestion.

Indeed, it is not rigorous enough to say “first step”. This has been replaced by “early step”.

Introduction:

Line 59: "Silence" should be "Silencing"

This has been replaced.

Results:

Lines 91-92: I am not clear about what this means. 20% of nompC and 20% of wild-type flies exhibit incomplete filling? So nompC is not different from wild-type?

Sorry for the mistake. Viscous foods led to incomplete emptying (not incomplete filling), as displayed in Video 4. The swallowing behavior differs between nompC mutants and wild-type flies, as illustrated in Figure 1C, Figure 1—figure supplement 1A-C and video 1&5.

When fed with 1% MC water solution (Figure 1—figure supplement 1E-H). We found that when fed with 1% MC watere solution, Tmc or piezo mutants displayed incomplete emptying, which could constitute a long time proportion of swallowing behavior; while only 20% of nompC flies and 20% of wild-type flies sporadically exhibit incomplete emptying, which is significantly different. Though the percent of flies displaying incomplete pump is similar between nompC mutant and wild-type files, you can find it quite different in video 1 and 5.

Line 94: Should read: “while for foods with certain viscosity, the pump of Tmc or piezo mutants might"

What evidence is there for weakened muscle motion? The phenotypes of all three mutants is quite similar, so concluding that they have roles in initiation versus swallowing strength is not well supported -this would be better moved to the discussion since it is speculative.

Muscles are responsible for pumping the bolus from the mouth to the crop. In the case of Tmc or piezo mutants, as evidenced by incomplete filling for viscous foods (see Video 4), we speculate that the loss of sensory stimuli leads to inadequate muscle contraction. The phenotypes observed in Tmc and piezo mutants are similar yet distinct from those of the wild-type or nompC mutant, as shown in Video 1 and 4. The phrase "due to weakened muscle motion" has been removed for clarity.

Line 146: If md-L neurons are also labeled by this intersection, then you are not able to know whether the axons seen in the brain are from md-L or md-C neurons. Line 148: cutting the labellum is not sufficient to ablate md-L neurons. The projections will still enter the brain and can be activated with optogenetics, even after severing the processes that reside in the labellum.

Please refer to the responses for reviewer #1 (Public Review):” A major weakness of the paper…” and Figure 4.

Line 162: If the fly head alone is in saline, do you know that the sucrose enters the esophagus? The more relevant question here is whether the md-C neurons respond to mechanical force. If you could artificially inflate the cibarium with air and see the md-C neurons respond that would be a more convincing result. So far you only know that these are activated during ingestion, but have not shown that they are activated specifically by filling or emptying. In addition, you are not only imaging md-C (md-L is also labeled). This caveat should be mentioned.

We followed the methods outlined in the previous work (Chen et al., Cell Rep., 2019, PMID:31644916), which suggested that md-C neurons do not respond to sugars. While we aimed to mechanically stimulate md-C neurons, detecting signal changes during different steps of swallowing is challenging. This aspect could be further investigated in subsequent research with the application of adequate patch recording or two-photon microscopy (TPM).

Figure 3: It is not clear what the pie charts in Figure 3 A refer to. What are the three different rows, and what does blue versus red indicate?

Figure 3A illustrates three distinct states driven by CsChrimson light stimulation of md-C neurons, with the proportions of flies exhibiting each state. During light activation, flies may display difficulty in filling, incomplete filling, or a normal range of pumping. The blue and red bars represent the proportions of flies showing the corresponding state, as indicated by the black line.

Figure 4: Where are the example traces for J? The comparison in K should be average dF/F before ingestion compared with average dF/F during ingestion. Comparing the in vitro response to sucrose to the in vivo response during ingestion is not a useful comparison.

Please refer to the answers for reviewer #2 question d).

Reviewer #2 (Recommendations For The Authors):

Suggested experiments that would address some of my concerns listed in the public review include:

a) high resolution SEZ images of MN-LexA lines crossed to LexAop-GFP to demonstrate their specificity

b) more detail on the P2X2 experiment. It is hard to make suggestions beyond that without first seeing the details.

c) presenting average GCaMP traces for all calcium imaging results

d) to rule out taste stimulation of md-C (Figure 4K) I would suggest performing more extensive calcium imaging experiments with different stimuli. For example, sugar, water, and increasing concentrations of a neutral osmolyte (e.g. PEG) to suppress the water response. I think that this is more feasible than trying to get an in vitro taste prep to be convincing.

Please refer to the responses for public review of reviewer #2.

Reviewer #3 (Recommendations For The Authors):

Below I list my suggestions as well as criticisms.

(1) It would be excellent if the authors could demonstrate whether varying levels of food viscosity affect md-C activation.

That is a good point, and could be studied in future work.

(2) It is not clear whether an intersectional approach using TMC-GAL4 and nompC-QF abolishes labelling of the labellar multidendritic neurons. If this is the case, please show labellar multidendritic neurons in TMC-GAL4 only flies and flies using the intersectional approach. Along with this question, I am concerned that labellum-removed flies could be used for feeding assay.

Intersectional labelling using TMC-GAL4 and nompC-QF could not abolish labelling of the labellar multidendritic neurons (Author response image 4). Labellum-removed flies could be used for feeding assay (Figure 3—figure supplement 1B-C, video 5), but once LSO or cibarium of fly was damaged, swallowing behavior would be affected. Removing labellum should be very careful.

Author response image 4

(3) Please provide the detailed methods for GRASP and include proper control.

Please refer to the responses for public review of reviewer #1.

(4) The authors hypothesized that md-C sequentially activates MN11 and 12. Is the time gap between applying ATP on md-C and activation of MN11 or MN12 different? Please refer to the responses for public review of reviewer #3. The time gap between applying ATP on md-C and activation of MN11 or MN12 didn’t show significant differences, and we think the reason is that the ex vivo conditions could not completely mimic in vivo process.

I found the manuscript includes many errors, which need to be corrected.

(1) The reference formatting needs to be rechecked, for example, lines 37, 42, and 43.

(2) Line 44-46: There is some misunderstanding. The role of pharyngeal mechanosensory neurons is not known compared with chemosensory neurons.

(3) Line 49: Please specify which type of quality of food. Chemical or physical?

(4) Line 80 and Figure 1B-D Authors need to put filling and emptying time data in the main figure rather than in the supplementary figure. Otherwise, please cite the relevant figures in the text(S1A-C).

(5) Line 84-85; Is "the mutant animals" indicating only nompC? Please specify it.

(6) Figure 1a: It is hard to determine the difference between the series of images. And also label filling and emptying under the time.

(7) S1E-H: It is unclear what "Time proportion of incomplete pump" means. Please define it.

(8) Please reorganize the figures to follow the order of the text, for example, figures 2 and 4

(9) Figure 4A. There is mislabelling in Figure 4A. It is supposed to be phalloidin not nc82.

(10) Figure 4K: It does not match the figure legend and main text.

(11) Figure 4D and G: Please indicate ATP application time point.

Thanks for your correction and all the points mentioned were revised.

Reviewer #4 (Recommendations For The Authors):

The figures need improvement. 1A has tiny circles showing pharynx and any differences are unclear.

The expression pattern of some of these drivers (Supplement) seems quite broad. The tmc nompC intersection image in Figure 1F is nice but the cibarium images are hard to interpret: does this one show muscle expression? What are "brain" motor neurons? Where are the labellar multi-dendritic neurons?

Tmc nompC intersection image show no expression in muscles. Somata of motor neurons 12 or 11 situated at SEZ area of brain, while somata of md-C neurons are in the cibarium. Image of md-L neurons was posted in response for reviewer #3 (Recommendations For The Authors):

Why do the assays alternate between swallowing food and swallowing water?

Thank for your suggestion, figure 1A has been zoomed-in. The Tmc nompC intersection image in Figure 2F displayed the position of md-C neurons in a ventral perspective, and muscles were not labelled. We stained muscles in cibarium by phalloidin and the image is illustrated in Figure 4A, while we didn’t find overlap between md-C neurons and muscles. Image of md-L neurons were posted as Author response image 4.

In the majority of our experiments, we employed water to test swallowing behavior, while we used methylcellulose water solution to test swallowing behavior of mechanoreceptor mutants, and sucrose solution for flies with md-C neurons expressing GCaMP since they hardly drank water when their head capsules were open.

How starved or water-deprived were the flies?

One day prior to the behavioral assays, flies were transferred to empty vials (without water or food) for 24 hours for water deprivation. Flies who could not survive 24h deprivation would be deprived for 12h.

How exactly was the pumping frequency (shown in Fig 1B) measured? There is no description in the methods at all. If the pump frequency is scored by changes in blue food intensity (arbitrary units?), this seems very subjective and maybe image angle dependent. What was camera frame rate? Can it capture this pumping speed adequately? Given the wealth of more quantitative methods for measuring food intake (eg. CAFE, flyPAD), it seems that better data could be obtained.

How was the total volume of the cibarium measured? What do the pie charts in Figure 3A represent?

The pump frequency was computed as the number of pumps divided by the time scale, following the methodology outlined in Manzo et al., 2012. Swallowing curves were plotted using the inverse of the blue food intensity in the cibarium. In this representation, ascending lines signify filling, while descending lines indicate emptying (see Figure 2D, 3B). We maintain objectivity in our approach since, during the recording of swallowing behavior, the fly was fixed, and we exclusively used data for analysis when the Region of Interest (ROI) was in the cibarium. This ensures that the intensity values accurately reflect the filling and emptying processes. Furthermore, we conducted manual frame-by-frame checks of pump frequency, and the results align with those generated by the time series analyzer V3 of ImageJ.

For the assessment of total volume of ingestion, we referred the methods of CAFE, utilizing a measurable glass capillary. We then calculated the ingestion rate (nL/s) by dividing the total volume of ingestion by the feeding time.

The changes seem small, in spite of the claim of statistical significance.

The observed stability in pump frequency within a given genotype underscores the significance of even seemingly small changes, which is statistically significant. We speculate that the stability in swallowing frequency suggests the existence of a redundant mechanism to ensure the robustness of the process. Disruption of one channel might potentially be partially compensated for by others, highlighting the vital nature of the swallowing mechanism.

How is this change in pump frequency consistent with defects in one aspect of the cycle - either ingestion (activation) or expulsion (inhibition)?

Please refer to Figure 2, 3. Both filling and emptying process were affects, while inhibition mainly influences emptying time (Figure 1—figure supplement 1).

for the authors:

Line 48: extensively

Line 62 - undiscovered.

Line 107, 463: multi

Line 124: What is "dysphagia?" This is an unusual word and should be defined.

Line 446: severe

Line 466: in the cibarium or not?

Thanks for your correction and all the places mentioned were revised.

Reviewer #3 (Public Review):

Strengths:<br /> There are many reports on the effect of chemical properties of foods on feeding in fruit flies, but only limited studies reported how physical properties of food affect feeding especially pharyngeal mechanosensory neurons. First, they found that mechanosensory mutants, including nompC, Tmc, and Piezo, showed impaired swallowing, mainly the emptying process. Next, they identified cibarium multidendritic mechanosensory neurons (md-C) are responsible for controlling swallowing by regulating motor neuron (MN) 12 and 11, which control filling and emptying, respectively.

Weaknesses:<br /> While the involvement of md-C and mechanosensory channels in controlling swallowing is convincing, it is not yet clear which stimuli activate md-C. Can it be an expansion of cibarium or food viscosity, or both? In addition, if rhythmic and coordinated contraction of muscles 11 and 12 is essential for swallowing, how can simultaneous activation of MN 11 and 12 by md-C achieve this? Finally, previous reports showed that food viscosity mainly affects the filling rather than the emptying process, which seems different from their finding.

eLife assessment

This valuable study investigates the role of mechanosensory feedback during swallowing in adult Drosophila. The authors provide convincing evidence that three mechanotransduction channel genes are required for ingestion rhythms and localize the role of these genes to a specific subpopulation of pharyngeal mechanosensory neurons. However, there is incomplete evidence to support the conclusions that these sensory neurons are necessary for swallowing, respond to stretch during swallowing, and connect to the motor neurons that control swallowing. This work may be of interest to neuroscientists interested in motor control of feeding behavior.

Reviewer #1 (Public Review):

Qin et al. set out to investigate the role of mechanosensory feedback during swallowing and identify neural circuits that generate ingestion rhythms. They use Drosophila melanogaster swallowing as a model system, focusing their study on the neural mechanisms that control cibarium filling and emptying in vivo. They find that pump frequency is decreased in mutants of three mechanotransduction genes (nompC, piezo, and Tmc), and conclude that mechanosensation mainly contributes to the emptying phase of swallowing. Furthermore, they find that double mutants of nompC and Tmc have more pronounced cibarium pumping defects than either single mutants or Tmc/piezo double mutants. They discovered that the expression patterns of nompC and Tmc overlap in two classes of neurons, md-C and md-L neurons. The dendrites of md-C neurons warp the cibarium and project their axons to the subesophageal zone of the brain. Silencing neurons that express both nompC and Tmc leads to severe ingestion defects, with decreased cibarium emptying. Optogenetic activation of the same population of neurons inhibited filling of the cibarium and accelerated cibarium emptying. In the brain, the axons of nompC∩Tmc cell types respond during ingestion of sugar but do not respond when the entire fly head is passively exposed to sucrose. Finally, the authors show that nompC∩Tmc cell types arborize close to the dendrites of motor neurons that are required for swallowing and that swallowing motor neurons respond to the activation of the entire Tmc-GAL4 pattern.

Strengths:<br /> -The authors rigorously quantify ingestion behavior to convincingly demonstrate the importance of mechanosensory genes in the control of swallowing rhythms and cibarium filling and emptying<br /> -The authors demonstrate that a small population of neurons that express both nompC and Tmc oppositely regulate cibarium emptying and filling when inhibited or activated, respectively<br /> -They provide evidence that the action of multiple mechanotransduction genes may converge in common cell types

Weaknesses:<br /> -A major weakness of the paper is that the authors use reagents that are expressed in both md-C and md-L but describe the results as though only md-C is manipulated<br /> -Evidence that the defects they see in pumping can be specifically attributed to md-C is based on severing the labellum and allowing md-L neurons to degrade.<br /> -GRASP is known to be non-specific and prone to false positives when neurons are in close proximity but not synaptically connected. A positive GRASP signal supports but does not confirm direct synaptic connectivity between md-C/md-L axons and MN11/MN12.<br /> -MN11/MN12 LexA lines are not included in the manuscript and their expression patterns (shared with the reviewers in the author response) do not appear to contain any motor neurons. Double labeling with previously described MN11 and MN12 motor neuron Gal4 lines is needed to support the claim that these LexA lines in fact label MN11 and MN12.<br /> -As seen in Figure Supplement 2, the expression pattern of Tmc-GAL4 is broader than md-C alone. Therefore, the functional connectivity the authors observe between Tmc expressing neurons and MN11 and 12 cannot be traced to md-C alone<br /> -Example traces of md-C calcium imaging during ingestion in vivo are not included, and evidence that md-C neurons respond to mechanical force is lacking<br /> -A positive control (perhaps demonstrating that sugar sensory neurons respond to sucrose in this preparation) is needed to assess whether the lack of response to sucrose ex vivo in Figure 4K is informative<br /> -Proximity between md-C neurons and muscles is not evidence that they sense stretch<br /> -Reporting of posthoc tests needs to be improved throughout the manuscript, as it is not clear which comparisons are noted with asterisks in the figures.

