2,797 Matching Annotations
  1. Last 7 days
    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Manuscript number: RC-2022-01758

      Corresponding author(s): Harbison, Susan and Souto-Maior, Caetano

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

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

      1. General Statements [optional]

      This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

      We thank the reviewers for their time and care in evaluating our manuscript. They raise several important points, which we have addressed, resulting in a greatly improved manuscript. Please note that we numbered the comments from both reviewers for ease of reference, as we cross-referenced comments in some cases. Reviewer comments are in italics; our responses are provided in plain text.

      2. Point-by-point description of the revisions

      This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

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

      *Summary*:

      *The authors of this work generated a Sleep Advanced Intercross Population from 10 extreme sleeper Drosophila Genetics Reference Panel. This new out-bred population was subjected to a artificial selection with the aim of understanding the genes underlying the sleep duration differences between three populations: short-sleep, unselected, and long-sleep. Using analysis of variance the authors identified up to nearly 400 of genes that were significant selected over the various generations and showed opposite trends for long and short sleep, thus potentially relevant for the regulation of sleep duration. 85 of these genes were consistent between male and females sub-populations, suggesting a small number of genetic divergences may underlie sex-independent mechanisms of sleep.

      Given the time-course nature of the generational data obtained, the authors studied potential correlations and interactions between these 85 identified candidate genes. Initially, the authors used pairwise Spearman correlation, noticing how this method could not filter most of pairwise interaction (around 40% of all possibilities were significant). To overcome the linear limitations of the previous approach, the authors implemented a more complex, non-linear Gaussian process model able to account for pairwise interactions. This new approach was able to identify a smaller number of different, and potentially more informative, correlations between the candidate genes previously identified.

      Lastly, with genetic manipulations, the authors show in vivo that a subset of the candidate genes is causally related with the sleep duration as well as partially validating some of the correlation identified by their new model.

      The authors conclude that, given the non-linear and complex nature of biological systems, simplistic linear approaches may not suffice to fully capture underlying mechanisms of complex traits such as sleep.

      *Major comments*

      1. Most of the the work presented focus on the computational and statistical analysis of different populations submitted (or not) to a process of artificial selection for short or long sleep duration. As such, the amount of potentially relevant biological conclusions to be tested is mostly unfeasible. The authors already present additional experiments to partially support some, though not all, of their findings. Given the manuscript is written as a method innovation, these additional experiments illustrate the potential uses of the method described. *

      Our response: The reviewer raises a very important point, one that is at the very impetus of our work. We agree that it is not possible to test all combinations of genes in all contexts to determine whether they influence sleep or not. In contrast to the situation for circadian rhythms, where the core clock is controlled by just four genes, recent work has concluded that sleep is a set of complex traits influenced by large numbers of genes. Robust computational methods are needed to identify the complex interactions among genes. The current manuscript is a first step towards achieving this goal.

      *(OPTIONAL) However, since the one of the focuses of this work in identifying potential gene interactions, it would be interesting if the authors could test a "double knockout" and perhaps demonstrate evidence for epistasis between two of the identified genes. Having access to single mutants, this experiment should be realistic. However, I have no hands-on experience working with Drosophila and I am unable to accurately estimate the amount of resources and time such and experiment could take. My initial guess would be 3-6 months work should suffice. *

      Our response: The reviewer makes an interesting proposal. While such an experiment would provide some additional information, our method does not make any prediction about what a double knockout would do, either to the sleep phenotypes or to gene expression.

      2. In regards to the gene CG1304, it seems to be an important example used throughout the manuscript. It should be carefully re-analyzed as was considered for interaction analyses without showing opposite trends for short- and long-sleep populations (see minor comments on figure 2).

      Our response: We are not entirely certain that we understand the reviewer’s point. We note that significant genotype-by-selection-scheme interactions may not manifest as opposite trends and this is not what is being tested for significance. The likelihood ratio is a test for a significant effect of including sel x gen coefficients for both short and long schemes; therefore, GLM significance may mean that either one or the two selection schemes are significantly different from controls, not from each other. We could, for instance, apply three different tests: one (i) comparing between long and short flies; the second (ii) __comparing short flies to controls; and the third (iii) __comparing long to controls and find that the first test is significant — i.e. short is different from long — and that the two others are not — i.e. neither scheme is found to be different from controls. The opposite could also happen: short and long flies may not be different from each other, but with both being different from controls.

      Since we are interested in identifying differences of either to controls, our choice of statistical test is equivalent to performing tests (ii) __and (iii)__ without the need to perform and correct for multiple tests. While there are caveats to this choice (like all choices), linear model-based differential expression analysis has its own caveats, and has limited ability to pick up arbitrary trends, so it serves as a coarse-grained filter for large shifts since it’s too costly (computationally) to run the Gaussian process on 50 million pairwise combinations.

      *3. One major comment would be that the claim that the Gaussian process method is more sensitive and specific than simpler approaches, though intuitively understandable, does not seem to be fully correct from a strict statistical point of view, given the lack of a gold standard reference to compare if the new method is indeed picking more true positives/negatives. I would reconsider re-rephrasing such statement in the absence of a biologically relevant validation set. *

      Our response: We agree with the reviewer that there is no ‘gold standard’ reference data set with which to compare our findings. We have softened this language a bit in response, where it occurs in both the Abstract and the Results.

      Under Abstract, we changed “Our method not only is considerably more specific than standard correlation metrics but also more sensitive, finding correlations not significant by other methods” to “Our method appears to be not only more specific than standard correlation metrics but also more sensitive, finding correlations not significant by other methods.”

      Under Results, we changed “Therefore, computing correlations between genes using covariance estimates from the Gaussian Processes greatly increases specificity over direct correlations. Furthermore, the Gaussian processes are not only more specific but more sensitive…” to “Therefore, computing correlations between genes using covariance estimates from the Gaussian Processes appears to increase specificity over direct correlations. Furthermore, the Gaussian Processes appear to be more sensitive…”

      *4. Finally, the study appears to be well powered and it is clear that the authors were careful in their explanation of the statistical methods. However, I could not find the copy of the code/script used for the model. Without it, it would be very difficult to fully reproduce the results as both the language used (Stan) and the method itself are not common in the sleep research field. *

      Our response: We thank the reviewer for noticing this, and apologize for this oversight. The code used for analysis has been deposited in GitHub under: https://github.com/caesoma/Multiple-shifts-in-gene-network-interactions-shape-phenotypes-of-Drosophila-melanogaster.

      We have noted the script location in the Data Availability statement. We added a statement to read “All scripts used for the model have been deposited in Git Hub https://github.com/caesoma/Multiple-shifts-in-gene-network-interactions-shape-phenotypes-of-Drosophila-melanogaster.”

      * * *Minor comments* * 5. The statistical cut-off used for gene expression hierarchical GLMM after BH correction was of 0.001, which is 50 times more strict than the common 0.05. Could the authors comment on how this choice may impact the results compared to those available in the literature and on the rational for choosing such a value.*

      Our response: A FDR of 0.05 would increase the number of genes identified (3,544 for females; 1,136 for males, with 462 overlapping). The FDR of 0.001 is consistent with the lowest threshold typically used for gene expression data collected during other artificial selection experiments (Mackay et al., 2005; Morozova et al., 2007; Edwards et al., 2006), though thresholds as high as 0.20 have been used (Sorensen et al., 2007). We have added to the last statement to the Methods and Materials section under “Generalized Linear Model analysis of expression data” to read “Model p-values were corrected for multiple testing using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995), with significance defined at the 0.001 level, consistent with the lower threshold applied in other artificial selection studies (Mackay et al., 2005; Morozova et al., 2007; Edwards et al., 2006).”

      *6. Heritability calculations are not mentioned in the methods. Could it be useful to include a small paragraph? Could a small comment be done on the differences in h2 for the short sleep replicates which show ~10x difference? *

      Our response: We thank the reviewer for noticing this omission and apologize for the oversight. We have added the following statements to the Methods and Materials under “Quantitative genetic analyses of selected and correlated phenotypic responses.”

      “We estimated realized heritability h2 using the breeder’s equation:

      h2 = ΣR/ΣS

      where ΣR and ΣS are the cumulative selection response and differential, respectively (Falconer and Mackay, 1996). The selection response is computed as the difference between the offspring mean night sleep and the mean night sleep of the parental generation. The selection differential is the difference between the mean night sleep of the selected parents and the mean night sleep of the parental generation.”

      Additionally, we thank the reviewer for noticing the large difference in the realized heritability between the short sleeping population replicates; the heritability for replicate 1 is a typo and should be 0.169, not 0.0169. Hence, the heritabilities of both replicate populations are quite similar, i.e., 0.169 for replicate 1 and 0.183 for replicate 2. We have corrected this error in the Results.

      7. In regards to the model implementation, what would be the implications of not enforcing positive semi-definiteness on the co-variance matrix, given than that these are strictly positive semi-defined?

      Our response: All covariance matrices are by definition positive semi-definite (PSD), since they cannot yield negative values for the probabilities associated to them, so it would not be possible to relax that assumption generally. The only choice we could make would be on the number of genes included (M) in each multi-channel gaussian process model, and this in turn would by design enforce positive semi-definiteness on an matrix of size MN, (N being the number of generations). As noted in the appendix, “enforcing” positive semi-definiteness on smaller blocks of a larger 2D-array of covariances (which is not itself a covariance matrix) does not imply the latter is PSD and therefore seems like a softer constraint. In practice scaling up to a model where M >> 40 is not trivial from a computational and inference point of view, so the choice of smaller M is in a way imposed on us, and fortunately it is the less limiting one. We provide the appendix as a general clarification on the subtleties of Gaussian Processes, but a comprehensive assessment is beyond the multidisciplinary scope of this article and would require a narrower mathematical/statistical description in a standalone methodological article or technical note.

      1. *The methods mention that PCA projection were performed on the first 3 components, however only the first two are showed. *

      Our response: PCA was performed on 10 components, although the algorithms will commonly compute all components and return only the selected number. The variance of the third component is smaller than ~5% (that of the second PC). In practice PC1 is by itself enough to show the clear separation of expression per sex with ~65% of the variance; PC2 is in fact only shown to improve visualization. Plots of the remaining components will not show clear separation among samples as the variance explained is so small. We have corrected the Methods to indicate that PCA was performed on 10 components rather than 3.

      *9. Figure 1 refers to the mean night sleep time of the population. Could some measurement of variability (se or sd) be represented to provide a general idea of the distribution of the values? Additionally, the standard deviation of associated with the CVe estimates are mentioned but not showed explicitly. Could they maybe be added to the text as to illustrate how much such deviations were reduced? *

      Our response: We thank the reviewer for this comment. Including either the standard errors or standard deviations on the plot of the response to selection (Figure 1A) makes visualization unwieldy; thus we have added an additional supplemental table, Supplementary Table S15, that contains the mean night sleep, standard deviation, and number of flies measured for each generation in each replicate population. We also added a plot of the standard deviation in night sleep per generation to Supplemental Figure S2 (letter “Q” in the figure) so that the reduction over time in each population can be seen.

      Under “Data Availability,” We added the following: “Night sleep phenotypes per selection scheme/sex/generation/population replicate are listed in Table S15.”

      *10. Figure 2 shows the linear model fits for gene CG1304. I find this gene on the list of significant genes for both sexes (tables S5/6), but it does not seem to be one that shows opposite trend for short- and long-sleep (tables S7/8). Surprisingly, it shows up again on table S10! However, the text introducing the figure reads like this should be one of the 85 sex-independent genes. Would it be best to provide an example of what a significant gene looks like? *

      Our response: As mentioned in our response to comment #2 above, significance in the likelihood-ratio test does not imply opposite trends between long and short selection schemes, but between a model that includes specific slope coefficients for selection scheme by generation (both long and short) compared to a reduced model where the only slope is one associated to generation and therefore independent of selection scheme.

      11. *Figure 3 would be interesting to have both the GP correlations and the Spearman correlations to illustrate the methodological differences. I would be curious to see at least one pairwise expression scatter-plot as well just to see how they correlate in one plot. *

      __Our response: __Table S11 contains all (significant and nonsignificant) GP and Spearman values side-by-side for comparison. High correlations are likely to conform to the Spearman assumptions of a monotonic relationship; nevertheless, this will not be so for the majority of genes since the difference in the number of Spearman and GP-significant genes is tenfold or more, so it would be misleading to focus on individual-gene relationships without taking into consideration the transcriptome wide results for any method employed.

      We would like to stress that there is nothing particularly special about CG1304 in and of itself; furthermore, there are no “representative” genes or figures in this manuscript. Instead, CG1304 is chosen because its GLM and GP fits are illustrative of the limitations and capabilities of each model to pick up certain kinds of trends, and especially because it is especially instructive of how correlations arise from the GP model, which may not be intuitively clear to all readers.

      12. Figures 3S3/4 are described as showing single- and multi-channel models don't change substantially. Would this be expected and why?

      Our response: This is not necessarily expected, as scaling up from a single to a multi-channel model will add additional parameters as well as constraints, like positive the semi-definiteness mentioned in comment #7 above. If that seemed to have considerable impact on the fits it could challenge our assumption that the signal variance parameters estimated from the single-channel are good priors for the same parameters in the two-channel model (although this is not a hard constraint, so in the worst case the result could still only be a slight bias).

      *13. Having build different networks of pairwise associations of genes (projecting on a unified network as illustrated on figure 5), it could in interesting to compare the network topologies at a basic level such as node degrees, overlapping sub-networks, are they potentially scale free as previously described for biological systems, etc. *

      __Our response: __The reviewer makes an interesting point. Indeed summaries of the network could be useful information about the system level parameters, which are the main results of this paper. We now include the number of connections (i.e., the degree) to each gene in each of the four networks presented in Figure 5 in a new supplemental Table (Table S13). We also plot the distribution of node connectivity below. The distributions do not appear random (i.e., a normal distribution), and appear closer to a power-law or scale-free distribution. However, the small size and low average degree of these networks make a formal test unfeasible, and a recent study suggests that a log-normal distribution is in general more likely than a power-law distribution (Broido et al., Nat Comm, 2019), so we lack the evidence to claim that these networks are scale-free.

      We have added to the Results under “Gaussian Process model analysis uncovers nonlinear trends and specifically identifies covariance in expression between genes”: “Table S13 lists the number of connections (degrees) that each gene has with others in the network. The average number of connections for long-sleeper males was 2.6; the other three networks had average degrees of 2.0 or less (2.0 for long-sleeper females and short-sleeper males; 1.75 for short-sleeper females).”

      *14. On table S6 I noticed some gene symbols were loaded as dates (1-Dec) *

      Our response: We thank the reviewer for noticing this, the gene symbol is supposed to be dec. We have corrected this in Table S6 (now Table S7).

      1. *In results, the phenotypical response to artificial selection is sometimes described in minutes, other times in hours. Though this is an hurdle, it could make the values easier to compere if they were consistently formatted as minutes (hours). *

      Our response: We are unsure what the reviewer is referring to. We only see one sentence in which we used hours, and that was the concluding sentence under Results, “Phenotypic response to artificial selection.” The remainder of the manuscript refers to sleep times in minutes, phenotypes in all of the figures are plotted as minutes, and all of the supplemental material refers to times in minutes.