Overall, this work convincingly shows that swallowing and swallowing rhythms are dependent on several mechanosensory genes. Qin et al. also characterize a candidate neuron, md-C, that is likely to provide mechanosensory feedback to pumping motor neurons, but the results they present here are not sufficient to assign this function to md-C alone. This work will have a positive impact on the field by demonstrating the importance of mechanosensory feedback to swallowing rhythms and providing a potential entry point for future investigation of the identity and mechanisms of swallowing central pattern generators.

Author Response

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

Reply to comments:

(1) It was not clear why the phylogenetic analysis included non-validated GPCRs that clustered with the validated peptidergic receptors. Would restricting the phylogenetic analyses only to confirmed peptidergic GPCRs alter the topology of the tree and subsequent conclusions of independent expansion?

Thank you for this comment. In general, phylogenetic analyses become more robust if a larger diversity and fuller complement of sequences are included. With very sparse sampling, sequences that are homologous but not orthologous may be misleadingly grouped together, because intermediate sequences have been left out. For tree building, we thus did not want to focus only on experimentally validated receptors but also on all receptors that are phylogenetically related to the validated receptors. Only this approach can ensure a comprehensive exploration of the relationship of peptidergic receptors. The broader phylogenetic approach was also essential to identify orthologs to the experimentally validated Nematostella receptors across other cnidarian species.

(2) Clearly, other neuropeptide signaling systems in cnidarians remain to be discovered but this paper represents a huge step forward.

We appreciate this assessment of the paper. We agree that many systems remain to be discovered. Our paper will also help with the identification of further receptors both in Nematostella as well as other cnidarian species. Please note that we have made specific receptor-ligand predictions for several cnidarian species based on our phylogenetic analysis. Our phylogenies could also help prioritize the study of the remaining orphan Nematostella GPCRs.

(3) There are limitations in what can be interpreted from single cell transcriptomic data but the data nevertheless provide the foundations for future studies involving i). detailed anatomical analysis of neuropeptide and neuropeptide receptor expression in N. vectensis using mRNA in situ hybridization and/or immunohistochemical methods and ii). functional analysis of the physiological/behavioral roles of neuropeptide signaling systems in N. vectensis

We fully agree with this comment. The analysis of the available single-cell sequence resources clearly represents only the first step of anatomical and functional analyses. Our aim was to place the identified peptide-receptor interactions into a whole-organism context with cell type resolution, to highlight the potential complexity of peptidergic signaling in this organism and to facilitate the exploration and conceptualisation of our biochemical screen.

Comments to authors

(1) In future, when preparing manuscripts, please use page and line numbers; it makes the task below for reviewers much easier!

We appreciate the suggestion and will do this for future manuscripts.

(2) In the abstract the term "extensively wired" is used. In the context of neuropeptide mediated volume transmission this may not be an appropriate term to use because use of the word "wired" is likely to be associated with point-to-point type classical synaptic transmission; "extensively connected" would be better.

Thank you for this comment. We have changed the text in the abstract to “extensively connected”.

(3) Introduction: Please change "seven-transmembrane proteins and show a slower evolutionary rate than proneuropeptide..." to "seven-transmembrane proteins that show a slower evolutionary rate than proneuropeptide..."

Changed.

(4) Under the section "Creation of a Nematostella neuropeptide library, what is meant by "our regular expressions"? This needs to be rephrased to make it clearer what is meant.

We have now rephrased the relevant sentence to make our approach clearer.

“This predicted secretome was filtered with regular expressions to detect sequences with the repetitive dibasic cleavage sites (K and R in any combination) and amidation sites, using a custom script from a previous publication (Thiel et al., 2021).”

and later:

“Based on the MS data, we included the additional, non-dibasic N-terminal cleavage sites into our script that uses regular expressions to search for repetitive cleavage sites (Thiel et al., 2024) and re-screened the predicted secretome.”

(5) Under the section "Creation of a Nematostella neuropeptide library" the phrase "differ in the length of their N-terminus" needs to be changed to "differ in the length of their N-terminal region". The N-terminus is, as its name implies, one end of the peptide/protein so it can't have a length as such.

Changed.

(6) Under the section "Analysis of metazoan class A GPCRs and selection of N. vectensis neuropeptide-receptor candidates",

Change:

"For a more detailed analysis, we then reduced our sampled species to the cnidarian, the bilaterian with experimentally confirmed GPCRs and Petromyzon marinus, and the two placozoan species (Figure 2B)."

To

"For a more detailed analysis, we then reduced our sampled species to cnidarians, bilaterians with experimentally confirmed GPCRs and Petromyzon marinus, and two placozoan species (Figure 2B)."

Changed.

(7) Under the section "Analysis of metazoan class A GPCRs and selection of N. vectensis neuropeptide-receptor candidates" - change "We re-run" to "We re-ran"

Changed.

(8) Throughout the paper reference is made to a variety of neuropeptides that have or are predicted to have an N-terminal pyroglutmate. However, these are referred to without indicating this post-translational modification e.g. QGRFamide.

This should be corrected throughout the paper, in the text, and figures. Two abbreviations for pyroglutamate are used in the literature:

pQ, which shows that the encoded amino acid is Q (Glutamine)

pE, which shows that the post-translationally modified amino-acid is glutamate (E)

In the neuropeptide field, pQ seems to be more widely used than pE, so our recommendation would be to use pQ.

In the revised version we now write pyroQ whenever we refer to the actual peptide. We now only use the peptide name without indicating this modification when we refer to the precursor of these peptides.

(9) The title for Figure 5 is rather short and vague. A title like "Tissue-specific expression of neuropeptide precursors and receptors in Nematostella" seems more appropriate

We appreciate the reviewer's input, and we have made the change accordingly. The revised figure legend now reads: “Tissue-specific expression of neuropeptide precursors and receptors (GPCRs) in N. vectensis.”

(10) All of the figures in the paper have been saved in bitmap format (e.g. tiff), which means that the resolution of the figures may end up being poor in the published article. All of the figures in this paper should be saved in vector format (e.g. eps) so that there is no loss of resolution when the size of the file/figure is reduced.

We have now uploaded all figures in vector format (.eps or .pdf) to prevent any loss of resolution.

(11) In Figure 3 - supplement 2 - the neuropeptides are referred to here as PRGamides and GPRGamides. Some consistency is needed here. And in Figure B, the G of one of the GPRGamides is not shown in black.

Thank you for spotting this mistake. We now give the correct peptide sequence in parenthesis as "GPRGamide". We also highlighted the missing GPRGamide in the figure.

eLife assessment

This work identifies cnidarian neuropeptides and pairs them to their GPCR, then shows that neuropeptide signaling systems have evolved and diversified independently in cnidarians and bilaterians. Neuropeptide-receptor partners were experimentally identified using established and widely used methodologies including single cell mapping, providing compelling evidence for the conclusions of the paper. This impressive accomplishment provides fundamental new insights into the evolution of neuropeptide signaling systems and will be of broad interest to neurobiologists and evolution of development researchers.

Joint Public Review:

Neuropeptide signaling is an important component of nervous systems, where neuropeptides typically act via G-protein coupled receptors (GPCRs) to regulate many physiological and behavioral processes. Neuropeptides and their cognate GPCRs have been extensively characterized in bilaterian animals, revealing that a core set of neuropeptide signaling systems originated in common ancestors of extant Bilateria. Neuropeptides have also been identified in cnidarians, which are a sister group to the Bilateria. However, the GPCRs that mediate the effects of neuropeptides in cnidarians have not been identified.

In this paper the authors perform a phylogenetic analysis of GPCRs in metazoans and report that the orthologs of bilaterian neuropeptide receptors are not found in cnidarians. This indicates that neuropeptide signaling systems have largely evolved independently in cnidarians and bilaterians. To accomplish this, they generated a library of putative and known neuropeptides computationally identified in the genome of the cnidarian sea anemone Nematostella vectensis. These peptides were systematically screened for their ability to activate any of the 161 putative Nematostella GPCRs.

This work identified 31 validated GPCRs. These, together with GPCRs that cluster with them, were then used to demonstrate the independent expansion of GPCRs in cnidarian and bilaterian lineages. The authors then mapped validated receptors and ligands to the Nematostella single cell data to provide an overview of the cell types expressing these signaling genes. In addition, the authors have begun to analyze neuropeptide signaling networks in N. vectensis by showing potential signaling connections between cell types expressing neuropeptides and cell types expressing cognate receptors.

This work is the most extensive pharmacological characterization of neuropeptide GPCRs in a cnidarian to date and thus represents an important accomplishment, and is one that will improve our understanding of how peptidergic signaling evolved in animals and its impact on evolution of nervous systems. In addition, this impressive work transforms our knowledge of neuropeptide signaling systems in cnidarians and provides the foundations for extensive functional characterization neuropeptide systems in the context of nervous systems that exhibit radial symmetry, contrasting with the bilaterally symmetrical architecture of the majority of bilaterian nervous systems.

The reviewers did not detect any weaknesses in the work but asked that the authors comment on the following points, which they have done in the revised version.

(1) Clearly, other neuropeptide signaling systems in cnidarians remain to be discovered but this paper represents a huge step forward.

(2) There are limitations in what can be interpreted from single cell transcriptomic data but the data nevertheless provide the foundations for future studies involving i). detailed anatomical analysis of neuropeptide and neuropeptide receptor expression in N. vectensis using mRNA in situ hybridization and/or immunohistochemical methods and ii). functional analysis of the physiological/behavioral roles of neuropeptide signaling systems in N. vectensis.

Author Response

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

Reviewer #1 (Public Review):

Summary:

In the manuscript by Chen et al. entitled, "The retina uncouples glycolysis and oxidative phosphorylation via Cori-, Cahill-, and mini-Krebs-cycle", the authors look to provide insight on retinal metabolism and substrate utilization by using a murine explant model with various pharmacological treatments in conjunction with metabolomics. The authors conclude that photoreceptors, a specific cell within the explant, which also includes retinal pigment epithelium (RPE) and many other types of cells, are able to uncouple glycolytic and Krebs-cycle metabolism via three different pathways: 1) the mini-Krebs-cycle, fueled by glutamine and branched-chain amino acids; 2) the alanine-generating Cahill-cycle; and 3) the lactate-releasing Cori-cycle. While intriguing if determined to be true, these cell-specific conclusions are called into question due to the ex vivo experimental setup with the inclusion of RPE, the fact that the treatments were not cell-specific nor targeted at an enzyme specific to a certain cell within the retina, and no stable isotope tracing nor mitochondrial function assays were performed. Hence, without significant cell-specific methods and future experimentation, the primary claims are not supported.

Strengths:

This study attempts to improve on the issues that have limited the results obtained from previous ex vivo retinal explant studies by culturing in the presence of the RPE, which is a major player in the outer retinal metabolic microenvironment. Additionally, the study utilizes multiple pharmacologic methods to define retinal metabolism and substrate utilization.

Weaknesses:

A major weakness of this study is the lack of in vivo supporting data. Explant cultures remove the retina from its dual blood supply. Typically, retinal explant cultures are done without RPE. However, the authors included RPE in the majority of experimental conditions herein. However, it is unclear if the metabolomics samples included the RPE or not. The inclusion of the RPE, which is metabolically active and can be altered by the treatments investigated herein, further confounds the claims made regarding the neuroretina. Considering the pharmacologic treatments utilized with the explant cultures are not cell-specific and/or have significant off-target effects, it is difficult to ascertain that the metabolic changes are secondary to the effects on photoreceptors alone, which the authors claim. Additionally, the explants are taken at a very early age when photoreceptors are known to still be maturing. No mention or data is presented on how these metabolic changes are altered in retinal explants after photoreceptors have fully matured. Likewise, significant assumptions are made based on a single metabolomics experiment with no stable isotope tracing to support the pathways suggested. While the authors use immunofluorescence to support their claims at multiple points, demonstrating the presence of certain enzymes in the photoreceptors, many of these enzymes are present throughout the retina and likely the RPE. Finally, the claims presented here are in direction contradiction to recent in vivo studies that used cell-specific methods when examining retinal metabolism. No discussion of this difference in results is attempted. Response: We agree with the reviewer that in vivo studies could be very interesting indeed. However, technologically it will be extremely difficult to (repeatedly/continuously) sample the retina of an experimental animal and to combine this with an interventional study, with a subsequent metabolomic analysis. We do not currently have access to such technology nor are we aware of any other lab in the world capable of doing such studies. Moreover, virtually all prior studies on retinal metabolism have been done on explanted retina without RPE. This includes the seminal studies by Otto Warburg in the 1920s. As opposed to this, our retinal samples for also all the metabolomic analyses included the RPE, except for the no RPE condition that was used as a comparator for the earlier investigations.

We note that our metabolomic analysis was done for all five experimental conditions where each condition included at least five independent samples (each derived from different animals).

The reviewer is correct to say that our organotypic explant cultures are early post-natal, with explantation performed at post-natal day 9 and culturing until day 15. Since our retinal explant system has been validated extremely well over more than three decades of pertinent research (see for instance: Caffe et al., Curr Eye Res. 8:1083-92, 1989), we are confident that photoreceptors mature in vitro in ways that are very similar to the in vivo situation. As far as studies in adult retina (i.e. three months or older) are concerned, this is indeed an important question that will be addressed in future studies. Studies employing stable isotope labelling may also be very informative and are planned for the future, also in order to properly determine fluxes. This will likely require an extension to our NMR hardware with an 15N channel probe, something that we plan on implementing in the future.

We are aware that a number of questions relating to retinal metabolism are controversial and that the use of other methodology or experimental systems may lead to alternative interpretations. We have now included citations of other studies that use, for example, conditional and/or inducible knock-outs or in vivo blood sampling (e.g. Wang et al., IOVS 38:48-55, 1997; Yu et al., Invest Ophthalmol Vis Sci. 46:4728-33, 2005; Swarup et al., Am J Physiol Cell Physiol. 316:C121-C133, 2019; Daniele et al., FASEB Journal 36:e22428, 2022) and discuss the pros and cons of such approaches (e.g. in Lines 376-384; 454-472).