      16. *Over 99% of chains converged after three runs. Even though the reasons for the lack of convergence of these chains was not investigated, could this be a relevant effect? 1% of 3570 interactions is still 35 potential interactions. Do the non convergent chains relate with specific genes? *

      Our response: Bayesian MCMC inference is a stochastic algorithm, so there is a finite chance that any given run doesn’t converge, and that means that all eight parallel chains must converge and mix as measured by the stringent choice of R-hat metric being within 0.05 of unity. Relaxing the interval to 0.1 or 0.2 could still be acceptable, but we made the choice of a stringent threshold to avoid making interpretations on less-than-ideal runs. There is no evidence that there is any gene-specific problem, usually it would be one out of eight chains that would not mix well and throw off the diagnostic metrics (like relaxing the metrics, an acceptable approach could be accepting a run with 6-7 chains converging properly, but we decided to rerun all chains and only accept 100% convergence but accept a possible loss). Non-converging/nonmixing runs are likely to eventually do so, but since were are running tens of thousands of runs (3570 pairwise combinations × 3 schemes × 8 chains) a massively parallel implementation in a HPC cluster is required. Finally, seeing that 145 is ~4% of the total number of interactions, a naïve expectation would be that no more than one interaction would come out significant — while there is a chance that an interesting interaction was identified, the same can be said for potential false negatives computed using the GLM, which is a consequence of working at a high-throughput scale.

      17. The GO terms identified as significantly enriched after pvalue correction point to a clear association of the 85 genes identified with Serine proteases. Could this be discussed further to highlight biological findings of the work in the context of neuronal function or sleep regulation?

      Our response: The reviewer is correct, nine putative Serine proteases are significantly enriched among the 85 genes. All nine exhibit some expression in neurons and in epithelial cells, and all are expressed at the adult stage. The appearance of these enzymes is interesting given their role in proteolysis.

      We have updated the Discussion to read, “Interestingly, our Gene Ontology analysis identified nine genes from the 85-gene network with predicted Serine endopeptidase/peptidase/hydrolase activity: CG1304, CG10472, CG14990, CG32523, CG9676, grass, Jon65Ai, Jon65Aii, and Jon99Fii. All of these genes are expressed in neurons and epithelial cells, and all genes are expressed at the adult stage (Li et al., 2022). Serine proteases are a large group of proteins (257 in Drosophila) that perform a variety of functions (Cao and Jiang, 2018). Their predicted enzymatic activity suggests a putative role in proteolysis. This is an intriguing observation given pioneering work in mammals which suggested a role for sleep in exchanging interstitial fluid and metabolites between the brain and cerebral spinal fluid (Xie et al., 2013). Recent work demonstrated that a similar function is conserved in flies via vesicular trafficking through the fly blood-brain barrier (Artiushin et al., 2018). It would be interesting to determine whether these genes function in this process.”

      *18. Could the authors discuss the little overlap between males/females and shot/long sleep for 145 gene pairs identified after the MCMC runs. Similarly, how can the network differences be explained from a biological/evolutionary perspective? *

      Our response: The reviewer asks an interesting question. We did not detect sex-specific responses to artificial selection for long or short sleep in the present experiment. Yet differences in gene expression network pairs between males and females exist, and as the reviewer mentions, we also observed differences in network pairs between long sleepers and short sleepers. These differences reflect an inescapable conclusion: a given sleep duration phenotype can originate from more than one gene expression network configuration.

      19. *In the mutational analyses it is pointed out that CG12560 and Jon65Aii only affect females significantly. However, in the following sentence, the authors claim these two genes had the greatest effect on both sexes, which seems contradictory, at least in the way it is described. *

      Our response: Our wording may have been confusing, given that it came after a comment about Jon65Aii. Our exact statement was “Effects of the Minos insertions on night sleep duration were stronger in females than in males; when sexes were examined separately, only mutations in CG12560 and Jon65Aii affected male night sleep duration.” This was meant to convey that the effects of all Minos insertions were the same directionally for both males and females, but that only CG12560 and Jon65Aii insertions had statistically significant effects on each sex separately. We have re-worded this sentence to read “All Minos insertions had the same directional effect on night sleep for both males and females, but only the CG12560 and Jon65Aii insertions had statistically significant effects on night sleep on each sex separately.”

      20. *Maybe a small comment on how unchanged expression could lead to the observed phenotypical variation could help understanding how Minos mutations effects are biological mediated for those not familiar with the method. This seems to be the authors expectation so, could it be non-functional proteins or something else? *

      Our response: The reviewer raises an interesting point. We did not observe changes in gene expression for CG13793, Cyp6a16, or hiw compared to w1118 controls. Thus far, we have examined gene expression relative to the control for a single timepoint, and only in pooled whole flies. Differential gene expression between the Minos mutants and controls might occur at a different timepoint, or in a small set of key neurons that would be undetectable when comparing whole flies.

      We expand on this in Results, under “Mutational analyses confirms the role of candidate genes and interacting expression networks in sleep”: “Potential reasons for the lack of a significant change in gene expression in the remaining lines include: the position of the insertion within the targeted gene, which has variable effects on its expression; the relatively low statistical power of the experiment; confining our observation to a single timepoint during the day; or pooling whole flies, which might obscure gene expression changes occurring at a single-tissue level.”

      *21. The assumption that interacting genes would have their expression ratio changed by the Minos insertion would hold on situation where the affected gene causally interferes with the candidates expression. As far as I understand, causality cannot be inferred by the proposed method. Thus in a situation where both genes are co-regulated by a third factor, no change in expression ratio is to expected. How would the authors re-interpret their final result when considering this direct vs indirect interaction distinction? *

      Our response: Our method only gives us the hypothesis that two genes interact based on their correlation, and that is what we test using the Minos insertions. We do not as yet have a way to identify a third gene or factor that might be regulating the two. Given the number of genes affecting sleep, it is quite likely that there are such factors, but we can only report and test what we’ve observed. Any interpretation based on an arbitrary third factor would be purely speculative.

      **Referees cross-commenting**

      22. *I agree with Reviewer #2 comments which, to me, reads as generally pointing out the lack of biological interpretation of the results (and thus connecting this study with previous literature). Adding this component would make the manuscript well-rounded and attractive to a wider audience. *

      Our response: We agree with both reviewers that additional biological interpretation of the results would make the manuscript more attractive to a wider audience. Accordingly, we have added the following paragraph to the Discussion: “The genes we identify herein overlap and extend previous work. Of the 1,140 genes implicated in the generalized linear model, 151 (13.2 percent) overlapped with previous candidate gene, random mutagenesis, gene expression, and genome-wide association studies of sleep and circadian behavior in flies (Pegoraro e t al., 2022; Dissel et al., 2015; Seugnet et al., 2017; Shalaby et al., 2018; Thimgan et al., 2010, Thimgan et al., 2018, He et al., 2013; Mallon et al., 2014; Roessingh et al., 2019, Feng et al., 2018; Lee et al., 2021; Khoury et al., 2020; Wu et al., 2018; Harbison et al., 2013; Harbison et al., 2009; Harbison et al., 2017; Harbison et al., 2019). Notably, previous studies identified the genes CG17574, cry, dro, mip120, Mtk, NPFR1, pdgy, PGRP-LC, Shal, and vari as affecting sleep duration (Feng e t al., 2018, Dissel et al., 2015; Pegoraro et al., 2022; Thimgan et al., 2018; Mallon et al., 2014; He et al., 2013; Khoury et al., 2020; Harbison et al., 2013). Two genes, ringer and mip120, overlapped with our previous study of DNA sequence variation in flies selected for long and short sleep (Harbison et al., 2017). In that study we identified a polymorphism in an intron of ringer that changed in allele frequency with selection, with increases in the population frequency of the ‘G’ allele with increasing sleep, while the frequency of the ‘A’ allele increased with decreasing sleep. When the selective breeding procedure was relaxed, the frequency of the ‘G’ allele increased in short-sleeping populations, paralleling an increase in sleep (Souto-Maior et al., 2020). One possibility is that this polymorphism contributes to the changes in gene expression in ringer that we observed in the present study. Of the 85 genes common to both sexes that we used in the gene interaction networks, 11 (13 percent) appear in other studies of sleep: CG10444, CG2003, CG5142, CG6785, CG9114, CG9676, CR42646, hiw, NPFR1, Tie, and wb (He et al., 2013; Seugnet et al., 2017; Wu et al., 2018; Harbison et al., 2013). Thus, our study corroborates genes known to affect sleep, and identifies new candidate genes for sleep as well.”

      Reviewer #1 (Significance (Required)):

      *This study proposes the application of advanced non-linear methods to study complex traits such as sleep. As implemented, Gaussian Processes are able to identify non-linear correlations between two biological features (e.g. transcripts) over time (e.g. generations), representing an attempt to push the analytical methods available beyond the single gene paradigm. As such, more than the relevance of the biological results themselves, the authors focus on the explaining and illustrating the application of methodological advances obtained, and its relevance to obtain a better understanding of biological systems.

      However the mathematical principles required to understand the implemented method are not trivial and require advanced knowledge of machine learning and statistics. This is a potential barrier, though not an impediment, to its quick and wide adoption by the community. In addition, even if demonstrated to be a valid method when working with Drosophila, the resolution required to perform such a study may be difficult to obtain with other model systems, which would likely require further refinement of the statistical approach.

      The main audience interested in this work would be basic sleep researchers. However, this work is also related to the understanding gene selection over an artificial evolutionary process, thus evolutionary and developmental biologist may be also be interested. The methodology itself, already used in other fields of study, is a general statistical tool that could be adopted by a broad range of researchers for a diversity of topics. As such, I believe with this work, the authors will be able to stimulate the development and/or rethinking of the available analytical methods to study complex biological systems, though this would likely be done either in collaboration with the authors themselves or by a specific subset of researchers who regularly work with advanced mathematical, statistical and computational principles.

      (disclaimer) My mathematical formation does not reach the PhD level expertise that may be required to fully understand the methodology described. I have never personally worked with D. melonogaster or used Gaussian Processes in a professional setting. As such, I may not be able to fully evaluate/appreciate the more detailed technical aspects of this work.

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

      Souto-Mairo et al. reports phenotypic and genotypic effects of artificially selecting for short and long sleep in flies. They generated an impressive time-series dataset where one could examine genetic and phenotypic changes across time (generations, total 13 generations) in response to the selection pressure. The authors explored the relationships between pairs of genes in addition to just identifying potential candidate genes involved in the regulation of the amount of sleep.

      Major points:

      1. Harbison et al 2017: This study seems to be a continuation of Harbison et al 2017. There needs to be a clearer approach in the text (introduction?) in elucidating how this study is really advancing the findings of Harbison et al., 2017. Do the two studies use the same selection lines? If not, how are they different? If they are not different, what might cause the phenotypes evolving differently? For example, day sleep, day bout number do not respond to the selection pressure similarly in both studies etc. *

      Our response: We would like to emphasize that this study is not a continuation of the Harbison et al., PLoS Genetics, 2017 paper, where we examined the changes in DNA sequence during artificial selection, and it does not use the same selection lines. The fact that the two studies are different can be seen from an examination of Figure 1A of the current study and Figure 1A of the Harbison et al 2017 study. The trajectories of each population across generation are very different. Out of convenience, we used the same nomenclature to refer to the populations in both studies (L1, L2, S1, S2, etc.), and apologize if this is the source of the confusion. Both studies do originate from the same outbred population, however, and to get to the broader question that the reviewer is asking, should one expect to see the same correlated responses to selection for night sleep among selection lines originating from the same outbred population? The answer is no, not unless the selected trait and the responding trait have a genetic correlation of 1.0. We previously estimated the correlation between day sleep and night sleep to be between 0.29 - 0.38 and between day bout number and night sleep to be -0.05 (Harbison et al., 2013; Harbison et al. 2009). In the Harbison et al. 2017 study we noted that day sleep and day bout number had correlated responses to selection for night sleep, but neither have correlated responses in the current study. The relatively low genetic correlations between these two measures and night sleep explain why we do not see a consistent correlated response among studies.

      We didn’t really elaborate on these observations in the manuscript, and so have added to the Results under “Correlated response of other sleep traits to selection for night sleep” the following: “These correlated responses concur with previous observations we made in selected populations originating from the same outbred population for night sleep and night average bout length, and night sleep and sleep latency (Harbison et al., 2017). However, unlike the previous study, we did not see a correlated response between night sleep and day sleep, and night sleep and day bout number (Harbison et al., 2017). The lack of correlated response reflects the relatively low genetic correlations these two traits have with night sleep (Harbison et al., 2013; Harbison et al., 2009).”

      2. Zeitgeber Time (ZT) for RNA collection: It is puzzling that the study reports that the RNA was collected at 12 PM. I do not understand what this information means; especially in a project where one is working with sleep. The authors might want to report ZT. Also, why a particular ZT was chosen should be discussed. These genes are potential sleep-relevant genes - hence it is not too esoteric to think that the ZT of data collection matters a lot as some of them might be cycling. To get a more appropriate picture, multiple time points of data collection might be even better. The authors seem to have ignored this crucial aspect of a clock/sleep study - time of data collection and how time of data collection might shape your findings.

      Our response: We agree with the reviewer that it would be better to have multiple timepoints for collection, but this is difficult to implement in practice as it would require an additional 5,280 flies per generation (4 pools of 10 flies per sex per population) for 12 timepoints as recommended by Hughes et al., JBR, 2017. We mention collection time in the Methods and Materials because we are aware of the changes in gene expression over the circadian day. 12PM is the midpoint between the start of the lights-on and lights-off period (i.e., ZT6), and was chosen arbitrarily. We have added the ZT notation to the Methods and Materials for clarity.

      3. Short sleeping flies: Are there reports of flies sleeping this less? "We found 2,830 interactions; 8 of these were one of the 3,570 between the 85 genes, but none of them overlapped with the 145 gene pairs found to be different from controls. The gene interactions we observed may therefore be unique to extreme sleep." What is extreme sleep? How does this study then claim to have identified evolution of potential sleep-relevant gene expression for normal, physiologically relevant sleep?

      Our response: Our statement was not very well worded, and we thank the reviewer for noticing this. What we intended to say was that the lack of overlap between our data and a known protein-protein interaction database may due to the interactions being unique to sleep as opposed to some other complex trait. We have re-worded this statement to say “The gene interactions we observed may therefore be unique to sleep.”