Reviewer #2 (Public Review):

Summary:

The authors aim to learn about retinal cell-specific metabolic pathways, which could substantially improve the way retinal diseases are understood and treated. They culture ex vivo mouse retinas for 6 days with 2 - 4 days of various drug treatments targeting different metabolic pathways or by removing the RPE/choroid tissue from the neural retina. They then look at photoreceptor survival, stain for various metabolic enzymes, and quantify a broad panel of metabolites. While this is an important question to address, the results are not sufficient to support the conclusions.

Strengths:

The questions the authors are exploring at extremely valuable and I commend the authors and working to learn more about retina metabolism. The different sensitivity of the cones to various drugs is interesting and may suggest key differences between rods and cones. The authors also provide a thoughtful discussion of various metabolic pathways in the context of previous publications.

Weaknesses:

As the authors point out, ex vivo culture models allow for control over multiple aspects of the environment (such as drug delivery) not available in vivo. Ex vivo cultures can provide good hints as to what pathways are available between interacting tissues. However, there are many limitations to ex vivo cultures, including shifting to a very artificial culture media condition that is extremely different than the native environment of the retina. It is well appreciated that cells have flexible metabolism and will adapt to the conditions provided. Therefore, observations of metabolic responses obtained under culture conditions need to be interpreted with caution, they indicate what the tissue is doing under those specific conditions (which include cells adapting and dying).

Chen et al use pharmacological interventions to the impact of various metabolic pathways on photoreceptor survival and "long term" metabolic changes. The dose and timing of these drug treatments are not examined though. It is also hard to know how these drugs penetrate the tissue and it needs to be validated that the intended targets are being accurately hit. These relatively long-term treatments should be causing numerous downstream changes to metabolism, cell function, and survival, which makes looking at a snapshot of metabolite levels hard to interpret. It would be more valuable to look at multiple time points after drug treatment, especially easy time points (closer to 1 hr). The authors use metabolite ratios to make conclusions about pathway activity. It would be more valuable to directly measure pathway activity by looking a metabolite production rates in the media and/or with metabolic tracers again in time scales closer to minutes and hours instead of days.

It is not clear from the text if the ex vivo samples with RPE/choroid intact are analyzed for metabolomics with the RPE/choroid still intact or if this is removed. If it is not removed, the comparison to the retina without RPE/choroid needs to be re-interpreted for the contribution of metabolites from the added tissue. The composition of the tissue is different and cannot be disentangled from the changes to the neural retina specifically.

While the data is interesting and may give insights into some rod and cone-specific metabolic susceptibility, more work is needed to validate these conclusions. Given the limitations of the model the authors have over-interpreted their findings and the conclusions are not supported by the results. They need to either dramatically limit the scope of their conclusions or validate these hypotheses with additional models and tools.

Response: We thank the reviewer for the insightful comments and agree that some of our interpretations may have been phrased too determinedly. We have therefore rephrased and toned down our conclusions in many instances in the text, and changed the manuscript title to now read “Retinal metabolism: Evidence for uncoupling of glycolysis and oxidative phosphorylation via Cori-, Cahill-, and mini-Krebs-cycle”.

Nevertheless, when considering the major known metabolic pathways and their possible impact on metabolite patterns after the experimental manipulations used here, we believe our interpretations to be consistent with the data obtained. Conversely, the previously suggested retinal aerobic glycolysis cannot explain most of the data we have obtained. Even further, also a predominant use of the classical “full” Krebs-cycle/OXPHOS would not explain the metabolite patterns found (e.g. alanine, N-acetylaspartate (NAA)). While this does not in itself mean that our interpretations are all correct, they seem plausible in view of the data at hand and will hopefully stimulate further research on retinal energy metabolism using complementary technologies that were not available to us for the purpose of this study.

We comment that our organotypic retinal explant cultures, while they do contain their very own, native RPE, do not comprise the choroidal vasculature (in our explantation procedure the RPE readily detaches from the choroid).

As far as the drugs used on retinal explants are concerned, we note that:

(1) all three compounds used are extremely well validated, with literally thousands of studies and decades of research to their credit (i.e., 1,9-dideoxyforskolin: >270 publications since 1984; Shikonin: >1000 publications since 1977; FCCP: >2800 publications since 1967),

(2) all experimental conditions show clear and differential drug effects, as shown, for instance, by the principal component analysis in Figure1I and the cluster analysis in Figure2A,

(3) the response patterns observed for key metabolites match the anticipated drug effects (e.g. decreased glucose consumption with 1,9-dideoxyforskolin; decreased lactate levels with Shikonin; lactate accumulation with FCCP).

One can therefore be reasonably certain that these drugs did penetrate the explanted retina and that their respective drug targets were hit. Assessing dose-responses would certainly be interesting, however, the aim of this initial study was not pharmacodynamics but a general manipulation of energy metabolism. Moreover, given the extensive validation of these drugs, off-target effects seem not very likely at the concentrations used.

We agree with the reviewer that using a longitudinal, time-series type of analysis could give additional insights. We note that each additional time-point will require retinae from 25 animals and a very resource-intensive and time-consuming metabolomic analysis, together with a significantly more complex multivariate analysis (metabolite, experimental condition, time). This is a completely new undertaking that is simply not feasible as an extension of the present study.

To look at pathway activity in more direct ways is very good idea, to this end we aim to implement in the future an idea put forward by the reviewers, namely 13C-labeling and additionally 15N-labeling and tracing for specific metabolic fuels (e.g. glucose, lactate and anaplerotic amino acids such as glutamate and branched chain amino acids).

The reviewer is of course correct to say that the culture condition is somewhat artificial and that this may have introduced changes in the metabolism. However, as noted above in the first response to reviewer #1, the organotypic retinal culture system, using a defined medium, free of serum and antibiotics, has been extremely well studied and validated for decades (cf. Caffé et al., Curr Eye Res. 8:1083-92, 1989). Importantly, this system allows to maintain retinal viability, histotypic organization, and function over many weeks in culture. Moreover, most previous studies on retinal metabolism have also used explanted retina – acute or cultured – i.e. experimental approaches that are similar to what we have used and that may be liable to their own artefactual changes in metabolism. This includes the seminal, 1920s studies by Otto Warburg, or the 1980s studies by Barry Winkler, the results of which the reviewers do not seem to doubt.

We further agree that studying retinal metabolism in a situation closer to in vivo conditions would be thrilling, however to our knowledge to date there is no retina model that fully mimics the complex interplay of the blood metabolome with metabolic tissue activity. This likely means that for each metabolic condition to study (e.g. hyperglycemia, cachexia, etc.), a fairly large number of animals will need to be sacrificed for the molecular investigation of ex vivo retinal biopsies, which would mean a tremendous animal burden.

We hope the reviewer will appreciate that the revised manuscript now includes numerous improvements, along with new, additional datasets and figures, references to further relevant literature, and – as mentioned above – a more cautious phrasing of our interpretations and conclusions, including a more careful wording for the manuscript title.

Reviewer #3 (Public Review):

Summary:

The neural retina is one of the most energetically active tissues in the body and research into retinal metabolism has a rich history. Prevailing dogma in the field is that the photoreceptors of the neural retina (rods and cones) are heavily reliant on glycolysis, and as oxygen tension at the level of photoreceptors is very low, these specialized sensory neurons carry out aerobic glycolysis, akin to the Warburg effect in cancer cells. It has been found that this unique metabolism changes in many retinal diseases, and targeting retinal metabolism may be a viable treatment strategy. The neural retina is composed of 11 different cell types, and many research groups over the past century have contributed to our current understanding of cell-specific metabolism of retinal cells. More recently, it has been shown in mouse models and co-culture of the mouse neural retina with human RPE cultures that photoreceptors are reliant on the underlying retinal pigment epithelium for supplying nutrients. Chen and colleagues add to this body of work by studying an ex vivo culture of the developing mouse retina that maintained contact with the retinal pigment epithelium. They exposed such ex vivo cultures to small molecule inhibitors of specific metabolic pathways, performing targeted metabolomics on the tissue and staining the tissue with key metabolic enzymes to lay the groundwork for what metabolic pathways may be active in particular cell types of the retina. The authors conclude that rod and cone photoreceptors are reliant on different metabolic pathways to maintain their cell viability - in particular, that rods rely on oxidative phosphorylation and cones rely on glycolysis. Further, their data support multiple mechanisms whereby glycolysis may occur simultaneously with anapleurosis to provide abundant energy to photoreceptors. The data from metabolomics revealed several novel findings in retinal metabolism, including the use of glutamine to fuel the mini-Krebs cycle, the utilization of the Cahill cycle in photoreceptors, and a taurine/hypotaurine shuttle between the underlying retinal pigment epithelium and photoreceptors to transfer reducing equivalents from the RPE to photoreceptors. In addition, this study provides robust quantitative metabolomics datasets that can be compared across experiments and groups. The use of this platform will allow for rapid testing of novel hypotheses regarding the metabolic ecosystem in the neural retina.

Strengths:

The data on differences in the susceptibility of rods and cones to mitochondrial dysfunction versus glycolysis provides novel hypothesis-generating conjectures that can be tested in animal models. The multiple mechanisms that allow anapleurosis and glycolysis to run side-by-side add significant novelty to the field of retinal metabolism, setting the stage for further testing of these hypotheses as well.

Weaknesses:

Almost all of the conclusions from the paper are preliminary, based on data showing enzymes necessary for a metabolic process are present and the metabolites for that process are also present. However, to truly prove whether these processes are happening, C13 labeling or knock-out or over-expression experiments are necessary. Further, while there is good data that RPE cultures in vitro strongly recapitulate RPE phenotypes in vivo, ex vivo neural retina cultures undergo rapid death. Thus, conclusions about metabolism from explants should either be well correlated with existing literature or lead to targeted in vivo studies. This paper currently lacks both.

Response: As mentioned above in the first answers to reviewers #1 and #2, we think of our study as a starting point that may provide novel directions for a whole series of investigations into retinal energy metabolism. Especially the use of novel technologies may in the future allow to decipher the different metabolic phenotypes of the 100+ distinct retinal cell types by in situ spatial metabolomics and lipidomics. Currently, we still have to limit the scope of our studies to only certain aspects of this topic. We thus agree that some of our interpretations need to be formulated more carefully and we have done so in the revised version of our manuscript. We also agree with the reviewer that carbon (13C) labelling and tracing studies will be very informative and will engage in such studies in the future. Besides 13C, we aim to further employ 15N labelled substrates, which is especially suitable to study the destiny of amino acids.

As far as our organotypic retinal explant system is concerned, it is arguably one of the best validated such systems available (see responses to reviewers #1 and #2). While the reviewer is correct to say that the neuroretina without RPE degenerates relatively quickly in vitro, in our system, with the neuroretina and its native RPE cultured together, we can routinely culture the retina for four weeks or more, without major cell loss (Söderpalm et al., IOVS 35:3910-21, 1994; Belhadj et al., JoVE 165, 2020). Thus, our retinal cultures with RPE do not undergo rapid death. Within the time-frame of the present study (6 days in vitro) culturing-induced cell death is minimal and unlikely to influence our analyses. For further, more detailed answers to the reviewers’ questions please see our detailed point-to-point response below.

We agree with the reviewer that eventually in vivo studies will be important to confirm our interpretations. As mentioned in our initial response to reviewer #1, such studies will be very challenging and new technologies may need to be developed before in vivo investigations can deliver the answers to the questions at hand (see answer to question Rev#3.17 below), especially if the cross-play between substrate availability from the blood metabolome and the retinal metabolic pathway activity shall be studied.

Recommendations For The Authors

Reviewer #1 (Recommendations For The Authors):

Rev#1.1. The animals should be screened for and lack rd8.

Response: This is a pertinent question from the reviewer. Ever since we first became aware of the presence of rd8 mutations in certain mouse lines from major vendors (e.g. Charles River, Jackson Labs) in around 2010, we have setup regular screening of all our mouse lines for this Crb1 mutation. Accordingly, the mouse lines used in this study were confirmed to be free of the rd8 / Crb1 mutation. A corresponding remark has now been inserted into the SI materials and methods section (Lines 37-38).

Rev#1.2. GLUT1 looks significantly different from in vivo to in vitro. Recommend co-staining with RHO and cone markers (PNA or CAR) to further delineate where it is being expressed. The in vitro cultures appear to have much shorter outer segments (OS). Considering OS biosynthesis is thought to drive a good deal of metabolic adaptations, how relevant is the in vitro model system to what is truly occurring in vivo?

Response: The GLUT1 staining shown in Figure 1 displays the in vivo situation. Since may not have been entirely clear from the previous figure legend, we have now labelled this as “in vivo retina” and distinguish it from “in vitro” samples in the legend to Figure 1 (Lines 774-778). As far as the comparison of GLUT1 staining in vivo (Figure 1A3) vs. in vitro (Figure S1C3) is concerned, in both situations a strong RPE labelling is clearly visible, with essentially no GLUT1 label within the neuroretina.

Nevertheless, to better delineate the expression of GLUT1 in the outer retina, we have now performed an additional co-staining with rhodopsin (RHO) as rod marker and peanut agglutinin (PNA) as cone marker, as suggested by the reviewer (new supplemental Figure S1). In brief, this co-staining confirms the strong expression of GLUT1 in the RPE, while there is essentially no GLUT1 detectable in rod or cone photoreceptors.

Retinal explants in long-term cultures do indeed have somewhat shorter outer segments compared to same age in vivo counterparts (Caffe et al., Curr Eye Res. 8:1083-1092, 1989). However, in the short-term cultures (6 DIV) and at the age studied here (P15) outer segments have only just started to grow out and are around 10 - 12 µm long, both in vitro and in vivo (cf. LaVail, JCB 58:650-661, 1978). Thus, the metabolism required for outer segment synthesis should be equivalent when in vitro and in vivo situations are compared. For considerations on outer segments in retinal explant cultures see also Rev#3.2 and Rev#3.29.

Rev#1.3. Also, recent publications have shown that GLUT1 is expressed in the neuroretina including rods, cones, and muller glia. Was GLUT1 not appreciated in these cells in your ex vivo samples and if so, why? Likewise, these same studies previously demonstrated GLUT1 resulted in rod degeneration but not cone. The results presented here differ significantly. Why the difference in results and is it secondary to the in vitro vs. in vivo setting? Furthermore, the authors state that they thought the no RPE situation would be similar to the GLUT1 inhibitor experimental condition but instead, they were vastly different. Is this secondary to the fact that GLUT1 is expressed outside the RPE.