      *Minor points:

      4. The article uses an unnecessarily defensive tone to establish their approach to understand underlying mechanisms of sleep 'better' than that of the others (in both introduction and discussion): "In spite the large amount of studies and data generated for many systems, identifying underlying processes is still very rare; this is clear indication that better methods are needed to obtain understanding of biological processes from data." The 'still very rare' part is just factually incorrect and misleading as far as sleep is concerned. Even if we just see Drosophila studies on sleep, there is a huge progress that's being made in terms of genes, neurons and circuits relevant for sleep: both in terms of baseline sleep as an output of the circadian clock and the rebound/homeostatic sleep. Most, if not all, of these elegant and pioneering studies from multiple, independent groups took approaches that did not require artificial selection regimes. As a substitution for their defense, the authors might attempt to present their findings in the context of the existing knowledge of sleep in flies. For example, what about genes already implicated in sleep? Do they show up in their analysis? For example, Sleepless, DATfmn, Sandman, AstA, AstA-receptor, Wide-awake etc. This could really help the manuscript.*

      Our response: We certainly did not intend for this statement to suggest that no progress had been made in the identification of genes and circuits for sleep, and we agree that elegant and pioneering approaches have made significant progress in our understanding of the phenomenon. Rather, we were thinking more in terms of fully described biochemical networks. To avoid this interpretation by other readers, we have altered the “still very rare” sentence in the Introduction to read: “Despite the large amount of studies and data generated for many systems, a full understanding of underlying processes has not yet been achieved…’

      We also agree with the reviewer that it would be helpful to put our work in the context of what is already known in flies. We have added the following paragraph to the Discussion to relate the work with previous work on sleep in flies: “The genes we identify herein overlap and extend previous work. Of the 1,140 genes implicated in the generalized linear model, 151 (13.2 percent) overlapped with previous candidate gene, random mutagenesis, gene expression, and genome-wide association studies of sleep and circadian behavior in flies (Pegoraro e t al., 2022; Dissel et al., 2015; Seugnet et al., 2017; Shalaby et al., 2018; Thimgan et al., 2010, Thimgan et al., 2018, He et al., 2013; Mallon et al., 2014; Roessingh et al., 2019, Feng et al., 2018; Lee et al., 2021; Khoury et al., 2020; Wu et al., 2018; Harbison et al., 2013; Harbison et al., 2009; Harbison et al., 2017; Harbison et al., 2019). Notably, previous studies identified the genes CG17574, cry, dro, mip120, Mtk, NPFR1, pdgy, PGRP-LC, Shal, and vari as affecting sleep duration (Feng e t al., 2018, Dissel et al., 2015; Pegoraro et al., 2022; Thimgan et al., 2018; Mallon et al., 2014; He et al., 2013; Khoury et al., 2020; Harbison et al., 2013). Two genes, ringer and mip120, overlapped with our previous study of DNA sequence variation in flies selected for long and short sleep (Harbison et al., 2017). In that study we identified a polymorphism in an intron of ringer that changed in allele frequency with selection, with increases in the population frequency of the ‘G’ allele with increasing sleep, while the frequency of the ‘A’ allele increased with decreasing sleep. When the selective breeding procedure was relaxed, the frequency of the ‘G’ allele increased in short-sleeping populations, paralleling an increase in sleep (Souto-Maior et al., 2020). One possibility is that this polymorphism contributes to the changes in gene expression in ringer that we observed in the present study. Of the 85 genes common to both sexes that we used in the gene interaction networks, 11 (13 percent) appear in other studies of sleep: CG10444, CG2003, CG5142, CG6785, CG9114, CG9676, CR42646, hiw, NPFR1, Tie, and wb (He et al., 2013; Seugnet et al., 2017; Wu et al., 2018; Harbison et al., 2013). Thus, our study corroborates genes known to affect sleep, and identifies new candidate genes for sleep as well.”

      Reviewer #2 (Significance (Required)):

      5. I believe that the authors should attempt to put this study in the context of what is already known in sleep in flies and how this study advances the knowledge. And how the knowledge generated by this study would help other sleep researchers, who, for obvious reasons, would like to employ techniques other than artificial selection and big data. The data is elegant. The work seems to be extremely laborious. Nonetheless, as it stands now, this manuscript is only very specific for an audience who work with artificial selection to understand underlying genetics of behavior. In fact, even within the fly sleep field, most people might not find this manuscript very useful.

      Our response: The reviewer may not have considered the wider application of this work. This framework is applicable to any data set of gene expression sampled across time, whether sampled across generation, as we did, or across the 24-hour circadian day, or sampled at other time intervals. We have added a statement to the Discussion to stress this fact: “The Gaussian Processes we apply herein have broad applications to other experimental designs, such as gene expression measured at varying time intervals over the circadian day, or time-based sampling of gene expression responses to drug administration.”

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Souto-Mairo et al. reports phenotypic and genotypic effects of artificially selecting for short and long sleep in flies. They generated an impressive time-series dataset where one could examine genetic and phenotypic changes across time (generations, total 13 generations) in response to the selection pressure. The authors explored the relationships between pairs of genes in addition to just identifying potential candidate genes involved in the regulation of the amount of sleep.

      Major points:

      1. Harbison et al 2017: This study seems to be a continuation of Harbison et al 2017. There needs to be a clearer approach in the text (introduction?) in elucidating how this study is really advancing the findings of Harbison et al., 2017. Do the two studies use the same selection lines? If not, how are they different? If they are not different, what might cause the phenotypes evolving differently? For example, day sleep, day bout number do not respond to the selection pressure similarly in both studies etc.
      2. Zeitgeber Time (ZT) for RNA collection: It is puzzling that the study reports that the RNA was collected at 12 PM. I do not understand what this information means; especially in a project where one is working with sleep. The authors might want to report ZT. Also, why a particular ZT was chosen should be discussed. These genes are potential sleep-relevant genes - hence it is not too esoteric to think that the ZT of data collection matters a lot as some of them might be cycling. To get a more appropriate picture, multiple time points of data collection might be even better. The authors seem to have ignored this crucial aspect of a clock/sleep study - time of data collection and how time of data collection might shape your findings.
      3. Short sleeping flies: Are there reports of flies sleeping this less? "We found 2,830 interactions; 8 of these were one of the 3,570 between the 85 genes, but none of them overlapped with the 145 gene pairs found to be different from controls. The gene interactions we observed may therefore be unique to extreme sleep." What is extreme sleep? How does this study then claim to have identified evolution of potential sleep-relevant gene expression for normal, physiologically relevant sleep?

      Minor points:

      The article uses an unnecessarily defensive tone to establish their approach to understand underlying mechanisms of sleep 'better' than that of the others (in both introduction and discussion): "In spite the large amount of studies and data generated for many systems, identifying underlying processes is still very rare; this is clear indication that better methods are needed to obtain understanding of biological processes from data." The 'still very rare' part is just factually incorrect and misleading as far as sleep is concerned. Even if we just see Drosophila studies on sleep, there is a huge progress that's being made in terms of genes, neurons and circuits relevant for sleep: both in terms of baseline sleep as an output of the circadian clock and the rebound/homeostatic sleep. Most, if not all, of these elegant and pioneering studies from multiple, independent groups took approaches that did not require artificial selection regimes. As a substitution for their defense, the authors might attempt to present their findings in the context of the existing knowledge of sleep in flies. For example, what about genes already implicated in sleep? Do they show up in their analysis? For example, Sleepless, DATfmn, Sandman, AstA, AstA-receptor, Wide-awake etc. This could really help the manuscript.

      Significance

      I believe that the authors should attempt to put this study in the context of what is already known in sleep in flies and how this study advances the knowledge. And how the knowledge generated by this study would help other sleep researchers, who, for obvious reasons, would like to employ techniques other than artificial selection and big data.

      The data is elegant. The work seems to be extremely laborious. Nonetheless, as it stands now, this manuscript is only very specific for an audience who work with artificial selection to understand underlying genetics of behavior. In fact, even within the fly sleep field, most people might not find this manuscript very useful.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      The authors of this work generated a Sleep Advanced Intercross Population from 10 extreme sleeper Drosophila Genetics Reference Panel. This new out-bred population was subjected to a artificial selection with the aim of understanding the genes underlying the sleep duration differences between three populations: short-sleep, unselected, and long-sleep. Using analysis of variance the authors identified up to nearly 400 of genes that were significant selected over the various generations and showed opposite trends for long and short sleep, thus potentially relevant for the regulation of sleep duration. 85 of these genes were consistent between male and females sub-populations, suggesting a small number of genetic divergences may underlie sex-independent mechanisms of sleep.

      Given the time-course nature of the generational data obtained, the authors studied potential correlations and interactions between these 85 identified candidate genes. Initially, the authors used pairwise Spearman correlation, noticing how this method could not filter most of pairwise interaction (around 40% of all possibilities were significant). To overcome the linear limitations of the previous approach, the authors implemented a more complex, non-linear Gaussian process model able to account for pairwise interactions. This new approach was able to identify a smaller number of different, and potentially more informative, correlations between the candidate genes previously identified.

      Lastly, with genetic manipulations, the authors show in vivo that a subset of the candidate genes is causally related with the sleep duration as well as partially validating some of the correlation identified by their new model.

      The authors conclude that, given the non-linear and complex nature of biological systems, simplistic linear approaches may not suffice to fully capture underlying mechanisms of complex traits such as sleep.

      Major comments

      Most of the the work presented focus on the computational and statistical analysis of different populations submitted (or not) to a process of artificial selection for short or long sleep duration. As such, the amount of potentially relevant biological conclusions to be tested is mostly unfeasible. The authors already present additional experiments to partially support some, though not all, of their findings. Given the manuscript is written as a method innovation, these additional experiments illustrate the potential uses of the method described.

      (OPTIONAL) However, since the one of the focuses of this work in identifying potential gene interactions, it would be interesting if the authors could test a "double knockout" and perhaps demonstrate evidence for epistasis between two of the identified genes. Having access to single mutants, this experiment should be realistic. However, I have no hands-on experience working with Drosophila and I am unable to accurately estimate the amount of resources and time such and experiment could take. My initial guess would be 3-6 months work should suffice.

      In regards to the gene CG1304, it seems to be an important example used throughout the manuscript. It should be carefully re-analyzed as was considered for interaction analyses without showing opposite trends for short- and long-sleep populations (see minor comments on figure 2)

      One major comment would be that the claim that the Gaussian process method is more sensitive and specific than simpler approaches, though intuitively understandable, does not seem to be fully correct from a strict statistical point of view, given the lack of a gold standard reference to compare if the new method is indeed picking more true positives/negatives. I would reconsider re-rephrasing such statement in the absence of a biologically relevant validation set.

      Finally, the study appears to be well powered and it is clear that the authors were careful in their explanation of the statistical methods. However, I could not find the copy of the code/script used for the model. Without it, it would be very difficult to fully reproduce the results as both the language used (Stan) and the method itself are not common in the sleep research field.

      Minor comments

      The statistical cut-off used for gene expression hierarchical GLMM after BH correction was of 0.001, which is 50 times more strict than the common 0.05. Could the authors comment on how this choice may impact the results compared to those available in the literature and on the rational for choosing such a value.

      Heritability calculations are not mentioned in the methods. Could it be useful to include a small paragraph? Could a small comment be done on the differences in h2 for the short sleep replicates which show ~10x difference?

      In regards to the model implementation, what would be the implications of not enforcing positive semi-definiteness on the co-variance matrix, given than that these are strictly positive semi-defined?

      The methods mention that PCA projection were performed on the first 3 components, however only the first two are showed.

      Figure 1 refers to the mean night sleep time of the population. Could some measurement of variability (se or sd) be represented to provide a general idea of the distribution of the values? Additionally, the standard deviation of associated with the CVe estimates are mentioned but not showed explicitly. Could they maybe be added to the text as to illustrate how much such deviations were reduced?

      Figure 2 shows the linear model fits for gene CG1304. I find this gene on the list of significant genes for both sexes (tables S5/6), but it does not seem to be one that shows opposite trend for short- and long-sleep (tables S7/8). Surprisingly, it shows up again on table S10! However, the text introducing the figure reads like this should be one of the 85 sex-independent genes. Would it be best to provide an example of what a significant gene looks like?

      Figure 3 would be interesting to have both the GP correlations and the Spearman correlations to illustrate the methodological differences. I would be curious to see at least one pairwise expression scatter-plot as well just to see how they correlate in one plot.

      Figures 3S3/4 are described as showing single- and multi-channel models don't change substantially. Would this be expected and why?

      Having build different networks of pairwise associations of genes (projecting on a unified network as illustrated on figure 5), it could in interesting to compare the network topologies at a basic level such as node degrees, overlapping sub-networks, are they potentially scale free as previously described for biological systems, etc.

      On table S6 I noticed some gene symbols were loaded as dates (1-Dec)

      In results, the phenotypical response to artificial selection is sometimes described in minutes, other times in hours. Though this is an hurdle, it could make the values easier to compere if they were consistently formatted as minutes (hours).

      Over 99% of chains converged after three runs. Even though the reasons for the lack of convergence of these chains was not investigated, could this be a relevant effect? 1% of 3570 interactions is still 35 potential interactions. Do the non convergent chains relate with specific genes?

      The GO terms identified as significantly enriched after pvalue correction point to a clear association of the 85 genes identified with Serine proteases. Could this be discussed further to highlight biological findings of the work in the context of neuronal function or sleep regulation?

      Could the authors discuss the little overlap between males/females and shot/long sleep for 145 gene pairs identified after the MCMC runs. Similarly, how can the network differences be explained from a biological/evolutionary perspective?

      In the mutational analyses it is pointed out that CG12560 and Jon65Aii only affect females significantly. However, in the following sentence, the authors claim these two genes had the greatest effect on both sexes, which seems contradictory, at least in the way it is described.

      Maybe a small comment on how unchanged expression could lead to the observed phenotypical variation could help understanding how Minos mutations effects are biological mediated for those not familiar with the method. This seems to be the authors expectation so, could it be non-functional proteins or something else?

      The assumption that interacting genes would have their expression ratio changed by the Minos insertion would hold on situation where the affected gene causally interferes with the candidates expression. As far as I understand, causality cannot be inferred by the proposed method. Thus in a situation where both genes are co-regulated by a third factor, no change in expression ratio is to expected. How would the authors re-interpret their final result when considering this direct vs indirect interaction distinction?

      Referees cross-commenting

      I agree with Reviewer #2 comments which, to me, reads as generally pointing out the lack of biological interpretation of the results (and thus connecting this study with previous literature). Adding this component would make the manuscript well-rounded and attractive to a wider audience.

      Significance

      This study proposes the application of advanced non-linear methods to study complex traits such as sleep. As implemented, Gaussian Processes are able to identify non-linear correlations between two biological features (e.g. transcripts) over time (e.g. generations), representing an attempt to push the analytical methods available beyond the single gene paradigm. As such, more than the relevance of the biological results themselves, the authors focus on the explaining and illustrating the application of methodological advances obtained, and its relevance to obtain a better understanding of biological systems.

      However the mathematical principles required to understand the implemented method are not trivial and require advanced knowledge of machine learning and statistics. This is a potential barrier, though not an impediment, to its quick and wide adoption by the community. In addition, even if demonstrated to be a valid method when working with Drosophila, the resolution required to perform such a study may be difficult to obtain with other model systems, which would likely require further refinement of the statistical approach.

      The main audience interested in this work would be basic sleep researchers. However, this work is also related to the understanding gene selection over an artificial evolutionary process, thus evolutionary and developmental biologist may be also be interested. The methodology itself, already used in other fields of study, is a general statistical tool that could be adopted by a broad range of researchers for a diversity of topics. As such, I believe with this work, the authors will be able to stimulate the development and/or rethinking of the available analytical methods to study complex biological systems, though this would likely be done either in collaboration with the authors themselves or by a specific subset of researchers who regularly work with advanced mathematical, statistical and computational principles.

      (disclaimer) My mathematical formation does not reach the PhD level expertise that may be required to fully understand the methodology described. I have never personally worked with D. melonogaster or used Gaussian Processes in a professional setting. As such, I may not be able to fully evaluate/appreciate the more detailed technical aspects of this work.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      We are very grateful to the reviewers for their constructive comments. In response to their critiques, we have made extensive modifications to the manuscript, including documenting new experiments and analyses, and improving data presentation. Here we provide a point-by-point response to the reviewers’ comments.

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

      Summary:

      It is well established that localization of oskar (osk) RNA in the Drosophila ovary proceeds in multiple steps. The first step depends upon dynein and results in delivery of osk into the oocyte. The second step involves kinesin-driven transport of osk to the oocyte posterior pole. The manuscript by Gáspár et al brings together several lines of evidence that support an tantagonistic relationship with respect to motor binding between two osk-interacting proteins, Egalitarian (Egl) and Staufen (Stau). As staufen RNA and protein accumulate in the oocyte, Egl dissociates from osk, down-regulating dynein and enabling the second stage of osk transport to begin.

      Major comments:

      In general the experimental results support the conclusions drawn, and the paper includes a strong mix of in vitro and in vivo approaches. Nevertheless I have a few concerns.

      (1)In Fig 1D it is apparent that stau KD increases the speed of both plus-end and minus-end runs to a highly significant degree, not just minus-end runs. The stimulating effect of loss of Stau on speed of plus-end runs is not mentioned in the text, and it perhaps muddies the argument that Stau is simply a negative regulator of dynein-dependent minus-end directed transport. This result needs to be explicitly discussed in the text.

      We thank the reviewer for this important comment. Indeed, our previous analysis of the overall population of oskar RNPs showed that plus-end-directed runs had increased velocity in the absence of Staufen (although the magnitude of the effect was considerably smaller than observed for minus-end-directed runs). The reviewer’s comment prompted us to analyze the effects on motility in more detail. In particular, we have now stratified the data based on the RNA content of the RNPs to control for effects of Staufen depletion on RNA copy number of the motile oskar RNPs. These analyses, which are documented in Fig 1B-F of the revised manuscript and discussed between lines 96-143, indicate that the previous velocity and run length data was somewhat confounded by the Staufen-depleted condition having a lower fraction of moving complexes with a large RNA content, which generally move more slowly. Accounting for this effect shows that impairing Staufen has no significant effect on plus-end-directed run lengths, whereas minus-end-directed run lengths are substantially increased. The velocity of runs is also specifically increased in the minus-end direction in the Staufen-depleted background for RNPs that have a relative RNA content of 1 or 2 units, which represent the majority of the RNP population in that genotype. Whilst RNPs with larger RNA content (2 relative units) do have significantly higher plus-end-directed velocity compared to the same category in the control, the effect is of much smaller magnitude than observed for minus-end-directed movements by this population. To help clarify these results, magnitudes of the effects are now shown in the new Fig. 1 E and F.

      These data strengthen the case that Staufen predominantly affects minus-end-directed motion. Given many documented examples of the interdependence of dynein and kinesin on bidirectional cargoes (Hancock et al. 2014), it is conceivable that the modest effects on plus-end-directed velocity for a subset of RNPs arise indirectly from the influence of Staufen on dynein activity. However, we agree with the reviewer that we should not rule out the alternative possibility that Staufen has additional roles in regulating oskar transport, including potentially modulating kinesin-1 directly. We have therefore added a section to the Discussion that covers this issue (lines 496-514).

      (2) I recognize the importance of quantitative imaging to rigorously measure small differences in localization patterns. Nevertheless I find the data in Fig 3 extremely difficult to interpret. Presumably there is standard deviation everywhere there is green signal, but the magenta signal that corresponds to SD is not visible in most places that are green. I suggest adding to Fig 3 a single representative image for each genotype to illustrate each localization pattern, as well as a much clearer explanation of the quantitative imaging data. Perhaps the quantitative images could be moved to a supplemental figure.

      Reviewer 2 also suggested that we include representative images in addition to the quantitative readout. We have now replaced the old Figure 3 with a new one showing representative examples of oskar distribution in the different genotypes and moved the quantitative images to the supplement (Figure S4). We have also improved the legends and labeling of this supplementary figure to add clarity.

      **Minor comments:**

      (1)Color/density scales should be added to Figs 1A and S1A, otherwise the yellow/white signal at the posterior could be interpreted as something other than high abundance.

      We thank the reviewer for spotting this. We have now added a color scale to the relevant figures.

      (2)In Fig 4A and 4C, I find it odd to have different halves of images photographed under different intensity settings and would prefer duplicate whole images.

      We used this layout to illustrate in the most compact way possible the (co)localization of the two RBPs and oskar RNA in the nurse cell and oocyte compartments, where signal intensities can differ dramatically. Following the reviewer’s comment, we now show whole images with different intensity settings (Figure 4 A, A’, C, C’).

      (3)The references to Fig 3G on page 13 should be corrected to Fig 4G.

      We thank the reviewer for spotting this error, which has now been corrected.

      Reviewer #1 (Significance (Required)):

      The paper represents a substantial advance over existing knowledge and it extends our understanding about how RNAs can shuttle between different motor proteins to achieve a localized pattern. However, the Mohr et al 2021 PLoS Genetics paper covers some of the same ground. As that paper has now been published for several months, I believe a revised version of this paper should discuss that other work more prominently, making it apparent where the two studies concur and where this study extends the conclusions of the other one. If there are any contradictions between the two, those should be made explicit as well.

      We had discussed the Mohr et al. study in our manuscript, which came out when our work was in preparation. Following the reviewer’s comment, we now address explicitly how our study differs from theirs and how our work extends their findings. The relevant paragraphs in the Discussion begin on lines 437 and 496. Briefly, a key point of difference is that Mohr et al. focused on the Transport and Anchoring Sequence (TAS) (including its ability to associate with Egl) and other Staufen recognition sites (SRSs) in oskar mRNA. Their study also includes an experiment examining the effect of Egl overexpression on oskar localization (as described in our original submission). In contrast, our study directly examines the interplay between the RBPs Staufen and Egl on oskar RNPs. We are the first to show that Staufen directly antagonizes dynein-based transport and that this is associated, at least in part, with an ability to impair Egl association with RNPs. Moreover, we provide insights into the in vivo role of Egl/BicD in recruitment vs activation of dynein on RNPs and how the activity of Staufen is coordinated in space and time via Egl-mediated delivery of stau mRNA, which constitutes a novel type of feed-forward mechanism. We do not believe there are any contradictions between the two studies.

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

      In this manuscript, Gáspár et al. investigated the molecular mechanisms underlying the switching of motors for osk mRNA transport in the Drosophila ovary: from dynein in the nurse cells to kinesin-1 in the oocyte. They demonstrated that it requires two RNA-binding proteins, Egalitarian (Egl) and Staufen (Stau) to achieve the posterior localization of osk mRNA in the oocyte. Their data show that Egl is responsible for the stau mRNA transport into the oocyte, while Stau protein inhibits Egl-dependent dynein transport in the oocyte. Thus, they proposed a feed-forward mechanism in which Egl transports mRNA encoding its own antagonist Stau into the oocyte and thus achieves the switch of the osk mRNA transport from dynein to kinesin-1.

      The antagonistic interaction between Egl and Staufen is well documented both in vitro and in vivo. All the results are carefully analyzed, but the data presentation is not reader-friendly. Overall, our main concern is about the role of Staufen in osk mRNA transport.

      **Here are specific points:**

      (1)According to the model, lack of Stau should result in failure of displacing Egl from the RNP complex and thus more dynein-driven transport in the oocyte. However, the increase of minus-end run length in stau-RNAi is very small (Figure 1E). It makes us wonder whether Stau is not a dominant inhibitor of Egl/dynein transport of osk RNPs. On the other hand, the speed increase of minus-end run in stau-RNAi is more dramatic than the run length (Figure 1D-1E). Does it mean that in stau-RNAi dynein-driven osk transport has a shorter duration of run? Additionally, in Figure 1D, there is a statistically-significant increase of plus-end-directed transport velocity in stau-RNAi. While the author did mention that in the results "analysis of the speed and length of oskar RNP runs in ooplasmic extracts indicated that Khc activity was not compromised upon staufen knock-down", it does not explain the increased velocity towards the plus-end.

      We thank the reviewer for these insightful comments.

      We and others (Zimyanin et al. 2008; Gaspar et al., 2014) have shown that there is only a small posterior-directed bias in oskar RNP transport in the wild-type ooplasm at mid-oogenesis. Thus, small increases in minus-end-directed transport parameters are expected to be sufficient for anterior mislocalization of a subset of RNPs, as is seen in stau mutants (note that we would not expect a dramatic increase in minus-end-directed motile properties in the stau RNAi condition, as a significant fraction of oskar RNA is targeted posteriorly). To allow the readers to better judge the magnitude of the effects, we now include the percentage change in mean velocity and run length values on the graphs (new Figure 1E and F).

      Regarding the reviewer’s question about the run duration, indeed it is shorter for the minus-end directed runs in the absence of Staufen. In the motor field, it is typical to present velocity and run length only because duration is dependent on these two parameters.

      Reviewer 1 also made a similar comment about plus-end directed velocity of RNPs. As we wrote in response to their comment, we have now stratified the data based on the RNA content of the RNPs to control for effects of Staufen depletion on RNA copy number of the motile oskar RNPs. These analyses, which are documented in Fig 1 B-F of the revised manuscript and discussed between lines 96-143, indicate that the previous velocity and run length data were somewhat confounded by the Staufen-depleted condition having a lower fraction of moving complexes with a large RNA content, which generally move more slowly. Accounting for this effect shows that impairing Staufen has no significant effect on plus-end-directed run lengths, whereas minus-end-directed run lengths are substantially increased. The velocity of runs is also increased only in the minus-end direction in the Staufen-depleted background for RNPs that have a RNA content of 1 or 2 relative units, which represent the majority of the RNP population in that genotype. Whilst RNPs with larger RNA content (2 relative units) do have significantly higher plus-end-directed velocity compared to the same category in the control, the effect is of much smaller magnitude than observed for minus-end-directed movement for this population.

      These data strengthen the case that Staufen predominantly affects minus-end-directed motion. Given many documented examples of the interdependence of dynein and kinesin on cargoes (Hancock et al., 2014), it is conceivable that the modest effects on plus-end-directed velocity arise indirectly due to the influence of Staufen on dynein activity. However, we agree with the reviewer that we should not rule out the alternative possibility that Staufen has additional roles in regulating oskar transport, including potentially modulating kinesin-1 activity directly. We have therefore added a section to the Discussion that covers this issue (lines 496-514).

      (2) What happened to osk mRNP transport in nurse cells with Staufen overexpression? The authors briefly mentioned that "GFP-Staufen overexpression has no major effect on the localization of oskar (Fig S1F-I)" on page 10. This is quite puzzling, as the authors propose that Staufen antagonized the Egl/dynein-driven transport. If the model holds true, we would expect to see that overexpression of Staufen causes less osk transport in nurse cells and thus less osk accumulated in the oocyte. Can the authors examine the osk mRNP transport in nurse cells in control and in GFP-Staufen overexpressing mutant and quantify the total amount of osk mRNA in the oocyte in control and after GFP-Staufen overexpression?

      We showed in the initial submission that strong overexpression of GFP-Staufen in early oogenesis (e.g. with osk-Gal4) disrupts oskar localization, including causing ectopic accumulation in the nurse cells (Fig S7F and G, now marked with arrowheads). Fig S1F-I, to which the reviewer refers, documents an experiment in which the expression of GFP-Staufen was directly driven by the maternal tubulin promoter (i.e. not through the UAS-Gal4 system; now indicated in Fig. S1F). We had assumed that the difference in behavior of the different GFP-Staufen transgenes was caused by the timing and the amount of overexpression – maternal Gal4 drivers are capable of very strong and, in the case of osk-Gal4, early expression of UAS transgenes. Prompted by the reviewer, we have now examined GFP-Staufen expression in these lines in more detail. This confirmed our previous assumptions about timing and levels of ectopic expression. We now included a new panel Fig S7I to document the expression of maternal tubulin promoter-driven GFP-Staufen and have updated the manuscript to include details about the mode of Staufen overexpression used in different experiments (lines 205, 408-417).

      (3)Is osk mRNP transport in the nurse cells affected by stau-RNAi? The authors showed the Khc association with oskar mRNPs in the nurse cells in Figure 1C. We hope they could quantify the velocity and run length of the osk mRNP particles in nurse cells and compare control with stau-RNAi.

      We have never succeeded in making squashes of nurse cells that maintain oskMS2 RNA transport. Therefore, we are unable to evaluate directional transport of oskar in these cells. However, Staufen does not accumulate to appreciable levels in the nurse cells, as shown by Little et al., 2015 and also Figure 4A and A’ (left panels). Moreover, we did not detect significant colocalization between Staufen and oskar in the nurse cells (Fig. 4B). Therefore, depletion of Staufen with RNAi is not expected to influence motility of oskar in this part of the egg chamber.

      (4)The kymograms of in vitro motility assays (Figure 2A and Figure S2) clearly showed two different moving populations, fast and slow. Did the authors include both types of events in their quantifications? What are the N numbers for each quantification? What do the dots mean in Figure 2B-2G? Does each dot represent a single track in the kymograph? If so, we believe that the sample sizes are too small for in vitro motility assay.

      For completeness, we did not exclude particles from our analysis based on their speed of movement. We have now made this point clear in an updated section of the Methods (lines 799-802), which provides additional information on particle inclusion criteria.

      We did document in the legends what the dots represent (values for single microtubules). We have now also included information on the number of complexes analyzed, which is 586-1341 single RNA particles or 1247-2207 single dynein particles per condition. These sample sizes are considerably larger than those used in most in vitro motility studies.

      (5)The in vitro motility assay showed that Staufen impairs dynein-driven transport of osk 5'-UTR (Figure 2). Based on these data, it is unclear whether the effect of Staufen is osk mRNA-dependent or Egl-dependent. We suggest performing the motility assay in the absence of osk 5'-UTR and Egl. Dynein, dynactin, and BicD should be sufficient to constitute the processive dynein complex in vitro. The addition of Staufen to the dynein complex will help to understand whether Staufen could directly affect dynein activity. We bring up this point because we noticed that the Staufen displacement of Egl in osk RNPs does not alter the amount of dynein complex associated (Figure 6), implying that Staufen inactivates dynein activity on the RNP complex, independently of Egl-driven dynein recruitment.

      We cannot look at transport of dynein in the presence of only dynactin and full-length BicD as BicD is not activated (and thus unable to effectively bind dynein and dynactin) without Egl and RNA (McClintock et al. 2018, Sladewski et al. 2018). However, the reviewer’s comment prompted us to investigate the effect of Staufen on dynein-dynactin motility that is stimulated by the constitutively active truncated mammalian BicD2, so called BicD2N (Schlager et al. 2014, McKenney et al. 2014). We find that Staufen partially inhibits DDB motility but not to the extent seen with the full-length BicD in the presence of Egl and RNA (new main figure panels 2H and I, and Figure S3). As stated between lines 187-188, these data suggest that Staufen inhibits both the activation of dynein-dynactin motility by BicD proteins, as well as stimulation of this event by Egl and RNA. This finding is also incorporated in a new section of the Discussion that covers possible roles of Staufen in addition to competing for Egl’s binding to RNA (between lines 496-514). We are very grateful to the reviewer for suggesting this approach, as it has provided significant new insight into Staufen’s function.

      (6)In Figure 4, it is hard to see any colocalization between GFP and osk mRNA. And the authors compared overexpressed Egl-GFP (driven by mat atub-Gal4 in mid-oogenesis) with Staufen-GFP under its endogenous promoter. An endogenous promoter-driven Egl-GFP would be much more appropriate for the comparison.

      Colocalization between GFP and oskar signals is seen as white in Fig. 4A and C. We have now added arrows to highlight a few examples of colocalization. The degree of colocalization was quantified in an unbiased fashion (shown in panels Fig 4B and D).

      Regarding the expression of Egl-GFP: it was driven directly by the aTub84B promoter and not by matTub-Gal4. Western blot analysis performed in response to the reviewer’s comment shows that Egl-GFP is expressed at similar levels to endogenous Egl in this line (new Fig. S5I).

      (7)In a recent publication (Mohr et al., 2021), a different model was proposed, in which Egl mediates transport, and Staufen facilitates the dissociation from the transport machinery for posterior anchoring. Although the authors referred to their paper in the discussion, they should acknowledge the differences and try to reconcile it (at least in the discussion).