Response: We are aware that there is a controversy regarding GLUT1 expression in the neuroretina, please see also our response to question Rev#3.1 below. As far as our immunostaining for GLUT1 on in vivo retina is concerned, we find an unambiguous and very marked expression of GLUT1 in RPE cells, at both basal and apical sides. Compared to the RPE, the neuroretina appears devoid of GLUT1 staining. However, at very high gamma values a faint staining in the neuroretina becomes visible, a staining which from its appearance – processes spanning the entire width of the retina – is most compatible with Müller glia cells. Under normal circumstances we would have dismissed such a faint staining as background and false positive. Given the sometimes very contradicting reports in the literature, we cannot fully exclude a weak expression of GLUT1 also in cells other than the RPE, with Müller glial cells perhaps being the most likely candidate. At any rate, GLUT1 expression in the neuroretina can only be much weaker than in the RPE, making its relevance for overall retinal metabolism unclear.

As far as recent publications studying GLUT1 in the retina are concerned, we know of the study by Daniele et al. (FASEB Journal 36:e22428, 2022), which used a rod-specific, conditional knock-out of GLUT1 and found a relatively slow rod degeneration. We are not aware of a selective GLUT1 knock-out in cones, nor are we aware of conditional GLUT3 knock-outs in the retina. For further discussion of the Daniele et al. study please see Rev#3.13.

The reviewer is right, initially we were thinking that, since GLUT1 was expressed only (predominantly) in RPE, the metabolic response to GLUT1 inhibition should look similar to the no RPE situation. However, this initial hypothesis did not consider a key fact: The RPE builds the blood retinal barrier and the tight-junction coupled RPE cells are a barrier to any larger molecule, including glucose. Removing the barrier by removing the RPE dramatically increases the availability of glucose to the retina, a phenomenon that is likely exacerbated by the expression of the high affinity/high capacity GLUT3 on photoreceptors (cf. Figure S1A). In other words, when the RPE is removed the outer retina is “flooded” with glucose and we believe that this is probably the main factor that explains why the metabolic response to GLUT1 inhibition (1,9-DDF group) is so different from the no RPE condition.

We have now included an additional corresponding explanation in the discussion (Lines 422-429). Furthermore, we have added an entire new subchapter to the discussion to debate the expression of glucose transporters in the outer retina (Lines 454-472).

Rev#1.4. Shikonin's mechanism of action via protein aggregation and lack of specificity for PKM2 vs PKM1 at 4uM is an experimental limitation that needs to be taken into account. All treatments utilized are not cell-specific.

Response: While the reviewer is correct to say that Shikonin may have multiple cellular targets and a diverse range of possible applications as an anti-inflammatory, antimicrobial, or anticancer agent (cf. Guo et al., Pharmacol. Res. 149:104463, 2019), numerous studies support its specificity for PKM2 over PKM1, at concentrations ranging from 1 – 10 µM (Chen et al., Oncogene 30:4297-306, 2011; Zhao et al., Sci. Rep. 8:14517, 2018; Traxler et al., Cell Metab. 34:1248-1263, 2022). We settled for 4 μM as an intermediate concentration, considering its effectiveness and specificity in previous studies. We have now inserted references detailing the specificity and concentration range of Shikonin into the SI Materials and Methods section (Line 62).

The concern that “all treatments” are not cell-specific is debatable. Certainly, any given compound may have off-target effects, yet, since the compounds we used in our study have all been studied for decades (see above, initial response to Reviewer #2), their off-target profile is well established and unlikely to play an important role here. Moreover, in our study the cell specificity does not come from the compounds used but from where their targets are expressed. As shown in Figure 1A and in Figure S1C, Shikonin´s target PKM2 is almost exclusively expressed in photoreceptor inner segments. Hence, it seems very reasonable to expect that the vast majority of the metabolomic changes observed by Shikonin treatment are related to photoreceptors. We note that this assertion would still be true even if there was a low-level expression of PKM2 in other retinal cell types and/or if Shikonin had moderate off-target effects on other enzymes since the bulk of the effect on the quantitative metabolomic dataset would still originate from PKM2 inhibition in photoreceptors.

Rev#1.5. What was the method of cone counting in Figure 1?

Response: Cones were counted per 100 µm of retinal circumference based on an arrestin-3 staining (cone arrestin, CAR).

This information is now included in the SI Materials and Methods section under “Microscopy, cell counting, and statistical analysis” (Lines 99-100).

Rev#1.6. How do you know that FCCP is not altering RPE ox phos, disrupting the outer retinal microenvironment and leading to cell death, and therefore, the effects seen are not photoreceptor-specific but rather downstream from the initial insult in RPE?

Response: We propose that FCCP will be acting on both photoreceptors and RPE cells (and all other retinal cell types) at essentially the same time, over the experimental time-frame. Thus, OXPHOS should be inhibited in all cells simultaneously. However, FCCP will primarily affect cells that actually use OXPHOS to a large extent, while cells relying on other metabolic pathways (e.g. glycolysis) will hardly be affected.

We believe the very strong effect of FCCP, seen exclusively in rod photoreceptors, to be a direct drug effect. While we cannot not fully exclude an indirect effect via the RPE – as proposed by the reviewer – we think this to be unlikely because:

(1) RPE viability was not compromised by FCCP treatment.

(2) If the reviewer´s hypothesis was correct, then also cone photoreceptors should have been affected (e.g. because now the RPE consumes all glucose, leaving nothing for cones). However, cones were essentially unaffected by the FCCP treatment, making a dependence on RPE OXPHOS unlikely. Especially so, because blocking GLUT1 and glucose import on the RPE with 1,9-DDF had only relatively minor effects on rod photoreceptor viability but strongly affected cones. This indicates that the RPE is mainly shuttling glucose through to photoreceptors, especially to cones, and this function does not seem to be impaired by FCCP treatment.

(3) We found that enzymes required for Krebs-cycle and OXPHOS activity (i.e. citrate synthase, fumarase, ATP synthase γ) are predominantly expressed in photoreceptors but virtually absent from RPE (Figure 3D, see also answer to following question).

(4) The density of mitochondria (i.e. the target for FCCP) is far lower in RPE than in photoreceptors, as evidenced also by the COX staining shown in Figure 1A. Hence, photoreceptors are far more likely to be hit by FCCP treatment than RPE cells.

To accommodate the reviewer´s concern, we have now added a further comment into the discussion (Lines 440-442).

Rev#1.7. While Figure 3D is interesting, it offers no significant insight into mechanisms as the enzyme levels are not being compared to control nor is mitochondrial fitness in these conditions being assessed, which would provide greater insight than just showing that these enzymes are present in the inner segments, which are known to be rich in mitochondria. Additionally, stating that the low ATP is secondary to decreased Krebs cycle activity and ox phos based on merely ATP levels is not supported by metabolite levels minus citrate nor ox phos enzyme levels or oxygen consumption. Also, citrate is purported to be decreased in the table in Figure 2 in the no RPE condition; however, Supplemental Figure 2 demonstrates this change is not significant then the same data is presented in Supplemental Figure 3 and it is statistically significant again. Why the difference in data and why is the same data being shown multiple times?

Response: The immunostaining shown in Figure 3D shows the in vivo retina, or in other words the localization of enzymes in the native situation. Since this may not have been obvious in the previous manuscript version, we have added a corresponding comment to the legend of Figure 3 (Line 806). The localization of the Krebs-cycle/OXPHOS enzymes citrate synthase, fumarase, and ATP synthase mainly to photoreceptors, but not (or much less) to RPE, is another piece of evidence supporting the idea that OXPHOS is predominantly performed by photoreceptors (see also answer to previous question Rev#1.6).

The decreased ATP levels (together with citrate, aspartate, NAA) shown in Figure 3 in the no RPE group, are an indication that photoreceptor Krebs-cycle activity may be decreased but not abolished in the absence of RPE. Importantly, GTP levels are not reduced in the no RPE group (Figure 2). Since large amounts of GTP can only by synthesized by either SUCLG-1 in the Krebs-cycle or by NDK-mediated exchange with ATP, the most plausible interpretation is that Krebs-cycle dependent ATP-synthesis was decreased in the no RPE situation, but that the (mini) Krebs-cycle or Cahill-cycle, notably the step from succinyl-CoA to succinate, was running. Since there is no RPE in this group, this strongly suggests important Krebs-cycle/OXPHOS activity in photoreceptors where the majority of the corresponding enzymes are located (see above).

We thank the reviewer for pointing out that the information on group comparisons may not have been presented with sufficient clarity. In the figures mentioned by the reviewer the data is shown and compared in different contexts: the table in Figure 2B and the data in Figure S3 (now renumbered to Figure S5) refer to two-way comparisons of treatment condition to control, to elucidate individual treatment effects. Meanwhile Figure S2 (now supplementary Figure S3) refers to a 5-way comparison for a general overview that puts all five groups in context with each other. These differences in comparisons and normalization to the respective common standards entail the use of different statistical tools, resulting in different p-values. The statistical testing approaches and thresholds are now disclosed in the figure legends, and additionally in the SI Materials and Methods section (Lines 145-155).

Rev#1.8. When were the ex vivo samples taken for metabolomics, and if taken when significant TUNEL staining and cell death have occurred, are the changes in metabolism due to cell death or a true indication of differential metabolism? Furthermore, it is unclear if the metabolomics samples included the RPE or not. Considering these treatments will affect most cells in the retina and the RPE, which is included in the ex vivo samples, it is difficult to ascertain that these changes are secondary to the effects on photoreceptors alone.

Response: The samples for metabolomics included the RPE (except for the no RPE condition) and were taken at the same time as the tissues for histological preparations and TUNEL assays, i.e. they were all taken at post-natal day 15. This has now been clarified in the SI Materials and Methods section (Lines 108-110).

We cannot entirely exclude an effect of ongoing cell death caused by the different drug treatments on the retinal metabolome. However, since in the experimental treatments cell death was still comparatively low (even in the FCCP condition, overall cell death was only around 10% of the total retina), and the metabolomic analysis considered the entire tissue, the impact of cell death per se on the total metabolome will be comparatively minor (≤ 10%, i.e. within the typical error margin of the metabolomic analysis).

As mentioned above, the drug treatments should in principle affect all retinal cells at the same time. However, only cells that express the drug targets (i.e. 1,9-DDF targets GLUT1 in RPE cells, Shikonin targets PKM2 in photoreceptors; cf. Figure 1A) should react to the treatment. Even FCCP, in the paradigm employed, will only affect those cells that rely heavily on OXPHOS. Our data indicates that while this is almost certainly the case for rods; cones, RPE cells, and essentially all of the inner retina, are not affected by FCCP treatment, strongly suggesting that OXPHOS is of minor importance for these cell populations.

Rev#1.9. Why were the FCCP and no RPE groups compared? If they have similar metabolite patterns as noted in Figure 2, would that suggest that FCCP's greatest effect is on the ox phos of RPE and the metabolite patterns are secondary to alterations in RPE metabolism? Also, the increase in citrate and decrease in NAD may be related to effects on RPE mitochondrial metabolism when comparing these groups, and the disruption of RPE metabolism may then result in PARP staining of photoreceptors.

Response: The reason for the pair-wise comparison of the no RPE and FCCP groups initially was indeed the similarity in metabolite patterns. This was now rephrased accordingly in the results section “Photoreceptors use the Krebs-cycle to produce GTP” (Lines 218-219). The interpretation that the reviewer proposed here is interesting, but does not conform with the data analysis of this and other group comparisons.

Instead, the similarity between the metabolic patterns found in the no RPE and FCCP groups further supports the idea that a lack of RPE decreases retinal OXPHOS and increases glycolysis. This interpretation is based on the following observations:

(1) Mitochondrial density in the RPE is far lower than in photoreceptors (see COX staining in Figure 1A), thus quantitatively the metabolite pattern caused by a disruption of OXPHOS (via FCCP treatment) will be dominated by metabolites generated by photoreceptors. For the same reason the depletion of retinal NAD+, and the concomitant increase in photoreceptor PAR accumulation after FCCP treatment, is unlikely to be due to changes in RPE.

(2) Similarly, citrate synthase (CS) was found to be almost exclusively expressed in photoreceptor inner segments, with little expression in RPE (Figure 3D). Hence, the quantitative increase of citrate levels after FCCP treatment can only originate in photoreceptors.

(3) The comparison of the control (with RPE) against the no RPE group suggested an increase in (aerobic) glycolysis in the absence of RPE, evidenced notably by a retinal accumulation of lactate, BCAAs, and glutamate (Figure 3A). The very same metabolite pattern is seen for the FCCP treatment (Figure 1B) indicating a marked upregulation of glycolysis (Figure 6C). The latter observation suggests that photoreceptors, after disruption of OXPHOS switch to an exclusively glycolytic metabolism, which, however, rods cannot sustain (Figure 1C, D).

(4) Glucose consumption and lactate release is increased in the no RPE group vs. control (new Supplementary Figure 4). A similar increase in glucose consumption and lactate production is seen in the FCCP group suggesting that also the no RPE situation disrupts OXPHOS in photoreceptors.

Rev#1.10. The conclusions being reached are difficult to interpret secondary to the experimental procedures and the fact that the treatments are not cell-specific and RPE is included with the neuroretina as well. Likewise, stating FCCP is altering the Krebs cycle in the neuroretina is difficult to believe as there are no changes in the Krebs cycle when compared to the control, which also has RPE.

Response: We agree with the reviewer, that some of the conclusions may have been somewhat speculative. Accordingly, we have toned down our conclusions in several instances in the text, notably in abstract, introduction, and discussion.

When it comes to Krebs cycle intermediates a key limitation of our study is indeed the lack of carbon-tracing and metabolic flux analysis as noted by the reviewers, a limitation that we now highlight more strongly in the discussion of the revised manuscript (Lines 545-549). While it is highly probable that the flux of Krebs cycle intermediates is altered by FCCP, our steady-state data does not show significant changes in the metabolites citrate, fumarate, and succinate. However, our study does show a highly significant decrease in GTP levels, which as explained above, is a key indicator of Krebs cycle activity/inactivity. Moreover, while GTP levels were reduced also in the no RPE group, GTP was still significantly higher in the no RPE group compared to the FCCP treatment. Our interpretation of this finding is that there is Krebs-cycle/OXPHOS activity in the neuroretina, which is abolished by FCCP.

Rev#1.11. Supplemental Figure 4C and D states that GAC inhibition affected only photoreceptors, but GAC is expressed throughout the retina and so the inhibition is altering glutamine-glutamate homeostasis throughout the retina. Clearly, based on histology, one can see that the architecture of the retina, especially at the highest dose, is lost likely because all cells are being affected. So it is not photoreceptor-specific and even at low doses one can see that the inner retina is edematous. Moreover, with such a high amount of TUNEL staining in the ONL, are rods more affected than cones?