      We now further discuss our work in the light of the findings by Mohr et al. (a request also made by Reviewer 1) (in paragraphs starting on lines 436 and 496). In our opinion, the data of Mohr et al. in fixed material cannot discriminate between effects of Staufen (or the TAS) on transport vs anchorage. In contrast, our dynamic imaging in vitro and ex vivo shows unambiguously that Staufen can modulate transport processes. As accumulation of RNA at the cortex is dependent on directional transport, we do not think it necessary to invoke a separate anchorage role of Staufen. We have now raised the possibility that transport and cortical localization are two facets of the same underlying process in the hope that this will stimulate further investigation (lines 455-459).

      (8)In the feed-forward model, Egl is required for the staufen mRNA transport from the nurse cells to the oocyte. Are Egl-GFP dots colocalized with staufen mRNAs in the nurse cells?

      We showed in Fig 7I of the original submission that Egl-GFP puncta are colocalized with stau mRNAs in nurse cells. Indeed, this is a key piece of evidence for our model. These data are now in Figure 7F.

      Furthermore, to our understanding, in this model, the translation of the staufen mRNA would be critical for the switching motors between dynein and kinesin-1. In this sense, staufen mRNA translation is either suppressed in the nurse cells or only activated in the oocytes. I think the authors should at least address this point in the discussion.

      This is another excellent suggestion. We have now included in the Discussion (from line 525) the point that Staufen translation may be suppressed during transit to the oocyte or that the protein may be translated en route but only build up to meaningful levels where the RNA is concentrated in the oocyte.

      **Minor points:**

      1)I hope the authors would show the osk mRNA localization in egl mutant in in individual stage 9 egg chambers. I can only find the osk mRNA in egl-RNAi early stage egg chambers (Figure 7E), in which osk mRNA still shows an accumulation in the oocyte, although to a much lesser extent compared to control. In another publication (Sanghavi et al., 2016), it seems that the knockdown of Egl by RNAi causes some retention of osk mRNA in the nurse cells; but there are still noticeable amount of osk mRNA in the oocyte (Figure 3A-B). We wonder whether the authors could quantify the amount of osk mRNA both in the nurse cells and in the oocyte of control and egl-RNAi. Also I wonder whether the authors could comment on fact that some osk mRNA transported into the oocyte. Could it be due to an egl-independent transport mechanism?

      egl null mutants do not reach stage 9 due to a defect in retention of oocyte fate, hence the use of egl RNAi in our study and the one by Sanghavi et al. Whilst we can’t rule out a (minor) Egl-independent mechanism for localizing oskar RNA in the oocyte, to date no other pathway has been implicated in the delivery of this or any other mRNA from the nurse cells. We favor a scenario in which residual oskar accumulation in the oocyte in egl RNAi egg chambers is due to incomplete depletion of Egl protein in the knockdown condition. We have noted this in the relevant figure legend and also clarify that the RNAi is a tool for knockdown in line 383 of the Results section.

      The below plot shows a quantification of oskar mRNA localization in egl and control RNAi egg chambers, which the reviewer was wondering about.

      In the egl RNAi egg-chambers, there is a significant increase in the mean signal intensity of oskar mRNA in the nurse cells, while oskar mRNA levels are substantially reduced in the oocyte, in line with the findings of Sanghavi et al., 2016.

      2)It is always nice to how the average distribution of osk mRNA (e.g., Figure 3, Figure S1, and Figure S3). But we recommend having a representative image of each genotype (a single egg) next to the average distribution. It will help the readers to better appreciate the differences among these genotypes.

      This suggestion was also made by Reviewer 1. We have added representative images to Figure 3 and moved the images depicting average distributions to the supplement (Fig S4). We have also improved the legend and labeling for Fig S4.

      3)The figure legends are overall hard to read and sometimes impossible to get information about the experiments (for example, Figure 4 legend). Can the authors improve their figure legends making them reader-friendly?

      We have edited the legends to make them clearer, including an extensive reworking of those for Figure 4. We thank the reviewer for encouraging us to do this.

      4)For moderate overexpression, the authors used P{matα4-GAL-VP16} (FBtp0009293). However, there are two different transgenic lines associated with FBtp0009293 (V2H and V37), which have slightly different expression levels. The authors should specify which line they used in the experiments.

      The matTub-Gal4 transgene we used in our study is inserted in the 2nd chromosome. We now mention this in the Methods section (line 567). We received this line from another lab many years ago, with no additional information provided.

      5) On page 13 "PCR on egg-chambers co-expressing Egl-GFP and either staufen RNAi or a control RNAi (white) in the germline (Fig 3G)", it should be Figure 4G.

      We apologize for this mistake, which has now been fixed.

      Reviewer #2 (Significance (Required)):

      see above

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

      Some additional experimental evidence is needed to solidify the conclusions and provide definitive support for this model, as discussed below.

      Biochemical experiments using UV crosslinking and GFP immunoprecipitation followed by quantitative PCR were performed to show that Staufen antagonizes the association of Egl with oskar mRNA in vivo. -The authors need to show the quantitative analysis, which was not present in the figure, specifically the effects of Staufen RNAi compared to control.

      These quantitative data, which are key for our model, were shown in the original submission (Fig 4G in the original and revised manuscript). We mistakenly called out the panel as 3G in the original submission. We apologize for this error, which has now been dealt with.

      Is the ability of Staufen to antagonize and displace Egl dependent on Staufen binding to Oscar RNA? Will a Staufen mutant that can't bind to RNA also displace Egl? Alternatively, the mechanism may be independent of RNA binding and perhaps due to protein-protein interactions.

      While the details of how Staufen displaces Egl are certainly an interesting topic for future research, we consider that addressing this goes well beyond the scope of this study, which already covers a lot of ground. Staufen contains four double stranded RNA-binding domains, and deleting or mutating all of these domains is likely to interfere with overall folding of Staufen, thus confounding the interpretation of the results.

      As an alternative approach to elucidating RNA-dependent vs RNA-independent roles of Staufen, we have now assessed the effect of the protein on in vitro motility of dynein-dynactin complexes formed in the presence of a constitutively active truncation of mammalian BicD2 (BicD2N). We find that Staufen partially inhibits motility of these ‘DDB’ complexes but not to the extent seen with the full length BicD in the presence of Egl and RNA (new Fig 2H, I and S3). As stated in the manuscript (lines 187-188) these data suggest that Staufen inhibits both the activation of dynein-dynactin motility by BicD proteins, as well as stimulation of this event by Egl and RNA. We believe these experiments provide significant new insight into Staufen’s function. This finding is also incorporated into a new section of the Discussion dealing with potential roles of Staufen in addition to displacing Egl from RNPs.

      A key question addressed is how does Staufen play a role in directing Oscar RNA localization to the posterior pole. The spatiotemporal control of Staufen at stage 9 seems to be a critical step. A number of experiments are performed to show that Staufen RNA enters the oocyte and accumulates to anterior pole through a process dependent on Egl (Fig. 7).

      -Definitive evidence is needed to show the role of 3'UTR of Stau and Egl binding. As it stands now, no evidence is presented to prove that delivery of staufen RNA via Egl, rather than dumping of Staufen protein into oocytes is the necessary trigger for the switch. It is well known that Staufen protein is also transported through ring canals to deliver Staufen into oocytes. There is no need to invoke an additional mechanism of Egl mediated staufen mRNA delivery. A key experiment is to perturb the Egl interaction with staufen 3'UTR and show this is a necessary component to impact oscar. Related to this comment, they should first perform biochemistry IP and PCR to demonstrate association of Egl with staufen RNA, and then somehow perturb this interaction to assess effects on oscar RNA localization. For example, is the 3'UTR of staufen RNA necessary for this mechanism? What if staufen RNA was ectopically localized in some inappropriate manner, for example localized to posterior pole? Would this prevent the switch of oscar RNA to move to posterior pole? The key question is: is it necessary that translation of Stau be coupled to Egl in order to drive the switch.

      Mapping of the Egl-binding site in stau mRNA is a major undertaking requiring the production and evaluation of multiple new transgenic fly lines. We feel that this would constitute an entirely new study. Moreover, multiple lines of evidence already support a functional interaction between Egl and stau mRNA, notably the presence of Egl on stau RNPs (previously Fig. 7I, now Fig. 7F), the strongly impaired accumulation of stau mRNA in the oocyte of egl RNAi egg chambers, and the ability of Egl overexpression to reposition a subset of the stau mRNA population at the anterior cortex.

      We have now performed new experiments and analyses to test the alternative hypothesis that Staufen protein is transported into the oocyte in the absence of stau mRNA transport. We find that disrupting Egl function with RNAi impairs localisation of both stau mRNA and protein in the proto-oocyte (new Figure 7A-D). As Egl has no known function in protein transport, these data argue against an RNA-independent mechanism for Staufen protein delivery. Moreover, we showed that both stau mRNA and Staufen are enriched in early oocytes lacking oskar mRNA, the main target of Staufen protein in the female germline. This result shows that Staufen protein is not appreciably transported from the nurse cells to the oocyte by hitchhiking on its RNA targets.

      Whilst Mhlanga et al. 2009 did report transport of large GFP-Staufen particles through ring canals, the line used (matTub4>GFP-Staufen from the St Johnston lab, which was also used for our rescue experiments) is known to make protein aggregates which is not the case for the endogenous protein (Zimyanin et al., 2008 and our new Figures 7B and S7E-I) and are therefore likely to be artefactual. Neither we, nor previous studies (Little et al., NCB, 2015), detected endogenous Staufen protein in nurse cells.

      Finally, the reviewer asks if coupling Staufen translation to Egl-mediated enrichment of stau mRNA in the oocyte is important: we showed in the original submission that strong overexpression of GFP-Staufen by Gal4 drivers leads to mislocalization of Staufen in the nurse cells of early egg-chambers, presumably due to saturation of the Egl-based transport machinery. In these egg-chambers, we observed defects in RNA enrichment in the primordial oocyte and defects in oogenesis, consistent with the need to exclude Staufen protein from the nurse cells.

      These findings are now presented in new panels of the updated Figures 7 and S7, with the corresponding section of the manuscript revised accordingly (lines 408-417). We think that altogether these lines of evidence strongly support our model that Egl transports stau mRNA into the developing oocyte and that this process is pivotal for oskar RNA localization.

      **Minor comments**

      "Substantially more oskar mRNA was co-immunoprecipitated with Egl-GFP from extracts of egg-chambers expressing staufen RNAi compared to the control (Fig 3G). -This data is not shown in 3G, but rather only in Fig. S4H which needs quantitative analysis shown.

      This point stems from us calling out the wrong panel in the first submission; this has now been addressed, as described above. We apologize for the error.

      "Addition of recombinant Staufen to the Egl, BicD, dynein and dynactin assembly mix significantly reduced the number of oskar mRNA transport events (Fig. 2A and B)."

      -In Fig. 2A, the Y axis shows velocity not number of transport events

      Fig 2A is a kymograph that is representative of the overall effect, where the Y-axis represents time. The reviewer may be referring to Fig 2B but this shows the frequency of processive oskar RNA movements (expressed as ‘number / micron / minute’), not velocity (micron/minute).

      Fig. 3. - This is very unclear figure as to what is being shown. More details are needed to explain the figure, and add arrows to help reader note what is being described.

      We have changed this figure to show representative images of individual egg chambers, as requested by the other two reviewers. The original Fig 3 is now moved to the Supplement as Fig S4. We have added arrows to the figure to indicate the anterior mislocalization of oskar mRNA and edited the legend for clarity.

      Staufen may also be required for the efficient release of the mRNA from the anterior cortex. This may reflect a role of Staufen in the coupling of the mRNA to the kinesin-dependent posterior transport pathway. This could be discussed as another aspect of the inhibition of dynein and handoff to kinesin.

      This is an interesting idea but it does not fit with our observation that Staufen depletion does not alter the association of oskar RNPs with kinesin-1 (originally Fig. 1C, now Fig. 1D). We do, however, now include in the Discussion a section on other ways, in addition to promoting Egl disassociation, that Staufen might orchestrate oskar mRNA transport.

      Reviewer #3 (Significance (Required)):

      This elegant manuscript by Gaspar et al provides new insight into the spatiotemporal regulation of Staufen mediated localization of oscar mRNA to the posterior pole in Drosophila oocytes. Here the authors demonstrate the competitive displacement of the RNA binding protein Egalitarian, which antagonizes dynein dependent localization at the anterior pole. This work done in this well characterized model of mRNA localization in Drosophila oocytes has broader implications for how the bidirectional transport of mRNAs is regulated in other polarized and highly differentiated cells, where very little is know about how mRNA transport direction might be regulated by opposing activities of kinesin and dynein motors. The strengths of this study are the integration of microscopy, biochemisty and genetic mutants to provide very nice experimental support for the two major aspects to the proposed model: 1) the competition between Staufen and Egl on oscar RNA which affects localization, 2) evidence for Egl mediated localization of staufen RNA into the oocyte as a key trigger for competitive displacement to bias localization of oscar RNA via kinesin. However, some additional experimental evidence is needed to solidify the conclusions and provide definitive support for this model, as discussed in other section.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Some additional experimental evidence is needed to solidify the conclusions and provide definitive support for this model, as discussed below.

      Biochemical experiments using UV crosslinking and GFP immunoprecipitation followed by quantitative PCR were performed to show that Staufen antagonizes the association of Egl with oskar mRNA in vivo. -The authors need to show the quantitative analysis, which was not present in the figure, specifically the effects of Staufen RNAi compared to control.

      Is the ability of Staufen to antagonize and displace Egl dependent on Staufen binding to Oscar RNA? Will a Staufen mutant that can't bind to RNA also displace Egl? Alternatively, the mechanism may be independent of RNA binding and perhaps due to protein-protein interactions.

      A key question addressed is how does Staufen play a role in directing Oscar RNA localization to the posterior pole. The spatiotemporal control of Staufen at stage 9 seems to be a critical step. A number of experiments are performed to show that Staufen RNA enters the oocyte and accumulates to anterior pole through a process dependent on Egl (Fig. 7). -Definitive evidence is needed to show the role of 3'UTR of Stau and Egl binding. As it stands now, no evidence is presented to prove that delivery of staufen RNA via Egl, rather than dumping of Staufen protein into oocytes is the necessary trigger for the switch. It is well known that Staufen protein is also transported through ring canals to deliver Staufen into oocytes. There is no need to invoke an additional mechanism of Egl mediated staufen mRNA delivery. A key experiment is to perturb the Egl interaction with staufen 3'UTR and show this is a necessary component to impact oscar. Related to this comment, they should first perform biochemistry IP and PCR to demonstrate association of Egl with staufen RNA, and then somehow perturb this interaction to assess effects on oscar RNA localization. For example, is the 3'UTR of staufen RNA necessary for this mechanism? What if staufen RNA was ectopically localized in some inappropriate manner, for example localized to posterior pole? Would this prevent the switch of oscar RNA to move to posterior pole? The key question is: is it necessary that translation of Stau be coupled to Egl in order to drive the switch.

      Minor comments

      "Substantially more oskar mRNA was co-immunoprecipitated with Egl-GFP f rom extracts of egg-chambers expressing staufen RNAi compared t o t he control (Fig 3G). -This data is not shown in 3G, but rather only in Fig. S4H which needs quantitative analysis shown.

      "Addition of recombinant Staufen to the Egl, BicD, dynein and dynactin assembly mix significantly reduced the number of oskar mRNA transport events (Fig. 2A and B)."

      -In Fig. 2A, the Y axis shows velocity not number of transport events

      Fig. 3. - This is very unclear figure as to what is being shown. More details are needed to explain the figure, and add arrows to help reader note what is being described.