Response: In our hands the immunostaining for Glutaminase C (GAC) labelled predominantly cone inner segments, the OPL, and perhaps bipolar cells (Figure S1A). The deleterious effects mentioned by the reviewer are only seen at the highest concentration of the GAC inhibitor compound 968. This concentration (10 µM) is 100-fold higher than the dose that produces a significant loss of cones in the outer retina (0.1 µM). We therefore think that this data points to the extraordinary reliance of cones on glutamine and glutamate. As can be seen from the images (Figure S4C) illustrating the effects of 0.1 and 1 µM Compound 968 treatment, the ONL thickness is not significantly reduced by the GAC inhibitor. This strongly indicates that at these doses the rods are not affected by GAC inhibition.

Rev#1.12. The no RPE vs 1,9 DDF data may be interpreted as preventing glucose transport in the RPE increases BCAA catabolism by the RPE, which has been shown to utilize BCAA in culture systems. To this end, when the RPE is not present, the BCAA is increased as compared to the control with RPE.

Response: Our original interpretation of this data was that after GLUT1 inhibition and a correspondingly reduced retinal glucose uptake, the retina switched to an increasing use of anaplerotic substrates, including BCAAs. This is supported by the concomitant upregulation of the Cahill-cycle product alanine and the mini-Krebs-cycle product N-acetylaspartate (NAA). Yet, we agree with the reviewer that BCAAs could also be consumed by the RPE. We have now changed our conclusion at the end of the results chapter “Reduced retinal glucose uptake promotes anaplerotic metabolism“ to also highlight this possibility (Lines 261-262).

Rev#1.13. It is unclear why so much effort is comparing the no RPE group to the treatment groups and not comparing the control group to the different treatment groups.

Response: Previous studies – including the seminal studies of Otto Warburg from the early 1920s – had always used retina without RPE. This “no RPE” situation is therefore something of a reference for our entire study, which is why we dedicated more effort to its analysis. We have now inserted a corresponding remark into the manuscript (Lines 182-184).

Rev#1.14. The conclusions are significantly overstated especially with regards to rods versus cones as these are not cell-specific treatments. For example, the control vs 1,9 DDF vs FCCP clearly shows that there is mitochondrial dysfunction due to decreased NAD, increased AMP/ATP ratio, decreased Asp but increased Gln, and a compensatory increase in lactate production.

Response: We agree with the reviewer and have tried to phrase our statements in more measured fashion. Notably, we have toned down our statements in the title, abstract, results, discussion, and several of the subchapter headings.

Rev#1.15. While metabolic conclusions are drawn on serine/lactate ratio, this ratio is driven by the drastic changes in lactate and not so much serine in the treatment conditions as it was rather stable. Likewise, substrates beyond glucose have the potential to fuel the TCA cycle and make GTP via SUCLG1, such as fatty acids, other AAs, etc. Therefore, this ratio may not tell the entire story about anaplerotic metabolism. Furthermore, knowing that RPE utilize BCAAs to fuel their TCA cycle, the no RPE condition may simply have increased BCAAs due to lack of metabolism by the RPE, which drives the GTP/BCAA ratio. To state that the neuroretina was utilizing BCAAs for anaplerosis is not well supported based on the current data. Similarly, what is to say that the GTP/lactate ratio in the no RPE situation is not driven by the fact that the RPE is no longer present to act as acceptor of retinal lactate production or that more glucose is reaching the retina since the RPE is not present to accept and utilize that produced. Glucose uptake was not assessed to further address these issues.

Response: We agree with reviewer that metabolite ratios may not tell the full story underlying retinal metabolism however based on the robustness of using quantitative and highly reproducible NMR data, they are an important part of the metabolomics toolbox. The reviewer correctly observed that the changes in lactate levels are more dramatic than in serine. Still, also serine was significantly increased in the no RPE, 1,9-DDF, and Shikonin groups. Together with the lactate changes (same or opposite direction) the resulting serine/lactate ratios display marked alterations.

When it comes to the supply of other potential energy substrates mentioned by the reviewer, i.e. fatty acids or amino acids other than BCAAs, these are only supplied in minimal amounts in the defined, serum free R16 medium (Romijn, Biology of the Cell, 63, 263-268, 1988) and – if used to any important extent – would be rapidly depleted by the retina. Thus, for a culture period of 2 days in vitro between medium changes these energy sources are not available and thus cannot be used by the retina.

Our conclusion that the retina is using anaplerosis is based not only on the observations made in the no RPE group but also on, for instance, the metabolite ratios seen in the 1,9-DDF treatment group. In this group decreased glycolytic activity may correspond to increased serine synthesis and anaplerosis.

As far as glucose uptake is concerned, we have analysed the medium samples at P15 (equivalent to the retina tissue collection time point) and now present data that addresses this question more directly via the consumption of glucose from and release of lactate to the culture medium (New Supplementary Figure 4C, D). This new dataset provides another independent observation showing that:

(1) Glucose consumption/lactate release (i.e. aerobic glycolysis) is high in the no RPE situation but low in the control situation. In other words, retinal aerobic glycolysis is most likely stimulated by the absence of RPE.

(2) 1,9-DDF treatment decreases glucose consumption/lactate release as would be expected from a GLUT1 blocker. Since ATP and GTP production are high nonetheless, this indicates that other substrates (i.e. anaplerosis) were used for retinal energy production, in agreement with the analysis shown in Figure 6C.

(3) The FCCP treatment, which disrupts oxidative ATP-production, increases glucose consumption/lactate release in way similar to the no RPE situation. Yet, the no RPE retina can still generate sizeable amounts of GTP but not ATP. Together, this provides further evidence that neuroretinal OXPHOS is decreased in the absence of RPE.

Rev#1.16. The evidence for the mini-Krebs cycle is intriguing but weak considering it is based on certain enzymes being expressed in the photoreceptors, which had already been shown to be present in other publications, and a single ratio of metabolites that is increased in FCCP. One would expect this ratio to be increased under FCCP regardless. There is no stable isotope tracing with certain fuels to confirm the existence of the mini-Krebs cycle.

Response: We thank the reviewer for this suggestion. We agree that our evidence for the mini-Krebs-cycle (and the Cahill-cycle) may be to some extent circumstantial and additional technologies would help to obtain further supportive data. Still, here we would like to invite the reviewer to a thought experiment where he/she could try and interpret our data without considering the Cahill- or the mini-Krebs-cycle. At least we ourselves, when we engaged into such thought experiments, were unable to explain the data observed without these alternative energy-producing cycles. Most notably, we were unable to explain the strong accumulation of either alanine or N-acetyl-aspartate (NAA) when only considering glycolysis and (full) Krebs-cycle metabolism. Of course, this may still be considered “weak” evidence, and we expect that future studies including complementary technologies will either confirm or expand our interpretation of the existing data set.

The suggestion to perform stable isotope-labelled tracing with potential alternative fuels (e.g. glutamate, glutamine, pyruvate, etc.) is very attractive indeed. While such studies are likely to shed further light on the metabolic pathways proposed, this will entail very extensive experimental work, with multiple different conditions and concentrations and variety of analysis methods that is currently not feasible (e.g. a 1.7 mm NMR probe equipped with a 15N channel) as an extension of the present manuscript. Nevertheless, we will certainly consider this approach for future follow-up studies once such techniques are available and will screen for suited collaboration partners. A corresponding comment on such future possibilities has now been inserted into the discussion (Lines 545-549).

Rev#1.17. The discussion does not mention how this data contradicts a recent in vivo study looking at Glut1 knockout in the retina (Daniele et al. FASEB. 2022) or previous in vivo studies that suggest cones may be less sensitive to changes in glucose levels (Swarup et al. 2019). This is a key oversight.

Response: We thank the reviewer for pointing this out. We now included these studies in the revised discussion in a new subchapter on the expression of glucose transporters in the outer retina (Lines 454-472). For a critical review of the Daniele et al., 2022 study please also see our more detailed response to question Rev#3.13 below.

Rev#1.18. GAC is expressed in more than just cones so making cell-specific statements regarding fuel utilization is not well supported.

Response: Our immunostaining for GAC revealed a strong expression in cone inner segments (Figure S1A3). While this does not exclude (relatively minor) expression in other retinal cell types, cones are likely to be more reliant on GAC activity than other cell types. See also answer above.

Rev#1.19. Suggesting that rods utilize the mini-Krebs cycle based on AAT2 being seen in the inner segments without at least co-staining for RHO or PNA is weak evidence for such a cycle. AAT looks to be expressed in the inner segments of all photoreceptors.

Response: We have taken up this suggestion from the reviewer and now provide an additional co-staining for AAT1 and AAT2 with rhodopsin. Note that in response to a pertinent comment from Reviewer #3 we have changed the abbreviation for aspartate aminotransferase from “AAT” to the more commonly used “AST” throughout the manuscript.

New images showing a co-staining for AST1 and AST2 with rhodopsin now replace the former image set in Figure 7D. In brief, the new images show the expression of both AST1 and AST2 across the retina, with, notably an expression in the inner segments of photoreceptors but not in the outer segments, where rhodopsin is expressed.

Reviewer #3 (Recommendations For The Authors):

Rev#3.1. The staining for the glucose transporters GLUT1 and GLUT3 does not reflect what has previously been published by two different groups that were validated by cell-specific knockout mice. As mentioned by the author GLUT1 and GLUT3 have differences in transport kinetics, which would affect their metabolism. Therefore, the lack of GLUT1 in photoreceptors would suggest that photoreceptor metabolism is not faithfully replicated in this system. This difference from the previous literature should be discussed in the discussion.

Response: As the reviewer pointed out, the expression of GLUT1 in the retina is somewhat controversial, with much older literature showing expression on the RPE, while some more recent studies claim GLUT1 expression in photoreceptors. For a brief discussion of our GLUT1 immunostaining please see also our answer to question Rev#1.3 above.

Although the retinal expression of GLUT1 was besides the focus of our study, we feel we must address this point in more detail: In the brain the generally accepted setup for GLUT1 and GLUT3 expression is that low-affinity GLUT1 (Km = 6.9 mM) is expressed on glial cells, which contact blood vessels, while high-affinity GLUT3 (Km = 1.8 mM) is expressed on neurons (Burant & Bell, Biochemistry 31:10414-20, 1992; Koepsell, Pflügers Archiv 472, 1299–1343, 2020). This setup matches decreasing glucose concentration with increasing transporter affinity, for an efficient transport of glucose from blood vessels, to glial cells, to neurons. In the retina, the cells that contact the choroidal blood vessels are the tight-junction-coupled RPE cells. As shown by us and many others, RPE cells strongly express GLUT1 (cf. Figure 1A-3.). To warrant an efficient glucose transport from the RPE to photoreceptors, photoreceptors must express a glucose transporter with higher glucose affinity than GLUT1. We show that this is indeed the case with photoreceptors expressing GLUT3 (cf. Supplemental Figure 1-5.). While a part teleological explanation does not per se prove that our data is correct, at least our data is plausible. In contrast, the glucose transporter setup sometimes claimed in the literature is biochemically implausible, i.e. for the flow of metabolites (glucose) to go against a gradient of transporter affinities, and we are not aware of an example of such a setup occurring anywhere in nature.

However, at this point we cannot exclude low levels of GLUT1 expression on Müller glia cells or even photoreceptors. This expression could, for instance, be relevant in cases where cells were shuttling excess glucose – perhaps produced through gluconeogenesis – onwards to other retinal cells. Still, GLUT1 expression can only be minor when compared to RPE since a major expression would destroy the glucose affinity gradient (see above) required for efficient glucose shuttling into the energy hungry photoreceptors.

To address this request by the reviewer (and also reviewer #1) we now discuss the question of glucose transporter expression in the outer retina in a new subchapter of the discussion (Lines 454-472).

Rev#3.2. Photoreceptor metabolism and aerobic glycolysis are tied to photoreceptor function, as demonstrated by Dr. Barry Winkler. The authors should provide data or mention (if previously published) about photoreceptor OS growth and function in this system.

Response: The studies of Barry Winkler (e.g. Winkler, J Gen Physiol. 77, 667-692, 1981) confirmed the original work of Otto Warburg and expanded on the idea that the neuroretina was using aerobic glycolysis. Importantly, Winkler used a very similar experimental setup as Warburg has used, namely explanted rat retina without RPE. In light of our data where we compare metabolism of mouse retina with and without RPE – where retina cultured without RPE confirms the data of Warburg and Winkler – it appears most likely that the purported aerobic glycolysis occurs mostly in the absence of RPE but only to a lower extent in the native retina.

Photoreceptor outer segment outgrowth is somewhat slower in the organotypic retinal explant cultures compared to the in vivo situation (cf. Caffe et al., Curr Eye Res. 8:1083-1092, 1989 with LaVail, JCB 58:650-661, 1978; see also answer to reviewer #1). Importantly, organotypic retinal explant cultures and their photoreceptors are fully functional and remain so for extended periods in culture (Haq et al., Bioengineering 10:725, 2023; Tolone et al., IJMS 24:15277, 2023). This information has now been added to the manuscript discussion section, into the new subchapter “The retina as an experimental system for studies into neuronal energy metabolism” (Lines 367-395).

Rev#3.3. It is unclear from the description of the experiment in both the results and methods if 1,9DDF, Shikonin, and FCCP were added to both apical and basal media compartments or one or the other and should be specified. The details of what was on the apical compartment would be helpful, as the model is supposed to allow for only nutrients from the basal compartment (as indicated by the authors themselves). Is the apical compartment just exposed to air? How does this affect survival?

Response: In organotypic retinal explant cultures the RPE rests on the permeable culturing membrane such that the basal side is contact with the membrane and the medium below (far schematic drawing see Figure S1B), while the apical side is covered by a thin film of medium created by the surface tension of water (Caffe et al., Curr Eye Res. 1989; Belhadj et al., JoVE, 2020). This thin liquid film ensures sufficient oxygenation and is an important factor that allows the retinal explant to remain viable for several weeks in culture. If the retinal cultures were submerged by the medium, their viability – especially that of the photoreceptors – would drop dramatically and would typically be below 3-5 days. Therefore, in the retinal organotypic explant cultures used here, the nutrients and the drugs applied do indeed reach the outer retina from the basal side, i.e. similar as they would in vivo.

To address this question from the reviewer, corresponding clarifications have been inserted into the SI Materials and Methods section (Lines 64-66).

Rev#3.4. As the metabolomic data obtained was quantitative, several metabolites discussed should be analyzed in terms of ratios, for example, Glutathione and glutathione disulfide should be reported as a ratio. In addition as ATP, ADP, and AMP were measured, they can used to calculate the energy charge of the tissue.

Response: We thank the reviewer for these suggestions and have created corresponding graphs for GSH / GSSG ratio and energy charge. These new graphs have now been added to the SI datasets, to the new Supplementary Figure 4. To accommodate other requests from the Reviewers, this new Figure also contains additional new datasets on glucose and lactate concentrations (see further comments above and below). Please note that all later SI Figures have been renumbered accordingly.

In brief, the ratios for GSH/GSSG show no significant changes between control and the different experimental groups. Meanwhile, the adenylate energy charge of the retinal tissues show a significant decrease in the energy charge for the Shikonin group and the FCCP group. Note that in the new Supplementary Figure 4A, the dotted lines indicate the energy charge window typical for most healthy cells (0.7 – 0.95).

Rev#3.5. I think a missed opportunity when discussing the possible taurine/hypotaurine shuttle would be the impact on the osmosis of the subretinal space as taurine has been hypothesized as a major osmolyte.

Response: This is another interesting recommendation from the reviewer. To address this point, we have now introduced a corresponding paragraph and references in the discussion of the manuscript (Lines 503-504; 512-514).

Rev#3.6. In Figure 3, the distribution of these enzymes should also be studied under the no RPE condition as the culture treatment took several days for these metabolic changes to occur.

Response: The images shown in Figure 3D are from the in vivo retina. Since this may not have been very clear in the previous manuscript version, we have now added a corresponding explanation to the legend of Figure 3. As far as we can tell, the expression and localization of neuroretinal enzymes does not change in cultured retina, during the culture period (compare Figure 1A with Supplementary Figure S1C). However, when it comes to the metabolite taurine its production (localization) changes dramatically in the no RPE situation where taurine is essentially undetectable by immunostaining (not shown but see metabolite data in Figure 2A, Figure 3A).

Rev#3.7. In Figures 4 and 5, it is unclear why the experimental groups were not compared to the control and requires further explanation. Furthermore, the authors should justify the concentrations of drugs used as the cell death could have risen from toxicity to the drugs and not due to disruption of metabolism.

Response: The reviewer is right, the rationale for these comparisons may not have been laid out with sufficient clarity. In Figure 4 the no RPE and FCCP groups are compared because both groups showed similar metabolite changes towards the control situation. The no RPE to FCCP comparison thus focussed on the details of the – at first seemingly minor – differences between these two groups. This has now been clarified in the corresponding part of the results (Lines 218-219).

In Figure 5A, B we compare the no RPE and 1,9-DDF groups with each other, notably because the data obtained seemingly contradicted our initial expectation that these two groups should show similar metabolite patterns. Also here, we have now inserted an additional explanation for this choice of comparisons (Lines 252-253).

In Figure 5C, D we compare the Shikonin and FCCP groups with each other. The idea behind this comparison was that in the 1st group glycolysis was blocked while in the 2nd group OXPHOS was inhibited, or in other words here were compared what happened when the two opposing ends of energy metabolism were manipulated in opposite directions. This reasoning is now given in the results section (Lines 265-268).

As far as the choice of drugs and concentrations is concerned, we used only compounds that have been extremely well validated through up to five decades of scientific research (see initial response to Reviewer #2 above). We therefore are confident that at the concentrations employed the results obtained stem from drug effects on metabolism and not from generic, off-target toxicity. Then again, as we show, prolonged (i.e. 4 days) block of energy metabolism pathways does cause cell death.

Rev#3.8. In line 203, the authors discuss GTP as being primarily a mitochondrial metabolite, however, photoreceptors would require a localized source of GTP synthesis in the outer segments as part of phototransduction, and therefore GTP in photoreceptors cannot be a mitochondrial-specific reaction in photoreceptors. Furthermore, the authors mentioned NDK as being a possible source of GTP, but they do not show NDK localization despite it being reported in the literature to be localized in the OS.

Response: The question as to the source of GTP in photoreceptor outer segments is indeed highly relevant. For GTP production in mitochondria see the answer to the next question below (Rev#3.9). An early study showed nucleoside-diphosphate kinases (NDK) to be expressed on the rod outer segments of bovine retina (Abdulaev et al., Biochemistry 37:13958-13967, 1998). More recently NDK-A was shown to be strongly expressed in photoreceptor inner segments (Rueda et al., Molecular Vision 22:847-885, 2016). We now refer to both studies in the results section of the manuscript (Line 227-228).

Rev#3.9. In the "Impact on glycolytic activity, serine synthesis pathway, and anaplerotic metabolism" section, the authors claim in the no RPE group glycolytic activity was higher due to a depressed GTP-to-lactate ratio. However, this reviewer is under the impression that GTP production in photoreceptors is not mitochondrial specific, so this ratio doesn't make sense (I could be mistaken, however). A better ratio would have been pyruvate/lactate or glucose/lactate when discussing increased glucose consumption.

Response: We appreciate the reviewers’ comment, yet we do indeed believe we can show that GTP-production in our experimental context is mainly mitochondrial. As explained in the manuscript results section (“Photoreceptors use the Krebs-cycle to produce GTP”), there are essentially only two possibilities for a photoreceptor to produce sizeable amounts of GTP. In the mitochondria via SUCLG1 – i.e. an enzyme highly expressed in photoreceptor inner segments (Figure 5D) – and the cytoplasm via NDK from excess ATP. The claim about the depressed GTP-to-lactate ratio in the no RPE situation takes this into account. Importantly, since in the no RPE situation ATP-levels are significantly lower than GTP, here GTP can only be produced via SUCLG1 and OXPHOS. Moreover, this contrasts with the FCCP group where mitochondrial OXPHOS is disrupted and both ATP and GTP are depleted.

As far as ratios with pyruvate and glucose are concerned, we agree that these could potentially be very interesting to analyse. Unfortunately, in our retinal tissue 1H-NMR spectroscopy- based metabolomics analysis the levels of both pyruvate and glucose were below the detection limits which likely reflects their rapid metabolic turnover (cf. table S1). While this might be attributable to the marked consumption of these metabolites within the tissue, it does not allow for us to calculate the suggested ratios to lactate. Then again, in the supernatant medium which was collected at the same time point as the retina tissue, we can readily detect glucose and lactate levels, for this data please see the new Supplementary Figure 4.

Rev#3.10. Aspartate aminotransferase should be abbreviated as AST, as it is more commonly noted.

Response: In response to this comment from the reviewer, we have changed the abbreviation for aspartate aminotransferase from AAT to AST throughout the manuscript.

Rev#3.11. In the discussion the assumptions of the ex vivo culture systems should be clearly stated. One that was not mentioned, but affects the implications of the data, is that the retinas used in this study are from the developing mouse eye. Another important assumption that was made in this paper was that the changes in retinal metabolism were due to photoreceptors even though the whole neural retina was included.

Response: The reviewer is correct; we have added these two points to the discussion section of the manuscript. Notably, we now included a new subchapter “The retina as an experimental system for studies into neuronal energy metabolism” (Lines 367-395) to present different in vitro and in vivo test systems.

Rev#3.12. Starting at line 347: As the authors know, the RPE has been shown to be highly reliant on mitochondrial function, and disruption of RPE mitochondrial metabolism leads to photoreceptor degeneration (numerous papers have shown this). Furthermore, the lower levels of lactate detected in their explants when RPE was present suggests that lactate is actively transported out of the neural retina by the RPE.

Response: The reviewer is right about lactate being exported from the retina to the blood stream in vivo, or, in our in vitro study, to the culture medium. In the new dataset showing glucose and lactate concentrations in the culture medium (new Supplementary Figure 4C, D), we show that without RPE (no RPE group) and the retina releases more significantly lactate into the medium than control retina with RPE. At the same time the no RPE retina consumes more glucose than control retina.

Rev#3.13. Line 360: Again, in mouse photoreceptors (by bulk RNAseq and scRNAseq), there is no GLUT3 expression (encoded by slc2a3). It was also recently shown by Dr. Nancy Philp's lab that rod photoreceptors express GLUT1, encoded by slc2a1 (PMCID: PMC9438481). The differences reported in this study and previous studies should be discussed.

Response: Although this comment may not make us very popular, we are somewhat sceptical of RNAseq data (especially single cell RNAseq) since the underlying methodology – at the current level of technological development – is notoriously unreliable when it comes to the assessment of low abundance transcripts and suffers from apoor batch reproducibility, compared to NMR based metabolomics. Due to methodological constraints RNAseq have a propensity to display erroneously high or low expression. Moreover, and perhaps even more important, dissociated cells in scRNAseq studies undergo rapid gene expression changes that can significantly falsify the image obtained (Rajala et al., PNAS Nexus 2:1-12, 2023). Finally, it cannot be emphasized enough that mRNA expression profiles DO NOT equate protein expression and there are numerous examples for divergent expression profiles when mRNA and protein is compared.

The Daniele et al. study (FASEB Journal 36:e22428, 2022; PMCID: PMC9438481) used in situ hybridization to study the mRNA expression of GLUT1 (slc2a1) and GLUT3 (slc2a3). In line with our comment just above, the Daniele et al. study may provide for an example of divergence between mRNA and protein expression, since it seemingly showed only minor expression of GLUT1/slc2a1 in the RPE, i.e. precisely in the one cell type that is well-known for its very strong GLUT1 protein expression.

Furthermore, Daniele et al. used a conditional GLUT1 knock-out in photoreceptors induced by repeated Tamoxifen injections. The photoreceptor GLUT1 knock-out led to a relatively mild phenotype with only about 45% of the outer nuclear layer lost over a 4-months time-course. This is in stark contrast with the FCCP or the 1,9-DDF treatment, which would ablate nearly all rod photoreceptors in under one or two weeks, respectively.

As a side note, Tamoxifen is an oestrogen receptor antagonist (with partial agonistic behaviour) with a long history of causing retinal and photoreceptor damage. Notably, oestrogen receptor signalling is important for maintaining photoreceptor viability (Nixon & Simpkins, IOVS 53:4739-47, 2012; Xiong et al., Neuroscience 452:280-294, 2021). Therefore, the relatively minor effects of the conditional GLUT1 KO in photoreceptors found in Daniele et al. may have been confounded by direct tamoxifen photoreceptor toxicity. On a wider level, this possible confounding factor related to the use of Tamoxifen points to general problems associated with certain forms of genetic manipulations.

We now mention the controversy around the expression of glucose transporters in the retina, including the Daniele et al. study in a new subchapter of the discussion on "Expression of glucose transporters in the outer retina” (Lines 454-472).

Rev#3.14. Lines 370-372: FCCP caused a strong cell death phenotype in rods, however under stress rods upregulate the secretion of RdCVF, which leads to cone photoreceptor survival by the upregulation of aerobic glycolysis in cones. The data should be re-interpreted in the context of this previous literature.

Response: We thank the reviewer for this comment; however, we could not find a reference that would state that “…under stress rods upregulate the secretion of RdCVF”. What we did find was a reference stating that similar factors such as thioredoxins (TRX80) are secreted from blood monocytes under stress (Sahaf & Rosén, Antioxid Redox Signal 2:717-26, 2000). However, we consider these cells to be too dissimilar to rod photoreceptors to warrant a corresponding comment. Moreover, the research group who discovered RdCVF originally showed that rod-secreted RdCVF cannot prevent cone degeneration if the corresponding Nxnl1 gene is knocked-out in cones, arguing for a cell-autonomous mechanism of RdCVF -dependent cone protection (Mei et al., Antioxid Redox Signal. 24:909-23, 2016).

Since it is very possible that we may have missed the correct reference(s), we would welcome further guidance by the reviewer.

Rev#3.15. Line 374: 1,9-DDF caused a 90% loss of cones, however, previous studies by Dr. Nancy Philp have shown glucose deprivation in the outer retina affects primarily rod photoreceptors. The differences should be discussed.

Response: We thank the reviewer for directing us to these studies. As mentioned above (Rev#3.13.) the Daniele et al. 2022 study yielded only relatively mild effects for a rod-specific conditional GLUT1 KO on photoreceptor viability. Similarly, in an earlier study (Swarup et al., Am J Physiol Cell Physiol. 316: C121–C133, 2019) the Philp group found that also a GLUT1 KO in the RPE caused only a minor phenotype in the photoreceptor layer. We would argue that if glucose, and by extension aerobic glycolysis, were indeed of major importance for (rod) photoreceptor survival, the degenerative effect of these genetic GLUT1 ablations should have been devastating and should have destroyed most of the outer retina in a matter of days. The fact that this was not seen in both studies is another piece of independent evidence that rod photoreceptors do not rely to any major extent on glycolytic metabolism.

The two studies from the Philp lab (Swarup et al., 2019; Daniele et al., 2022) are now cited in the discussion (Lines 417-419 and 458-460).

Rev#3.16. Line 375: Yes Dr. Claudio Punzo and Dr. Leveillard Thierry along with other groups have shown glycolysis is required to maintain cone survival when under stress, however, the authors should emphasize that it is under stress that this is observed.

Response: In response to this comment we have now specifically extended our corresponding remark in the discussion of the manuscript (Lines 446-447).

Rev#3.17. The section "Cone photoreceptors use the Cahill-cycle". The presence of ALT in photoreceptors was surprising and suggests alternatives to the Cori reaction. However, previous measurements of glucose and lactate from localized in vivo cannulation of animal eyes suggest the majority of glucose taken up by the retina is released back to the blood as lactate. Again, this section should discuss this idea in terms of the previous literature.

Response: Here, we believe the reviewer is referring to studies performed in the late 1990s where, in anaesthetized cats, the lactate concentration in blood samples obtained from choroidal vein cannulation was compared against that in blood samples obtained from femoral arteries (Wang et al., IOVS 38:48-55, 1997). We note that a more relevant in vivo measurement of retinal glucose consumption and lactate production would likely require the simultaneous cannulation of the central retinal artery (CRA) and the central retinal vein (CRV). This would need to be combined with repeated (online) blood sampling, drug applications, and subsequent metabolomic analysis. We are not aware of any in vivo studies where such procedures have been successfully performed and further miniaturization and increased sensitivity of metabolomic analytic equipment will likely be required before such an undertaking may become feasible. Even so, such studies may not be feasible in small rodents (mice, rats) and may instead require larger animal species (e.g. dog, monkey) to overcome limitations in eye and blood sample size.

We have now extended the discussion of our manuscript with a new subchapter on “The retina as an experimental system for studies into neuronal energy metabolism”. Within this new subchapter we now present two different in vivo experimental approaches that addressed retinal energy metabolism (Lines 376-384). Moreover, we now present new data on retinal lactate release to the culture medium, showing, for instance, a strong increase in lactate release in the no RPE condition compared to control (new Supplementary Figure 4).

Rev#3.18. Lines 431-433: The study cited suggested that the mitochondrial AST was detected in other cells, in agreement with the data shown. However, the authors' statements in this section are misleading as they do not take into consideration the contribution of AST from other cell types.

Response: The reviewer is right, we found both AST1 and AST2 to be expressed not only in photoreceptor inner segments but also in the inner retina, especially in the inner plexiform layer (new Figure 6D). Since this might indicate mini-Krebs-cycle activity also in retinal synapses, we have added a corresponding comment to the discussion (Lines 540-543).