      Staufen may also be required for the efficient release of the mRNA from the anterior cortex. This may reflect a role of Staufen in the coupling of the mRNA to the kinesin-dependent posterior transport pathway. This could be discussed as another aspect of the inhibition of dynein and handoff to kinesin.

      Significance

      This elegant manuscript by Gaspar et al provides new insight into the spatiotemporal regulation of Staufen mediated localization of oscar mRNA to the posterior pole in Drosophila oocytes. Here the authors demonstrate the competitive displacement of the RNA binding protein Egalitarian, which antagonizes dynein dependent localization at the anterior pole. This work done in this well characterized model of mRNA localization in Drosophila oocytes has broader implications for how the bidirectional transport of mRNAs is regulated in other polarized and highly differentiated cells, where very little is know about how mRNA transport direction might be regulated by opposing activities of kinesin and dynein motors. The strengths of this study are the integration of microscopy, biochemisty and genetic mutants to provide very nice experimental support for the two major aspects to the proposed model: 1) the competition between Staufen and Egl on oscar RNA which affects localization, 2) evidence for Egl mediated localization of staufen RNA into the oocyte as a key trigger for competitive displacement to bias localization of oscar RNA via kinesin. However, some additional experimental evidence is needed to solidify the conclusions and provide definitive support for this model, as discussed in other section.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this manuscript, Gáspár et al. investigated the molecular mechanisms underlying the switching of motors for osk mRNA transport in the Drosophila ovary: from dynein in the nurse cells to kinesin-1 in the oocyte. They demonstrated that it requires two RNA-binding proteins, Egalitarian (Egl) and Staufen (Stau) to achieve the posterior localization of osk mRNA in the oocyte. Their data show that Egl is responsible for the stau mRNA transport into the oocyte, while Stau protein inhibits Egl-dependent dynein transport in the oocyte. Thus, they proposed a feed-forward mechanism in which Egl transports mRNA encoding its own antagonist Stau into the oocyte and thus achieves the switch of the osk mRNA transport from dynein to kinesin-1.

      The antagonistic interaction between Egl and Staufen is well documented both in vitro and in vivo. All the results are carefully analyzed, but the data presentation is not reader-friendly. Overall, our main concern is about the role of Staufen in osk mRNA transport.

      Here are specific points:

      (1)According to the model, lack of Stau should result in failure of displacing Egl from the RNP complex and thus more dynein-driven transport in the oocyte. However, the increase of minus-end run length in stau-RNAi is very small (Figure 1E). It makes us wonder whether Stau is not a dominant inhibitor of Egl/dynein transport of osk RNPs. On the other hand, the speed increase of minus-end run in stau-RNAi is more dramatic than the run length (Figure 1D-1E). Does it mean that in stau-RNAi dynein-driven osk transport has a shorter duration of run? Additionally, in Figure 1D, there is a statistically-significant increase of plus-end-directed transport velocity in stau-RNAi. While the author did mention that in the results "analysis of the speed and length of oskar RNP runs in ooplasmic extracts indicated that Khc activity was not compromised upon staufen knock-down", it does not explain the increased velocity towards the plus-end.

      (2)What happened to osk mRNP transport in nurse cells with Staufen overexpression? The authors briefly mentioned that "GFP-Staufen overexpression has no major effect on the localization of oskar (Fig S1F-I)" on page 10. This is quite puzzling, as the authors propose that Staufen antagonized the Egl/dynein-driven transport. If the model holds true, we would expect to see that overexpression of Staufen causes less osk transport in nurse cells and thus less osk accumulated in the oocyte. Can the authors examine the osk mRNP transport in nurse cells in control and in GFP-Staufen overexpressing mutant and quantify the total amount of osk mRNA in the oocyte in control and after GFP-Staufen overexpression?

      (3)Is osk mRNP transport in the nurse cells affected by stau-RNAi? The authors showed the Khc association with oskar mRNPs in the nurse cells in Figure 1C. We hope they could quantify the velocity and run length of the osk mRNP particles in nurse cells and compare control with stau-RNAi.

      (4)The kymograms of in vitro motility assays (Figure 2A and Figure S2) clearly showed two different moving populations, fast and slow. Did the authors include both types of events in their quantifications? What are the N numbers for each quantification? What do the dots mean in Figure 2B-2G? Does each dot represent a single track in the kymograph? If so, we believe that the sample sizes are too small for in vitro motility assay.

      (5)The in vitro motility assay showed that Staufen impairs dynein-driven transport of osk 5'-UTR (Figure 2). Based on these data, it is unclear whether the effect of Staufen is osk mRNA-dependent or Egl-dependent. We suggest performing the motility assay in the absence of osk 5'-UTR and Egl. Dynein, dynactin, and BicD should be sufficient to constitute the processive dynein complex in vitro. The addition of Staufen to the dynein complex will help to understand whether Staufen could directly affect dynein activity. We bring up this point because we noticed that the Staufen displacement of Egl in osk RNPs does not alter the amount of dynein complex associated (Figure 6), implying that Staufen inactivates dynein activity on the RNP complex, independently of Egl-driven dynein recruitment.

      (6)In Figure 4, it is hard to see any colocalization between GFP and osk mRNA. And the authors compared overexpressed Egl-GFP (driven by mat atub-Gal4 in mid-oogenesis) with Staufen-GFP under its endogenous promoter. An endogenous promoter-driven Egl-GFP would be much more appropriate for the comparison.

      (7)In a recent publication (Mohr et al., 2021), a different model was proposed, in which Egl mediates transport, and Staufen facilitates the dissociation from the transport machinery for posterior anchoring. Although the authors referred to their paper in the discussion, they should acknowledge the differences and try to reconcile it (at least in the discussion).

      (8)In the feed-forward model, Egl is required for the staufen mRNA transport from the nurse cells to the oocyte. Are Egl-GFP dots colocalized with staufen mRNAs in the nurse cells? Furthermore, to our understanding, in this model, the translation of the staufen mRNA would be critical for the switching motors between dynein and kinesin-1. In this sense, staufen mRNA translation is either suppressed in the nurse cells or only activated in the oocytes. I think the authors should at least address this point in the discussion.

      Minor points:

      1)I hope the authors would show the osk mRNA localization in egl mutant in in individual stage 9 egg chambers. I can only find the osk mRNA in egl-RNAi early stage egg chambers (Figure 7E), in which osk mRNA still shows an accumulation in the oocyte, although to a much lesser extent compared to control. In another publication (Sanghavi et al., 2016), it seems that the knockdown of Egl by RNAi causes some retention of osk mRNA in the nurse cells; but there are still noticeable amount of osk mRNA in the oocyte (Figure 3A-B). We wonder whether the authors could quantify the amount of osk mRNA both in the nurse cells and in the oocyte of control and egl-RNAi. Also I wonder whether the authors could comment on fact that some osk mRNA transported into the oocyte. Could it be due to an egl-independent transport mechanism?

      2)It is always nice to how the average distribution of osk mRNA (e.g., Figure 3, Figure S1, and Figure S3). But we recommend having a representative image of each genotype (a single egg) next to the average distribution. It will help the readers to better appreciate the differences among these genotypes.

      3)The figure legends are overall hard to read and sometimes impossible to get information about the experiments (for example, Figure 4 legend). Can the authors improve their figure legends making them reader-friendly?

      4)For moderate overexpression, the authors used P{matα4-GAL-VP16} (FBtp0009293). However, there are two different transgenic lines associated with FBtp0009293 (V2H and V37), which have slightly different expression levels. The authors should specify which line they used in the experiments.

      5)On page 13 "PCR on egg-chambers co-expressing Egl-GFP and either staufen RNAi or a control RNAi (white) in the germline (Fig 3G)", it should be Figure 4G.

      Significance

      see above

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      It is well established that localization of oskar (osk) RNA in the Drosophila ovary proceeds in multiple steps. The first step depends upon dynein and results in delivery of osk into the oocyte. The second step involves kinesin-driven transport of osk to the oocyte posterior pole. The manuscript by Gáspár et al brings together several lines of evidence that support an antagonistic relationship with respect to motor binding between two osk-interacting proteins, Egalitarian (Egl) and Staufen (Stau). As staufen RNA and protein accumulate in the oocyte, Egl dissociates from osk, down-regulating dynein and enabling the second stage of osk transport to begin.

      Major comments:

      In general the experimental results support the conclusions drawn, and the paper includes a strong mix of in vitro and in vivo approaches. Nevertheless I have a few concerns.

      (1)In Fig 1D it is apparent that stau KD increases the speed of both plus-end and minus-end runs to a highly significant degree, not just minus-end runs. The stimulating effect of loss of Stau on speed of plus-end runs is not mentioned in the text, and it perhaps muddies the argument that Stau is simply a negative regulator of dynein-dependent minus-end directed transport. This result needs to be explicitly discussed in the text.

      (2)I recognize the importance of quantitative imaging to rigorously measure small differences in localization patterns. Nevertheless I find the data in Fig 3 extremely difficult to interpret. Presumably there is standard deviation everywhere there is green signal, but the magenta signal that corresponds to SD is not visible in most places that are green. I suggest adding to Fig 3 a single representative image for each genotype to illustrate each localization pattern, as well as a much clearer explanation of the quantitative imaging data. Perhaps the quantitative images could be moved to a supplemental figure.

      Minor comments:

      (1)Color/density scales should be added to Figs 1A and S1A, otherwise the yellow/white signal at the posterior could be interpreted as something other than high abundance.

      (2)In Fig 4A and 4C, I find it odd to have different halves of images photographed under different intensity settings and would prefer duplicate whole images.

      (3)The references to Fig 3G on page 13 should be corrected to Fig 4G.

      Significance

      The paper represents a substantial advance over existing knowledge and it extends our understanding about how RNAs can shuttle between different motor proteins to achieve a localized pattern. However, the Mohr et al 2021 PLoS Genetics paper covers some of the same ground. As that paper has now been published for several months, I believe a revised version of this paper should discuss that other work more prominently, making it apparent where the two studies concur and where this study extends the conclusions of the other one. If there are any contradictions between the two, those should be made explicit as well.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Response to Reviewer Comments

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

      Summary:

      In developing systems, morphogens gradients pattern tissues such that cells along the patterning length sense varying levels of the morphogen. This process has a low positional error even in the presence of biological noise in numerous tissues including the early embryo of the Drosophila melanogaster. The authors of this manuscript developed a mathematical model to test the effect of noise and mean cell diameter on gradient variability and the positional error they convey.

      They solved the 1D reaction-diffusion equation for N cells with diameters and kinetic parameters sampled from a physiologically relevant mean and coefficient of variation (CV). They fit the resulting morphogen gradients to a hyperbolic cosine profile and determined the decay length (DL) and amplitude (A) for a thousand independent runs and reported the CV in DL and A.

      The authors found that CV in DL and A increases with increase in mean cell diameter. They propose a mathematical relationship between CV in DL scales as an inverse-square-root of N. Whereas the CV in DL and A is a weak function of CV of cell surface area (CVa) if CVa __They further looked at the shift in readout boundaries and compared four different readout metrics: spatial averaging, centroid readout, random readout and readout along the length of the cilium. Their results show that spatial averaging and centroid have a high readout precision.

      They finally showed that the positional error (PE) increases along the patterning length of the tissue and increases with increasing mean cell diameter.

      The authors also supported their theoretical and simulated results by looking at mean cell areas reported for in patterning tissues in literature which also have a higher readout precision with smaller cell diameters.

      Major comments:

      Most of the key conclusions are convincing. However, there are four major points that should be addressed. First, the authors conclude the section titled, "The positional error scales with the square root of the average cell diameter," by saying that morphogen systems with small cells can have high precision in absolute length scales, but not on the scale of one cell diameter. They state this would result in salt and pepper patterns in the transition zones. The authors should either support this with biological examples or explain why this is not observed experimentally.

      We thank the referee for pointing out this imprecise comment, which we have removed. The exact nature of transition zones between patterning domains is a subject of ongoing research in our group, and goes beyond the scope of the present work. We will be sharing our results on this aspect in a separate forthcoming publication.

      Second, perhaps the main conclusion of the paper is that morphogen gradients pattern best when the average cell diameter is small. The authors support this by reviewing the apical cell area of epithelial systems that are known to be patterned by morphogens and those that are not (presumably taking apical cell area as a proxy for cell diameter). However, the key parameter is not absolute cell diameter, but the cell diameter relative to the morphogen length scale. The authors should report the ratio of these two quantities in their literature analysis.

      Since cell areas and cell diameters are monotonically increasing functions of one another for reasonably regular cell shapes, we indeed consider apical cell areas as proxies for the cell diameter, as the referee correctly noted. Cell areas are more frequently reported in the literature than cell diameters, which is why we compiled these in our analysis.We have now revised our analysis of the effect of the cell diameter on patterning precision to further length scales relevant in the patterning process. We show by example of the Drosophila wing disc how the parallel changes in cell diameter and morphogen source size compensate for the increase in gradient length and domain size, which would otherwise reduce patterning precision over time as the readout positions shift away from the source to maintain the same relative position in the growing wing disc.

      Lamentably, accurate measurements of morphogen gradients in epithelial tissues are still rare. In fact, among the listed tissues that are patterned by gradients, we are only aware of measurements of the SHH and BMP gradients in the mouse NT (lambda = 20 µm) and of the Dpp gradients in the Drosophila wing and eye discs [Wartlick, et al., Science, 2011 & Wartlick et al., Development, 2014]. We agree that it would be great if experimental groups would measure this in more tissues. In this revised and extended analysis, we show that the positional error increases with the cell diameter in absolute terms, not only relative to any reference length, be it the gradient length or cell diameter.

      Third, as part of their literature analysis, the authors state that in the Drosophila syncytium, there are morphogen gradients, but they imply that because these gradients operate prior to cellularization, one cannot use the large distances between nuclei as counter evidence to their main conclusion. Rather than simply dismissing the case of the Drosophila syncytium, the authors should explain why this case does not apply, using reasoning based on their model assumptions.

      Our paper is concerned with patterning of epithelia (which we now make clearer in the manuscript), and we would not want to stretch our paper to other tissue types, as the reaction-diffusion process in them differs. But we do not share the referee’s sentiment that the syncytium would present a counter-example. Since our model explicitly represents kinetic variability between spatial regions bounded by cell membranes, which are absent in the syncytium, our model is not directly applicable to it. We now provide this argument in the discussion, as requested by the referee.

      At 100 µm [Gregor et al., Cell, 2007], the Bicoid gradient is 5 times longer than the SHH/BMP gradients in the mouse neural tube and more than 10 times the reported length of the WNT gradient in the Drosophila wing disc [Kicheva et al., Science, 2007]. The nuclei become smaller as they divide because the anterior-posterior length of the Drosophila embryo remains about 500 µm [Gregor et al., Cell, 2007], but even at the earliest patterning stage their diameter will not be larger than 10 µm at midinterphase 12 [Gregor et al., Cell, 2007, Fig. 3A].

      Fourth, related to the above: the authors then state that there are no morphogen gradients known during cellularization. Unless I am misunderstanding their point, this is untrue. The Dpp gradient acts during the process of cellularization and specifies at least three distinct spatial domains of gene expression. Furthermore, not long after gastrulation, EGFR signaling patterns the ventral ectoderm into at least two distinct domains of gene expression. What are the cell areas in that case?

      Unfortunately, the referee does not provide literature references, and we were not able to find anything in the literature ourselves. We have now rephrased the statement to “we are not aware of morphogen gradient readout during cellularisation”.