Grammatical and wording fixes:

Rev#3.19. Line 98 - "the recycling of the photopigment, retinal."

Response: We have inserted a comma after “photopigment”.

Rev#3.20. Results section and Figure 1 start without providing context for the model system where staining is being done.

Response: We have added this information to the beginning of the results section (Lines 105-106).

Rev#3.21. Supplementary Figure 2 is not mentioned in the main text - there is no context for this figure.

Response: Supplementary Figure 2 was originally referenced in the legend to Figure 2. We now mention supplementary figure 2 (now renumbered to supplementary figure S3) also in the main text, in the results section under “Experimental retinal interventions produce characteristic metabolomic patterns” (Line 148).

Rev#3.22. Volcano plot in Supplementary Figures 3, 5, 6, 7, and 8 don't indicate what Log2(FC) is in reference to.

Response: The log2 fold change (FC) is calculated as follows: log2 (fold change) = log2 (mean metabolite concentration in condition A) - log2 (mean metabolite concentration in condition B) where condition A and condition B are two different experimental groups being compared. This is now explained in the SI Materials and Methods (Lines 145-147) and indicated in abbreviated form in the figure legends. Please note that supplemental figures have now been renumbered due to the insertion of an additional, new Figure.

Rev#3.23. Line 331 - –a“d allowed to analyze the..." ”s incorrect phrasing.

Response: This phrasing was changed.

Rev#3.24. Line 343 "c“cled" ”

Response: This phrasing was changed.

Rev#3.25. Line 446 is misworded.

Response: This phrasing was changed.

Technical questions:

Rev#3.26. At what point after explant was the IHC done in Supplemental Figure 1? If early, but experiments are done later, there's’a chance things are more disorganized at the end of the experiment.

Response: Staining and metabolomics analysis were both done at the end of each experiment, at the same time, at P15. This is now mentioned in the SI materials and methods section (Lines 67, 108-110).

Rev#3.27. FCCP affects plasma membrane permeability, which is particularly critical in neurons that undergo repolarization and depolarization - –ow do we know FCCP on cell death via metabolism? See: https://www.sciencedirect.com/science/article/pii/S2212877813001233

Response: The reviewer is correct, a significant permeabilization of cell membranes in general would likely cause extensive neuronal cell death, unrelated to a disruption of OXPHOS. However, the FCCP concentration used here (5 µM) is at the lower end of what was used in the mentioned Kenwood et al. study (Mol Metab. 3:114-123, 2014) and the effect on cell membrane permeability in tissue culture is likely to be rather small, as opposed to what was seen by Kenwood et al. in cultures of individual cells. This view is supported by the fact that in our FCCP treatments, we did not observe any significant increases of cell death in any retinal cell type (including RPE) other than in rod photoreceptors. Together with the fact that only photoreceptors strongly express Krebs-cycle/OXPHOS related enzymes, this strongly suggests that the FCCP effects seen by us were due to disruption of OXPHOS.

Rev#3.28. Numerous metabolite comparisons are being made throughout the manuscript – what type of multiple hypothesis testing corrections are utilized? Only certain figures mention multiple hypothesis testing (e.g. Figure 6).

Response: In general, in this manuscript we used two different statistical methods: 1) For two-group comparisons, we used an unpaired, two-tailed t-test, which reports a p-value with 95% confidence interval without additional multiple hypothesis testing (e.g. in Figure 2, Suppl. Figures 4, 6, 7, 8). 2) For multiple group comparisons we used a one-way ANOVA analysis with Tukey’s multiple comparisons post-hoc test (except suppl. Figure 9 where Fisher´s LSD post-hoc test was used). The information on which statistical test was used for what dataset is now given in the figure legends and in the SI Material and Methods section.

Rev#3.29. For Figure 3, how do we know that the removal of RPE is causing the metabolite changes due to RPE-PR coupling? How do you rule out the fact that it isn’t just: I – a thicker physical barrier between media and the neural retina that is causing the changes, or II – removal of RPE from PR causes OS shearing and a stress response that alters metabolism?

Response: We believe these concerns can be ruled out: The RPE cells are linked by tight junctions and are not “just a thicker barrier” but a barrier that is almost impermeable for most metabolites unless they are carried by specific transporters. Outer segment shearing via RPE removal would indeed be a concern if we had used adult retina. However, we explanted that retina at P9 when it does not possess any sizeable outer segments yet. As a matter of fact, photoreceptors grow out outer segments only after P9.

Rev#3.30. While 1,9-dideoxyforskolin blocks GLUT1, it is known to have other effects, including on potassium channels. How do we know the effects of 1,9-dideoxyforskolin are specific to GLUT1? Utilizing a GLUT1 KO and showing no additional effects when adding 1,9-dideoxyforskolin would be helpful as a control.

Response: This is a good suggestion from the reviewer. We note that this is technically not easy to achieve as it would require an RPE-specific knock-out that should be inducible at a given experimental time-point, in a quantitative manner. The study by Swarup et al. (see above Rev#3.13.) used an RPE specific knock-out that was, however, not inducible. Moreover, if the corresponding inducible knock-out animals could be generated, then the stochastic nature of the inducing treatment would probably affect only a limited number of cells within a given cell population. In our experimental context, a less than quantitative knock-out would significantly complicate interpretation of results, even to the point that no additional insight might be gained.

Rev#3.31. The analysis in Figure 6, even with attempts to control drug treatments, is highly speculative. One really needs animals with predominately cones vs. predominately rods to do this analysis (e.g. with NRL mice).

Response: The reviewer is right, the analysis shown in Figure 6 was an explorative approach to try and deduce features of rod and cone metabolism. This is now mentioned in the results section (Lines 282-284). Since the experiments were not initially intended to address such questions, by necessity the interpretations remain speculative. The comparison of mouse mutants in which there are either no cones (e.g. cpfl1 mouse) or no rods (e.g. NRL knock-out mouse) may allow to disentangle the metabolic contributions of rods and cones. We appreciate the suggestion from the reviewer and have now inserted a relating suggestion for future studies into the discussion section (Lines 450-452).

Rev#3.32. Overall, much of the paper suggests intriguing pathways, but without C13 tracing or relevant genetic knock-outs, the pathways would have to be speculative rather than definitive.

Response: We agree with the reviewer that further research, including 13C and 15N-tracing studies, will be necessary to evaluate which pathway(s) are used by what retinal cell type under what condition. Still, the high robustness and quantitative nature of the NMR metabolomics data allows us to draw pathway conclusions based on metabolites that are unique to specific pathways/cell types or using ratios. We now relate to the advantages of such carbon-tracing studies in the discussion of the manuscript (Lines 545-549).

Stylistic suggestions:

Rev#3.33. This is a very dense paper to read. It would be helpful for each figure to have a summary diagram of the relevant metabolite changes and how they fit together. Further, for those not metabolism-inclined, defining the mini-Kreb’s, Cahill, and Cori cycles and their brief implications at some point early in the manuscript would be helpful.

Response: We have been thinking a lot about how we could add in the suggested summary diagrams into each figure. Unfortunately, whatever idea we contemplated would have significantly increased the complexity of the figures, while the actual benefit in terms of improved understandability was unclear.

However, we did include the suggestion from the reviewer to present the terms Cori, Cahill-, and mini-Krebs-cycle already in the introduction and we hope that this has improved the understandability of the manuscript overall (Lines 79-92).

Rev#3.34. More discussion about the step-by-step ways that the mini-Kreb’s reaction “uncouples” glycolysis from the Kreb’s cycle would be helpful. What do you mean by “uncouple” in this context?

Response: We thank the reviewer for this suggestion. Uncoupling in this context means that glycolysis and Krebs cycle are not metabolically coupled to each other via pyruvate. Instead both pathways can run independently from each other and in parallel, as long as the Krebs-cycle uses glutamate, BCAAs or other amino acids as fuels. We now also address this point already in the introduction of the manuscript (Lines 87-90).

Conceptual questions:

Rev#3.35. As the proposal that PR undergo heavy amounts of OXPHOS is controversial, it would be helpful for the authors to review the literature on lactate production by the retina and what studies have shown previously about retina use of lactate, specifically lactate making its way into TCA cycle intermediates, suggesting OXPHOS, in PRs.

Response: In response to this question we have added several new references to the introduction and discussion of the manuscript. The question of lactate production (aerobic glycolysis) vs. the use of OXPHOS is now discussed in Lines 77-81, Lines 367-384.

Rev#3.36. Why would cones die more in the no RPE condition? The authors suggest this has something to do with GLUT1 expression on RPE and the transport of glucose to cones. Even if we accept that cones are highly glycolytic, loss of RPE should expose the neural retina to even more glucose in your experimental set-up.

Response: This is a very interesting question from the reviewer. Indeed, loss of the RPE and blood-retinal barrier function should increase photoreceptor access to glucose, even more so if they are expressing high affinity GLUT3. In the discussion (Lines 420-424), we speculate that this may trigger the Crabtree effect, shutting down OXPHOS and causing the cells to exclusively rely on glycolysis. This, however, will likely not yield sufficient ATP to maintain their viability, so that they “starve” to death even in the presence of ample glucose. Since cones require at least twice as much ATP as rods, they may be more sensitive to a Crabtree-dependent shut-down of OXPHOS. However, if this speculation was correct then the question remains why the FCCP treatment, which abolishes OXPHOS more directly, does not cause cone death. Here, we again can only speculate that high glucose may have additional toxic effects on cones that are independent of OXPHOS. We now try to present this reasoning in the discussion (Lines 426-429).

eLife assessment

Chen and colleagues utilize an in situ explant model of the neural retina and retinal pigment epithelium (RPE), along with small molecule inhibition of key metabolic enzymes and targeted metabolomic analysis, to decipher key differences in metabolic pathways used by rods, cones, Muller glia, and the RPE. They conclude that rods are heavily reliant on oxidative metabolism, cones are heavily reliant on glycolysis, and multiple mechanisms exist to decouple glycolysis from oxidative metabolism in the retina. This study provides valuable metabolomic data and insights into the metabolic flexibility of different retinal cells. However, current evidence is still incomplete as several of the conclusions from the paper stand in contradiction to other published findings and the authors naturally suggests experiments that will be needed in the future to validate the hypothesized pathways and refute existing published data. Such future validation includes animal models with tissue specific knockout of the key enzymes probed in the study; inhibiting the targets of this study with more than 1 small molecule that is structurally different, and at different doses and timings; using retinal explants from matured animals; performing labeled metabolite tracing experiments; and direct assessment of mitochondrial function (via OCR) under various manipulations.

Reviewer #1 (Public Review):

Summary:

In the resubmitted manuscript by Chen et al. entitled, "Retinal metabolism: Evidence for uncoupling of glycolysis and oxidative phosphorylation via Cori-, Cahill-, and mini-Krebs-cycle", the authors look to provide insight on retinal metabolism and substrate utilization but using a murine explant model with various pharmacological treatments in conjunction with metabolomics. The authors conclude that photoreceptors, a specific cell within the explant, which also includes retinal pigment epithelium (RPE) and many other types of cells, are able to uncouple glycolytic and Krebs-cycle metabolism via three different pathways: 1) the mini-Krebs-cycle, fueled by glutamine and branched-chain amino acids; 2) the alanine-generating Cahill-cycle; and 3) the lactate-releasing Cori-cycle. While the authors have toned down some of their bold conclusions made in the original manuscript, they did very little in the way of providing additional well-controlled experiments, including cell-specific treatments, genetic knockouts, or stable isotope tracing to support their conclusions. Rather, the authors proceed to speculate more without additional data. The major issues raised by this reviewer were not adequately addressed. As such, the conclusions continue to be highly speculative and not well supported with evidence.

Strengths of resubmission:

The resubmission toned down some of its bold statements.

Weaknesses of resubmission:

Major weaknesses of this study persist including lack of in vivo supporting data. Also, retinal explant culture metabolomics are done in neuroretina with RPE attached, which are metabolically active and can be altered by the treatments investigated herein, further confounding the claims made regarding the neuroretina. While including the RPE in the explant model is commended, it needs to be separated from the retina prior to metabolomics to get a better sense of each tissues' metabolism. Also, melanin within RPE will hinder immunofluorescence signal, so one cannot state that RPE do not express certain enzymes based solely on immunofluorescence. Pharmacologic treatments are not cell-specific as the enzymes are expressed in numerous cells within the retina and RPE, and/or the treatments have significant off-target effects (such as shikonin). So, it is difficult to ascertain that the metabolic changes are secondary to the effects on photoreceptors alone, which the authors claim. Additionally, the explants are taken at a very early age when photoreceptors are known to still be maturing. No mention or data is presented on how these metabolic changes are altered in retinal explants after photoreceptors have fully matured. Likewise, significant assumptions are made based on a single metabolomics experiment with no stable isotope tracing to support the pathways suggested. In vivo, stable-isotope retinal metabolomics are being done and have been done, so stating this technology is beyond our field is false. Therefore, the conclusions reached in this manuscript are still not supported.

Reviewer #2 (Public Review):

Summary:

The authors aim to learn about retinal cell specific metabolic pathways, which could substantially improve the way retinal diseases are understood and treated. They culture ex vivo mouse retinas for 6 days with 2 - 4 days of various drug treatments targeting different metabolic pathways or by removing the RPE/choroid tissue from the neural retina. They then look at photoreceptor survival, stain for various metabolic enzymes/transporters and quantify a broad panel of metabolites. While this is an important question to address, the results are not sufficient to support the conclusions.

Strengths:

The questions the authors are exploring at extremely valuable and I commend the authors and working to learn more about retina metabolism. The different sensitivity of the cones to various drugs is interesting and may suggest key differences between rod and cones. The authors also provide a thoughtful discussion of various metabolic pathways in the context of previous publications.

Weaknesses:

As the authors point out, ex vivo culture models allow for control over multiple aspects of the environment (such as drug delivery) not available in vivo. Ex vivo cultures can provide good hints as to what pathways are available between interacting tissues. However, there are many limitations to ex vivo cultures, including shifting to a very artificial culture media condition that is extremely different than the native environment of the retina. It is well appreciated that cells have flexible metabolism and will adapt to conditions provided. Therefore, observations of metabolic responses obtained under culture conditions need to be interpreted with caution, they indicate what the tissue is doing under those specific conditions (which include cells adapting and dying).

Chen et al use pharmacological interventions are to the impact of various metabolic pathways on photoreceptor survival and "long term" metabolic changes. The dose and timing of these drug treatments are not examined though. It is also hard to know how these drugs penetrate the tissue and it needs to be validated that they intended targets are being accurately hit. These relatively long term treatments should be causing numerous downstream changes to metabolism, cell function and survival, which makes looking at a snap shot of metabolite levels hard to interpret. It would be more valuable to look at multiple time points after drug treatment, especially easy time points (closer to 1 hr). the authors use metabolite ratios to make conclusions about pathway activity. It would be more valuable to directly measure pathway activity by looking a metabolite production rates in the media and/or with metabolic tracers again in time scales closer to minutes and hours instead of days.

While the data is interesting and may give insights into some rod and cone specific metabolic susceptibility, more work is needed to validate these conclusions. Given the limitations of the model the authors have over interpreted their findings and the conclusions are not supported by the results. They need to either dramatically limit the scope of their conclusions or validate these hypotheses with additional models and tools.

Reviewer #3 (Public Review):

Summary:

The neural retina is one of the most energetically active tissues in the body and research into retinal metabolism has a rich history. Prevailing dogma in the field is that the photoreceptors of the neural retina (rods and cones) are heavily reliant on glycolysis, and as oxygen tension at the level of photoreceptors is very low, these specialized sensory neurons carry out aerobic glycolysis, akin to the Warburg effect in cancer cells. It has been found that this unique metabolism changes in many retinal diseases, and targeting disease-altered retinal metabolism may be a viable treatment strategy. The neural retina is composed of 11 different cell types, and many research groups over the past century have contributed to our current understanding of cell-specific metabolism of retinal cells. More recently, it has been shown in mouse models and co-culture of the mouse neural retina with human RPE cultures that photoreceptors are reliant on the underlying retinal pigment epithelium for supplying nutrients. Chen and colleagues add to this body of work by studying an ex vivo culture of the developing mouse retina that maintained contact with the retinal pigment epithelium. They exposed such ex vivo cultures to small molecule inhibitors of specific metabolic pathways, performing targeted metabolomics on the tissue and staining tissue with key metabolic enzymes to lay the groundwork for what metabolic pathways may be active in particular cell types of the retina. The authors conclude that rod and cone photoreceptors are reliant on different metabolic pathways to maintain their cell viability - in particular, that rods rely on oxidative phosphorylation and cones rely on glycolysis. Further, their data suggest multiple mechanisms whereby glycolysis may occur simultaneously with anapleurosis to provide abundant energy to photoreceptors. The data from metabolomics revealed several novel findings in retinal metabolism, including the use of glutamine to fuel the mini-Krebs cycle, the utilization of the Cahill cycle in photoreceptors, and a taurine/hypotaurine shuttle between the underlying retinal pigment epithelium and photoreceptors to transfer reducing equivalents from the RPE to photoreceptors. In addition, this study provides quantitative metabolomics datasets that can be compared across experiments and groups. The use of this platform will allow for rapid testing of novel hypotheses regarding the metabolic ecosystem in the neural retina.

Strengths:

The data on differences in susceptibility of rods and cones to mitochondrial dysfunction versus glycolysis provides novel hypothesis-generating conjectures that can be tested in animal models. The multiple mechanisms that allow anapleurosis and glycolysis to run side-by-side add significant novelty to the field of retinal metabolism, setting the stage for further testing of these hypotheses as well.

Weaknesses:

Almost all of the conclusions from the paper are preliminary, based on data showing enzymes necessary for a metabolic process are present and the metabolites for that process are also present. However, to truly prove whether these processes are happening (rather than speculation of the possibility they are happening), further experiments are necessary. As it currently stands, results from this study contradict results from other studies - in particular that cones, not rods, are most reliant of glycolysis. The authors attempt to address these contradictions, but without further experimentation, logical arguments carry only so much weight. At a minimum, the authors have argued that the small molecules they use are exquisitely specific for their intended targets, but validating results with a second small molecule that hits the same target but is structurally different would bolster their claims. Genetically knocking down the intended targets with interfering RNA technology would also be possible, as would explant cultures from knock-out animals. Without these studies to confirm target specificity, combined with the fact that conclusions from this study contradict existing studies in the literature, the results have to be categorized as speculative and hypothesis-generating rather than conclusive.

Author Response

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

We thank the reviewers and editors for their comments, as well as for the time dedicated to make useful suggestions that have contributed to improve the manuscript. We have responded to the concerns raised by the reviewers, and after that, we have also responded to the different points highlighted in the Recommendations for the authors:

Reviewer #1

While in vivo injury was used to assess regeneration from subsets of PNS neurons, different in vitro neurite growth or explant assays were used for further assessments. However, the authors did not assess whether the differential "regenerative" responses in vivo could be recapitulated in vitro. Such results will be important in interpreting the results.

We included a supplementary figure evaluating the neurite extension in vitro and updated the text accordingly.

Intriguingly, even in individual groups of PNS neurons, not all neurons regenerate to the same extent. It is known that the distance between the cell body and the lesion site affects neuronal injury responses. It would be interesting to test this in the observed regeneration.

Although it is true that the distance can affect the outcome, here we used a physiological model where all neurons are lesioned at the same point in the nerve. Not only distance is different for motoneurons, but also the microenvironment surrounding their somas and therefore the direct comparison of these neurons with sensory neurons is limited. We extended the discussion on this matter in the new manuscript.

Fig 1: The authors quantified the number of regenerating axons at two different time points. However, the total numbers of neurons/axons in each subset are different. The authors should use these numbers to normalize their regenerative axons.

Figure 1D shows the normalization of data from figure 1C (normalized against the number of control axons in each neuron type). This has been clarified in the text.

Fig 2-5: In explaining differential regeneration of individual groups of neurons, there are at least two possibilities: (1). Each group of neurons has different injury/regenerative responses; (2). The same set of injury/regenerative responses are differentially activated. Some data in this manuscript suggested the latter possibility. But some other data point in the opposite direction. It would be informative for the authors to analyze/discuss this further.

From our point of view, these two options can be considered differential response to injury and could be potentially used for the modulation of regeneration. However, if the second possibility is correct, the regenerative program could be more influenced by the time chosen to study the response. Given the importance of this, we added some discussion about this topic.

Fig 6: Is it possible to assess the regenerative effects of knockdown Med12 after in vivo injury?

It is possible, but it is out of the scope of this work. Here, we aimed to describe the regenerative response and validate our data by testing a potential target for specific regeneration. Future studies will focus on the modulation of this specific regeneration both in vitro and in vivo.

Reviewer #2

It seems that the most intriguing outcome of this paper revolves around the role of Med12 in nerve regeneration. The authors should prioritize this finding. Drawing a conclusion regarding Med12's role in proprioceptor regeneration based solely on this in vitro model may be insufficient. This noteworthy result requires further investigation using more animal models of nerve regeneration.

The main goal of this work was to compare the regenerative responses of different neuron subpopulations. We modulated Med12 to validate our data and the potential of our findings. Unfortunately, investigating in depth the role of Med12 in regeneration is out of the scope of this paper. For this reason, we did not prioritise this finding here. As this finding was striking, we strongly agree that the next step should be studying how it modulates regeneration.

One critique revolves around the authors' examination of only a single time point within the dynamic and continuously evolving process of regeneration/reinnervation. Given that this process is characterized by dynamic changes, some of which may not be directly associated with active axon growth during regeneration, and encompasses a wide range of molecular alterations throughout reinnervation, concentrating solely on a single time point could result in the omission of critical molecular events.

We agree that this is probably the main limitation of this study, as we discussed in the text. We chose 7 days postinjury as a standard time point widely described in literature and to have a correlate with our histological data. Although the main aim was to compare populations, analyzing an additional time point after injury could add valuable information.

Reviewer #3

No concerns were expressed by that reviewer.

Recommendations for the authors:

The authors should assess whether the differential "regenerative" responses in vivo could be recapitulated in vitro.

We included a supplementary figure evaluating the neurite extension in vitro and updated the text accordingly.

Optional:

It will be interesting to test if the distances between the cell body and the lesion site contribute to the observed differences in individual subsets of PNS neurons.

Figure 1D shows the normalization of data from figure 1C (normalized against the number of control axons in each neuron type). This has been clarified in the text.

Fig 2-5: In explaining differential regeneration of individual groups of neurons, there are at least two possibilities: (1). Each group of neurons has different injury/regenerative responses; (2). The same set of injury/regenerative responses are differentially activated. Some data in this manuscript suggested the latter possibility. But some other data point in the opposite direction. At least the authors should discuss these.

From our point of view, these two options can be considered differential response to injury and could be potentially used for the modulation of regeneration. However, if the second possibility is correct, the regenerative program could be more influenced by the time chosen to study the response. Given the importance of this, we added some discussion about this topic.

While the paper is technically well-executed, the conclusions and some of the findings appear to be incomplete and challenging to draw meaningful conclusions from. This manuscript presents some interesting findings, but the title is quite broad and may suggest that the authors have unveiled fundamental mechanisms explaining the varying regenerative abilities of peripheral axons. However, the results do not substantiate such a conclusion. Further comments and suggestions follow.

We eliminated the word “regenerative (response)” from the title, as it could lead to think that all changes seen in these neurons are related only to regeneration. We think that “Neuron-specific RNA-sequencing reveals different responses in peripheral neurons after nerve injury” highlights the differences between neurons that we found without misleading towards thinking that we described regenerative mechanisms in all neurons.

What's notably absent here is the validation of certain genes found with the ribosomes, especially those highlighted in the subsequent figures. The question arises as to whether the changes depicted in the figures align with changes in the DRGs in vivo. Is there concordance between the presence of these genes and their transcriptional changes? It would greatly enhance the study's value if the authors could show evidence of upregulation or downregulation of certain genes over time in tissue sections, utilizing techniques such as in situ hybridization or immunocytochemistry.

We selected some factors that were specifically upregulated in subsets of neurons to corroborated by immunohistochemistry these findings. Changes in the immunofluorescence of P75 in motoneurons and ATF2 in cutaneous mechanoreceptors, were evaluated in controls and animals that received a nerve crush one week before. Supplementary figures with the images have been added.

The authors discovered intriguing distinctions, such as the presence of specific signaling pathways unique to neurons projecting to muscle as opposed to those projecting to the skin. Among these pathways were those associated with receptor tyrosine kinases like VEGF, erbB, and neurotrophin signaling among others. The question now arises: do these pathways play a role in natural peripheral regeneration processes? To answer this, it is imperative to conduct in vivo studies. However, the authors employed an in vitro DRG neurite outgrowth assay to demonstrate that various types of neurons exhibit different responses to the presence of different neurotrophins. This does not reflect what actually happens in vivo. While neurotrophins indeed play a role in neuron survival and axon extension during development, their role in postnatal periods changes over time, and it remains unclear whether they play any role in the natural regenerative processes of the peripheral nerve. Therefore, this experiment may not be directly relevant in this case, especially during the early axon extension period of the regenerating axons. if the authors aim to establish a causal link with neurotrophin signaling, it becomes crucial to conduct in vivo experiments by manipulating the expression of key molecules like the receptors.

It has been widely described that different types of peripheral neurons have a differential expression of Trk receptors, even in the adult, and that these respond differentially to neurotrophins. In our study, we do not stablish a causal relationship between the expression of Trk and neurite extension, but instead we show (as many others) that distinct neurons respond differentially to these neurotrophins. The fact that in vivo studies fail to show a clear effect does not necessarily mean that neurotrophins are not specific. It might mean that their effect is not strong enough to be a useful guide in the complex microenvironment found after an injury. For instance, NGF acts on TrkA (present in some neurons), but in vivo it has been shown to accelerate the clearance of myelin debris in Schwann cells (Li et al., 2020), which could facilitate regeneration of all type of axons, masking any potential specific effect on the subtypes of neurons expressing TrkA. In contrast, in an in vitro setting on neuronal cultures, the specific neuronal effect can be more evident.

Additionally, it's worth noting that another paper utilizing the same methodology and experimental setup (PMID: 29756027, "Translatome Regulation in Neuronal Injury and Axon Regrowth" by Rozenbaum et al.) exists. Are there any significant differences or shared findings with that study?

This study shows the transcriptomic response after an injury 4, 12 and 24 hours after an injury in a very similar experimental setup. They focus on comparing the neuronal vs the glial response to the injury, using a Ribotag line that tags ribosomes from all neurons in the DRG rather than specific neuron subtypes. As the time postinjury (24h vs 7 days) and the cell types studied are different, we could not directly compare our results. We did see an upregulation in both datasets of previously described growth-associated genes (Jun, Atf3, Sox11, Sprr1a, Gal…). We included the article in the references for its relevance in the topic.

It would be helpful for readers to illustrate the finding of the fastest axon regeneration of nociceptors by showing fluorescence micrographs of the nerve samples in addition to the graphs shown in Fig. 1 C/D.

In figure 1B, we show fluorescence micrographs of the nerves 7 days postinjury. As explained in the results, we counted the number of axons at 2 distances from the injury, we did not analyse the fastest axon. This is due to technical reasons: 7 days after the injury the fastest axon has surpassed our evaluation point, which was the further distance that we could assess in our experimental setting in a consistent manner. If the reviewer thinks that we need to include more images from our evaluations (from 9 dpi for example), we could prepare a new figure.

The labeling in Fig. 2B is confusing. Is the CHAT immunoreactivity shown in the last panel illustrated by green or red signals? Is the red signal counterstaining with beta-tubulin?

The labelling was changed in the figure to increase clarity.

The references to the supplementary data throughout the manuscript are confusing. For example, where can the "Supp data 2" be found? (mention on p. 14 in the merged pdf file). Are they referring to the Excel spreadsheets?

We divided the supplementary material in supplementary figures/table (found in the pdf) and supplementary data. Supplementary data refers to excel spreadsheets found outside the pdf file. We hope this will be clearer after the final formatting of the article.

What does the following statement on p. 14 mean?: "The caveat in these analyses was that molecular classification by these approaches may be arbitrary, and not reflective of protein repurposing." This reviewer notes that these databases consider the fact that components participate in different pathways.

Indeed, we aimed to explain that many proteins participate in different pathways, and this is a limitation of the enrichment analysis. We modified the sentence in the text.

First paragraph on p. 15: The PPAR and AMPK pathways have much broader roles, and are not only "related to fatty acid metabolism". This factual inaccuracy should be corrected in the manuscript.

The sentence has been corrected.

The authors should consider showing increased TGF-beta signaling in their neurons after downregulation of Med12 given the previous implication of TGF-beta signaling in axon regeneration.

We tried to demonstrate the effect of our knockdown in TGF-beta pathway by analyzing the expression of typical targets from this pathway by qPCR in our cultures. However, we could not detect any difference. We think that this can have two explanations: (1) as only a few cells upregulate Med12 whereas many cells downregulate it, the effect is masked (presumably only proprioceptors will have a significant difference in this pathway and, thus, it would be very difficult to see the effect), or (2) Med12 is not exerting its effect through this pathway. We added a supplementary figure with these data and discussed it in the manuscript.

It would be helpful to eliminate typos and improve syntax/grammar/style.

We revised the text to improve style.