      Minor comments:

      Figs 1cd:

      The way the system is set-up: (DL = 20 micron, Patterning Length (LP) = 250 micron, Nominal cell diameter (D) = 5 micron) the DL/L ~ 0.08 which makes the exponential profile far to a small value around 100 micron. This means in all these simulations, the LP was only around 100 micron, cells beyond that saw nearly zero concentration.

      Because of this, when diameters were varied from 0.2 - 40 micron, there could be as few as 2.5 cells in the "patterning region" which could be responsible for higher variability in DL and A.

      Patterning in the neural tube works across several 100 µm. At x=100µm, there is still exp(-5)=0.0067 of the signal left, which likely well translates into appreciable numbers of the morphogen molecule (see [Vetter & Iber, 2022] for a discussion of concentration ranges cells might sense). Unfortunately, very little is known about absolute morphogen numbers in the different patterning systems — experimental data is available only on relative scales, not in absolute nu mbers. While more quantitative experiments are still outstanding, modeling work needs to be based on reasonable assumptions. The seemingly quick decay of exponential profiles (when plotted on a linear scale) can be deceiving. In fact, exponential profiles describe the same fold-change over repeated equal distances, which makes them biologically very useful for different readout mechanisms operating on different levels of morphogen abundance. Our simulations are not limited to a patterning length of 100µm. Our work merely shows that variable exponential gradients stay precise over a long distance. We draw no conclusion on whether cells are able to interpret the low morphogen concentrations that arise far in the patterning domain - this aspect certainly deserves further research.

      The referee’s observation is correct in that for a cell diameter of up to 40 µm, there are only few cells in the patterning domain (namely down to about six, for a length of 250µm, as used in the simulations). It is also correct that this is the reason why gradients in such a tissue have greater variability in lambda and C0. This is precisely the main point we are making in this study: The narrower the cells in a tissue of given size, the less variable the morphogen gradients, and the more accurate the positional information they carry. Conversely, the wider the cells in x direction, the more variable the gradients.

      Would any of the results change if DL/L was higher, around 0.2?

      As we consider steady state gradients, nothing changes if we fix the (mean) gradient decay length and only shorten the patterning domain, except for a small boundary effect at the far end of the tissue due to zero-flux conditions applied there. At a fixed gradient length, the steady-state gradients just extend further if DL/L is increased (for example to 0.2), reaching lower concentrations, but the shape remains unchanged, and so does the morphogen concentration at a given absolute readout position.

      To demonstrate what happens at DL/L = 0.2, as requested by the referee, we repeated simulations with an increased gradient decay length of DL=50 micrometers; the length of the patterning domain remained unchanged at L=250 micrometers. As it is not possible to include image files in this response, we have made the plots available at https://git.bsse.ethz.ch/iber/Publications/2022_adelmann_vetter_cell_size/-/blob/main/revision_increased_dl.pdf for the time of the reviewing process. The plots show the resulting gradient variability, which is analogous to Fig 1c,d in the original manuscript. For both gradient parameters, we still recover the identical scaling laws.

      The source region is 25 microns in length and all cell diameters above 25 micron get defaulted back to 25 micron which explains the flatness lines in the region beyond mu_delta/mu_DL> 1

      Thanks for pointing this out. We now mention this in the manuscript. Note that it’s the ratio mu_delta/L_s that matters, not mu_delta/mu_lambda. It just so happens in this case, that both are nearly equal, because L_s=5*mu_lambda/4 in our simulations.

      Results:

      Pg 2 (bottom left): In the git repository code, the morphogen gradients are fit to a hyperbolic cosines function (described in reference 19) which is not described in the main text. Having this in the main text would help readers understand why fig 1c has variation in d only, D only and all k parameters whereas fig 1d has variation with all individual parameters p, d and D and all k.

      The reason why the impact of CV_p alone on CV_lambda is not plotted in Fig 1c is that it is minuscule. We now mention this in the figure legend. This follows from the fact that the gradient length lambda is determined in the patterning domain, whereas the production rate p sets the morphogen concentration in the source domain, and thus, the gradient amplitude, but not its characteristic length. This is unrelated to the functional form used to fit the shape of the gradients, be it exponential or a hyperbolic cosine. We mention that we fit hyperbolic cosines to the numerical gradients in section Gradient parameter extraction in the Methods section, and we refer the interested reader to the original reference [Vetter & Iber, 2022], which contains all mathematical details, should they be needed.

      Figure 3b:

      In figures where markers are overlapping perhaps the authors can use a "dot" to identify one set of simulations and a "o" to identify the ones under it. The way the plots are set up currently makes it hard for the reader to understand where certain points on the plot are.

      We use a color code to represent the readout strategy and different symbols to represent the cell diameter in Fig 3b. We agree that for the smallest of the cell diameters, the diamond-shaped data points lie so close that they are not easy to tell apart at first sight. For this reason, we chose different symbol sizes. We would like to keep the symbols as they are to maintain visual consistency with the other figures, which we think is an important feature of our presentation that facilitates the interpretation. Note that all our figures are vector graphics, which allow the reader to zoom in arbitrarily deep, and to easily distinguish the data points. Note also that in this particular case, telling the data points apart is not necessary; recognizing that they are nearly identical is sufficient for the interpretation of our results.

      Methods:

      The Methods can be more descriptive to include certain aspects of the simulations such as adjusted lambda which is only described in the code and not the main text or supplementary.

      We apologize for this omitted detail. As shown in Fig. 8g in [Vetter & Iber, 2022], the mean fitted value of lambda drifts away from the prescribed value, depending on which of the kinetic parameters are varied, and by how much. To report the true observed mean gradient length in our results, we corrected for this drift in our implementation, as the referee correctly noticed. We now describe this in the methods section, and we have extended the methods also on other aspects.

      Git code:

      The git code function handles do not represent figure numbers and should be updated to make it easier for readers to find the right code

      Thank you for pointing this out — it was an oversight from an earlier preprint version. The function names now correspond to the figure numbers.

      Reviewer #1 (Significance (Required)):

      This manuscript contributes certain key aspects to the patterning domain. The three most important contributions of this work to the current literature are: (1) the scaling relationships developed here are important, (2) the idea that PE increases at the tail-end of the morphogen profile is nicely shown and (3) Comparison of various readout strategies.

      Thank you for the positive assessment.

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

      Summary:

      How morphogen gradients yield to precise patterning outputs is an important problem in developmental biology. In this manuscript, Adelmann et al. study the impact of cell size in the precision of morphogen gradients and use a theoretical framework to show that positional error is proportional to the square root of cell diameter, suggesting that the smaller the cells in a patterning field, the more precise patterns can be established against morphogen gradient variability. This result remains true even when cells average the morphogen signal across their surface or spatial correlations between cells are introduced. Thus, the authors suggest that epithelial tissues patterned by morphogen gradients buffer morphogen variability by reducing apical cell areas and support their hypothesis by examining several experimental examples of gradient-based vs. non-gradient-based patterning systems.

      Major comments:

      While the idea that smaller cells yield to more precise morphogen gradient outputs is attractive, it is unclear whether patterning systems use this strategy to make patterns more precise, as there are several mechanisms that could achieve precision. Do actual developmental systems use it as a mechanism to increase precision? Or precision is achieved through other mechanisms (for example, cell sorting as in the zebrafish neural tube; Xiong et al. Cell, 2013). Indeed, classical patterning work on Drosophila embryo suggest that segmentation patterns are of an absolute size rather by an absolute number of cells (Sullivan, Nature, 1987). According to the authors, the patterning stripes should be more precise when embryos have higher cell densities than in the wild-type, but stripes are remarkably precise in wild-type embryos. This is likely due to other precision-ensuring mechanisms (such as downstream transcriptional repressors, in this case).

      We want to emphasize that our predictions concern the precision of the gradients, not the precision of their readout, which can be strongly affected by readout noise, as we will show in a forthcoming paper. Cell sorting can sharpen boundaries in the transition zone, but this would not address errors in target domain sizes and is thus different from gradient precision as we discuss it here. Also, cell sorting as observed in the zebrafish neural tube requires higher cell motility than what is observed in most epithelial tissues. The work by Sullivan, Nature, 1987, is concerned with patterning of the early Drosophila embryo, and the stripes are defined already before cellularisation. We are unfortunately not aware of any work that quantified gradient precision at different cell densities in epithelia. This would, of course, be highly interesting data and would indeed put our predictions to a test. We are, to the best of our knowledge, the first to propose this principle with the present work. We have now made these points and distinctions clearer in the revised manuscript. Thank you for bringing this up.

      Their modeling approach is based on exponential gradients formed by diffusion and linear degradation, but in reality, actual morphogen gradients are affected by receptor and proteoglycan binding and are likely not simply exponential and/or interpreted at the steady state. Do the main results of the manuscript hold even for non-exponential gradients or before they reach a steady state?

      We can confirm that our results also hold for non-exponential gradients, as they emerge for example when morphogen degradation is self-enhanced (i.e., non-linear). This result will be published in a follow-up study [BioRxiv: 10.1101/2022.11.04.514993], which we now cite in the concluding remarks in the revised manuscript.

      The analysis of pre-steady-state gradients lies outside of the scope of the present work, and so the question as to whether our results are applicable to them as well, remains to be answered in future research. We have added a comment on this to the discussion.

      In their Discussion section, the authors note that several patterning systems, such as the Drosophila wing and eye discs, show smaller cells near the morphogen source relative to other regions in the tissue. This observation suggests a prediction of the authors' hypothesis that can be tested experimentally. In the Drosophila wing and eye discs genetic mosaics of ectopic morphogen sources (such as Dpp) can (and have) been made. Therefore, one could predict that the patterning outputs in a region of larger cross-sectional areas will be more imprecise than in the endogenous source. Since this is a theoretical paper, it is understandable that authors are not going to make this experiment themselves, but I wonder if they can use published data to test this prediction or at least mention it in the manuscript to offer the experimental biology reader an idea of how their hypothesis can be tested experimentally.

      We appreciate that the referee would like to help us inspire the experimental community. Unfortunately, the problem with the proposal is that Dpp has been shown to result in a lengthening of the cells (and thus a smaller cell width) [Widmann & Dahman, J Cell Sci, 2009]. The Dpp gradient thus ensures a small cell width close to its source, which makes it virtually impossible to test this proposal experimentally in the suggested way. Nevertheless, we have added brief comments on potential experimental testing of our predictions to the discussion.

      Other comments:

      The Methods section should be expanded and should include more details about how authors consider cell size in their simulations. As presented, I believe that experimental biologists will not be able to grasp how the analysis was done.

      We have expanded on the technical details of our model in the methods section, in particular in relation to the cell size, as requested. To avoid being overly redundant with existing published descriptions of the modeling details [Vetter & Iber, 2022], we focus here on a description of what has not been covered already, and refer the interested reader to our previous publication. It is inevitable for any kind of work, be it theoretical or experimental, to be less accessible to experts in other disciplines, but we believe that the presentation of our results is independent enough of modeling aspects to be accessible to experimental biologists, too.

      Authors use adjectives such as 'little' as 'small' without a comparative reference. For example in the abstract, the authors say that apical areas "are indeed small in developmental tissues." What does "small" mean? This should be avoided throughout the text.

      We thank the referee for raising this point. Where appropriate, we changed the phrasing accordingly to clarify what the comparative reference is. We leave all sentences unchanged where the statement holds in absolute terms. Note that in the substantially revised analysis on the impact of the different length scales involved in the patterning process, we now explicitly show with simulation data and theory that the absolute positional error increases with increasing absolute cell diameter.

      Reviewer #2 (Significance (Required)):

      Overall, I believe that the manuscript is well written and deserves consideration for publication. However, authors should consider the points outlined above in order to make their manuscript more accessible and relevant to the developmental biology community.

      Thank you for the positive assessment.

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

      In their mansucript "Impact of cell size on morphogen gradient precision" the authors Adelmann, Vetter and Iber numerically analyse a one-dimensional PDE-based model of morphogen gradient formation in tissues in which the cell sizes and cell-specific parameters locally affecting the gradient properties are varied according to predefined distributions. They find that the average cell size has the largest impact on the variance of the gradient shape and the read-out precision downstream, while other factors such as details of the readout mechanism have markedly less influence on these properties. In addition they demonstrate that averaging gradient concentrations over typical cell areas induces a shift of the readout position, which however appears to be insignificant (~1% of the cell diameter) for typical parameters.

      Overall this manuscript is in very good shape already and tackles an interesting topic. I still would like the authors to address the comments below before I would recommend any publication. My main criticism pertains to some of the authors' derivations which, as I find, partly do deserve more detail, and to their conclusions about gradient readout precision.

      Thank you for the positive assessment.

      MAJOR COMMENTS

      p. 1, left column: The positional error of the readout position does not only depend on the variation of the gradient parameters, as suggested by the first part of the introduction. A very important factor is also the fluctuations due to random arrival of molecules to the promoters that perform the readout due to the limited (and typically low) molecule number. In fact, for positions very distant to the source of the gradient, this noise source is expected to be dominant over gradient shape fluctuations. Importantly, these fluctuations also arise for non-fluctuating, "perfect" gradient inputs if copy numbers of the morphogen molecules are limited (which they always are). This important contribution to the noise is neglected in the work of the authors. This is OK if the purpose is focusing on the origin and influence of the gradient shape fluctuations, but that focus should be clearly highlighted in the introduction, saying explicitly that noise due to diffusive arrival of transcription factors is not taken into account in the given work (see, e.g., Tkacik, Gregor, Bialek, PLoS ONE 3, 2008)

      In the present work, only precision of the gradients, but not the readout itself is studied. We have now mentioned this more explicitly in the introduction. We also acknowledge the fact that the readout itself introduces additional noise into the system. We are currently finishing up work that addresses exactly this subject, which is outside of the scope of the present paper.

      What may have led to misinterpretation of the scope of our work is that we called x_theta the readout position. x_theta defines the location where cells sense (i.e., read out) a certain concentration threshold, and is not meant to be interpreted as the location of a certain readout (a downstream transcription factor) of the morphogen. We have made this distinction clearer in the revised manuscript.

      p.1, right column: Why exactly are the parameters p, d, D assumed to follow a log-normal distribution? Such a distribution has been verified for cell size, but the rationale behind choosing it also for the named parameters should be explained, in particular for D. Why would D depend on local properties of the cell? Which diffusion / transport mechanism precisely is assumed here?

      The motivations for the used log-normal distributions for the kinetic parameters are the following:

      The morphogen production rates, degradation rates and diffusivities must be strictly positive. This rules out a normal distribution. The probability density of near-zero kinetic parameters must vanish quickly, as otherwise no successful patterning can occur. For example, a tiny diffusion coefficient would not enable morphogen transport over biologically useful distances within useful timeframes. This rules out a normal distribution truncated at zero, because very low diffusivities would occur rather frequently for such a distribution. Given the absence of reports on distributions for p, d, D from the literature, we chose a plausible probability distribution that fulfills the above two criteria and possesses just two parameters, such that they are fully defined by a mean value and coefficient of variation. This is given by a lognormal distribution. Our results are largely independent of the exact choice of probability distribution assumed for the kinetic parameters, under the constraints mentioned above. To demonstrate this, we have repeated a set of simulations with a gamma distribution with equal mean and variance as used for the lognormal distribution. Below are some simulation results for a gamma distribution with shape parameters a = 1/CV^2 and inverse scale parameter b = mu*CV^2 with CV = 0.3 as used in the results shown in the paper. As can be appreciated from these plots, the results do not change substantially, and our conclusions still hold. As we believe this information is potentially relevant for the readership of our paper, we have added this result and discussion to the supplement and to the conclusion in the main text.

      We assume extracellular, Fickean morphogen diffusion with effective diffusivity D along the epithelial cells, as specified by Eq. 2. We now state this more explicitly just below Eq. 2 in the revised manuscript. Cell-to-cell variability in the effective diffusivity may arise from effects that alter the effective diffusion path and dynamics along the surface of cells, which we do not model explicitly, but lump into the effective values of D. Such effects may include different diffusion paths (different tortuosities) or transient binding, among others.

      Moreover, is there any relationship between A_i and p_i, d_i and D_i, or are these parameters varied completely independently? If yes, is there a justification for that?

      The parameters are all varied independently, as written in the paragraph below Eq. 2 on the first page (“drawn for each cell independently”). To our knowledge there is no reported evidence for correlations between cell areas, morphogen production rates, degradation rates, or transport rates across epithelia, that we could base our model on. The choice of independent cell parameters therefore represents a plausible model of least assumptions made. Note that we explore the effect of potential spatial correlations in the kinetic parameters between neighboring cells in the section “The effect of spatial correlation”, finding that such correlations, if at all present, are unlikely to significantly alter our results.

      p. 2, right column, section on "Spatial averaging": First of all, how is "averaging" exactly defined here? Do the authors assume that the cells can perfectly integrate over their surface in the dimensions perpendicular to their height? If yes, then this should be briefly mentioned here. Secondly, the shift \Delta x calculated by the authors ultimately seems to trace back to the fact that the cells average over an exponential gradient, whose derivative also is exponential, such that levels further to the anterior from the cell center are higher (on average) than levels to the posterior of it. I suppose, therefore, that a similar calculation for linear gradients would not lead to any shift. If these things are true they deserve being mentioned in this part of the manuscript because they provide an intuitive explanation for the shift. Thirdly, in Fig. 2A the cell sizes seem exaggerated with respect to the gradient length. This seems fine for illustrative purposes, but if it is the case it should be mentioned. Also, I believe that this figure panel would benefit from showing another readout case where the average concentration e.g. in cell 1 maps to its corresponding readout position, in order to show that this process repeats in every cell. Moreover, it could be indicated that in the shown case C_\theta matches the average concentration in cell 2 at the indicated position.

      Spatial averaging is defined as perfect integration along the spatial coordinate over a length of 2r (which can generally be equal to, or smaller than, or larger than one cell diameter) as detailed in the supplementary material. In simulations, we use the trapezoid method for numerical integration to get the average concentration a cell experiences along its surface area perpendicular to their height.

      The reviewer is correct, that the shift is a consequence of averaging over an exponential gradient. The average of an exponential gradient is higher compared to the concentration at the centroid of the cell, thus the small shift. This is mentioned e.g. in the caption of Fig. S1, but also in the main text (“spatial averaging of an exponential gradient results in a higher average concentration than centroid readout”). We have now added this information also to the caption of Fig. 2. As pointed out correctly by the referee, linear gradients would not result in such a shift. A brief comment on this has been added to the revised manuscript.

      We now mention that the cell size is exaggerated in comparison to the gradient decay length for illustration purposes in the schematic of Fig. 2a, as requested.

      Unfortunately, we had a hard time following the reviewer’s final point. We show a specific readout threshold concentration, C_theta, in Fig. 2a. A cell determines its fate based on whether its sensed (possibly averaged) concentration is greater or smaller than C_theta. In the illustration, cells 1 and 2 sense a concentration greater than C_theta, and all further cells sense a concentration smaller than C_theta. Cell fate boundaries necessarily develop at cell boundaries (here; between cells 2 and 3, red). Additionally, the readout position for a continuous domain, where morphogen sensing can occur at an arbitrary point along the patterning axis, is shown (blue). This position can be different from the one restricted to cell borders. Thus, different readout positions in the patterning domain result from the two scenarios, which is what the schematic illustrates. Given that our illustration seems to go well with the other referees, we are unsure in what way it could be improved.

      As for the significance of the magnitude of the shift for typical parameters as calculated by the authors: I believe that it could be said more explicitly and clearly that under biological conditions the calculated shift overall seems insignificant, as it amounts to a small fraction of the cell diameter.

      We have made this more explicit in the text.

      Finally, and most importantly: The term "spatial averaging" can have a different meaning in developmental biology than the one employed by the authors. While the authors mean by it that individual cells average the gradient concentration over their area, in other works "spatial averaging" typically means that individual cells sense "their" gradient value (by whatever mechanism) and then exchange molecules activated by it, which encode the read-out gradient value downstream, between neighboring cells, in order to average out the gradient values "measured" under noisy conditions. The noise reduction effect of such spatial averaging can be very significant, as evidenced by this (incomplete) list of works which the authors can refer to:

      - Erdmann, Howard, ten Wolde, PRL 103, 2009

      - Sokolowski & Tkacik, PRE 91, 2015

      - Ellison et al., PNAS 113, 2016

      - Mugler, Levchenko, Nemenman, PNAS 113, 2016

      The main point, however, is that this is a different mechanism as the one described by the authors, and this should be clearly mentioned in order to distinguished them. I would therefore also advise the authors to make the section title more precise here, by changing "Spatial averaging barely affects ..." to "Spatial averaging across the cell area barely affects ..." for clarity.

      Most theory development has previously indeed been done with the syncitium of the early Drosophila embryo in mind. However, most patterning in development happens in epithelial (or mesenchymal) tissues, where spatial averaging via translated proteins is not as straightforward and natural as in a syncitium. In fact, a bucket transport of a produced protein from cell to cell would be difficult to arrange (as upon internalization, degradation would have to be prevented), be subject to much molecular noise, and be rather slow. Our paper is concerned with patterning in epithelia, which we have now stated more clearly in the manuscript.

      Regarding the section title: Our analysis does not only cover spatial morphogen averaging over the cell area, but it also includes averaging radii below (in the theory) and far above (in the theory and in the new Fig. 4c, previously 3c) half a cell diameter. With cilia of sufficient length r, epithelial cells could potentially average over spatial regions extending further than their own cell area, without need for inter-cellular molecular exchange between neighboring cells. This is the kind of spatial averaging we explored here. Restricting the section title to the cell area only would therefore be misleading. However, we agree with the referee that the distinction between different meanings of “spatial averaging” is important, and we now emphasize our interpretation and the scope of our work more in the revised text.

      p. 3, Figure 3: It would be good to highlight the fact that the colours in panel A correspond to the bullet colors in the other panels also in the main text.

      We now added this also in the main text.

      As to the comparison of different readout strategies: How exactly were the different readout mechanisms compared on the mathematical side? More precisely: How was the readout by the whole area matched (in terms of fluxes) to the readout at a single point, be it in the center of the cell or a randomly chosen point? How was it ensured that the comparison is done at equal footing?

      Our model considers that a cell can sense a single concentration even if it is exposed to a gradient of concentrations. Assuming the French flag model is correct, a cell must make a binary decision based on a sensed concentration in order to determine its fate. The different readout strategies are hypothetical and simplified mechanisms for how a cell could, in principle, detect a local morphogen signal. It is unclear to us what the referee is referring to when mentioning “matching in terms of fluxes”, as there are no fluxes involved in the modeled readout strategies. We make no assumption on the underlying biochemical mechanism that would allow cells to implement one of the strategies. The main goal of this analysis was to determine whether various different sensing strategies had a significant effect on the precision of morphogen gradients experienced by cells. To assure that we can compare the different mechanisms at equal footing, we simulated gradients and then calculated from each gradient the readout concentration in each cell and for each of the methods.

      p. 3, right column: "... similar gradient variabilities, and thus readout precision": Linking to comment 1 above, this is strictly speaking only the case when the only source of fluctuations in the readout is the gradient fluctuations. I would therefore leave this statement out.

      To avoid confusion, we have removed parts of the sentence. Thank you for pointing this out.

      p. 3, section on positional error (right column): In this part I had most troubles following the thoughts of the authors.

      First of all, the measure that the authors use for the positional error is sigma_x / mu_lambda, i.e. the standard deviation of the readout position relative to the gradient length. The question is whether this is the correct measure. It should be specified what the motivation for normalizing by mu_lambda is. In the end, one could argue, what the cells really do care about would be that the developmental process can assign cell fates with single cell precision, for which the other measure shown in Eq. (6) is the representative one. Now in contrast to the former measure, the latter actually increases with decreasing cell diameter.

      We thank the referee for raising this point, and acknowledge that we have not presented this aspect well enough. We have rewritten the entire section and the discussion about biological implications. Instead of normalizing with a constant mean gradient length in the formulas and figures, which has left room for misinterpretation, we now instead varied all relevant length scales in the patterning system, to determine the impact of each of them independently on the positional error. We now show that the positional error increases (to leading order) proportionally to the mean gradient length, the square root of the cell diameter, the square root of the location in the patterned tissue, and inversely proportional to the length of the source domain. We support these new aspects with new simulation data (Fig. 2E-2H, Fig. 3D-G, Fig. S5, Fig. S6). As the positional error is now reported in absolute terms, rather than relative to a particular length scale, the question of the relevant scale is addressed. We now show that the absolute positional error increases with increasing absolute cell diameter.

      We believe that this extension provides additional important insight into what affects the patterning precision. We thank the referee very much for motivating us to expand our analysis.

      Secondly, even when the former measure (sigma_x / mu_lambda) is employed, Fig. 3(D) shows that while it decreases with decreasing cell diameters, in the regime of small diameters the std. dev. of the readout position becomes larger than the average cell diameter, which actually would mean that cell fates cannot be assigned with single-cell precision. While the authors later report both quantities for specific gradients, it should be clarified beforehand which of the measures is the relevant one.

      This has now been addressed by considering absolute length scales as discussed at length in our answer to the previous point.

      Moreover, in the following derivations, mu_x is not properly introduced. What exactly is the definition of that quantity? Is it the mean readout position? If yes, it is not clear why exactly it would be interesting and relevant to the cell. This should be properly explained in a way that does not require the reader to look up further details in another publication.

      The referee is correct in that mu_x is the mean readout position. We apologize for not being clear enough on this, and have now defined this in the introduction together with the definition of sigma_x.

      At the end of this section the authors come back to the sigma_x / mu_delta measure again and indeed point out that it increases with decreasing mu_delta, which causes a bit of confusion because the initial part of the section only talks about the increase of the pos. error with mu_delta. Overall I find that this section should be rewritten more clearly. Right now it leaves the reader with the "take home message" that small cells are good because they lead to smaller pos. error, but when the--in my opinion--relevant measure (sigma_x/mu_delta) is employed the opposite is the case. This is confusing and unclear about the authors' intentions in that part.

      See the answer above. The “take-home message” is now reformulated in absolute terms regarding the effect of cell diameter, rather than relative to a certain choice of reference scale. Our new analysis revealed a new relative ratio that determines the positional error, mu_lambda/L_s. We now discuss this relative measure also regarding its biological significance. Once again, we thank the referee for pointing us at this source of confusion, the elimination of which allowed us to improve our analysis.

      __Finally, the authors could also supplement the numbers that they name for the FGF8 and SHH gradients by the known numbers for the Bcd gradient in Drosophila, which has been studied excessively and constitutes a paradigm of developmental biology. Here mu_delta ~= 6.5 um, while mu_lambda ~= 100 um, such that mu_delta/mu_lambda While we appreciate that most theoretical work has been done for syncytia, this paper is concerned with patterning of epithelia, which have different patterning constraints, as also explained in a reply further above. We now make the scope of our work clearer in the revised manuscript. But as the referee points out, the diameter of the nucleus relative to the gradient length is such that gradients can be expected to be sufficiently precise.

      p. 4, section on the effect of spatial correlation: Here the authors chose to order the kinetic parameters in ascending or descending order. Is there any biological motivation for that particular choice? Other types of correlations seem possible, e.g. imposing the rule that successive parameter values are sampled starting from the previous value, p_i+1 = o_i +- delta_i+1 where delta_i+1 are random numbers with a defined variance.

      In the simulations we go from zero correlation (every cell has independent kinetic parameters) to maximal correlation (every cell has the same parameters, resulting effectively in a patterning domain that consists of a single effective “cell”), see Fig. S3. Biologically plausible correlations in between these extremes should retain the same kinetic variability levels (same CVs) which we took from the measured range reported in the literature. We accomplish this by ordering the parameters after independently sampling the parameters for each cell from probability distributions with the desired CV. The motivation for this approach is that this produces a type of maximal correlation that still reflects the measured biological cell-to-cell variability, to demonstrate in Fig. S3, that even such a maximal degree of spatial correlation does not qualitatively alter our results. The kind of correlation that the referee suggests introduces a spatial correlation length that lies in between the extremes that we simulated. Since even for maximal correlation using the ordering approach, we find our conclusions to still apply, we have no reason to expect that intermediate levels of correlation would behave any differently.

      The idea brought forward by the referee effectively introduces a correlation length scale. We discuss this case in the paper, noting that the positional error will scale as x~N , where N is the number of cells sharing the same kinetic parameters. A correlation length scale will be proportional to N and will therefore simply uniformly scale the positional error accordingly, but will likely not reveal any new insight beyond that.

      Moreover, using the idea of the referee as an additional way to introduce correlation is difficult to realise in practice, as we need to recover the mean and variance of the kinetic parameters, while ensuring strict positivity for each of them. A simple random walk, as proposed, would not lend itself easily to achieve this without introducing a bias in the distribution, because negative values need to be prevented. As explained in a reply further above, an important feature of the kinetic parameters is that they are not too small to prevent the formation of a meaningful gradient, which is not straightforward to ensure with the proposed method.

      We acknowledge that there are different types of correlations conceivable, but we expect these correlations to lie between the two extremes that we present in the paper, which show no qualitative difference in the results.

      p.5, Discussion: "..., but with nuclei much wider than the average cell diameter". To be honest, I could not completely imagine what is meant with this sentence. Intuitively, it seems that the nuclei cannot be larger than the cells, but I suppose that some kind of special anisotropy is considered here? In any case, this should be made precise.

      The main tissues that are patterned by gradients are epithelia. Our paper focuses on such tissues. It is a well-known feature of pseudostratified epithelia that nuclei are on average wider than the cell width averaged over the apical-basis axis. Nature solves this problem by stacking nuclei above each other along the apical-basal axis, resulting in a single-layered tissue that appears to be a multi-layered stratified tissue when only looking at nuclei. For a schematic illustration of this, see Fig. 1 in [DOI: 10.1016/j.gde.2022.101916]. An image search for “pseudostratified epithelia” on Google yields a plethora of microscopy images. Right at the end of the quote recited by the referee, we also cite our own study [Gomez et al, 2021], which quantifies this in Fig. 5.

      Moreover, I find that the conclusion that morphogen gradients "provide precise positional information even far away from the morphogen source" goes to far based on the authors' work, precisely for the fact input fluctuations due to limited morphogen copy number, which can become detrimentally low far away from the source, are not considered, neither the timescales needed to both establish and sample such low concentrations far away from the source. While thus, according to the work of the authors, the fluctuations in the morphogen signal may be favorably small, these other factors are supposed to exert a strong limit on positional information. This conclusion therefore seems unjustified and should be toned down, or even better taken out and replaced by a more accurate one, which only focuses on the gradient shape fluctuations, not on the conveyed positional information.

      There is no evidence so far that morphogen gradient concentrations become too low to be sensed by epithelial cells, to the best of our knowledge. What we show is that the gradient variability between embryos remains low enough that precise patterning remains possible. Whether the morphogen concentration remains high enough to be read out reliably by cells is a subject that requires future research. Genetic evidence from the mouse neural tube demonstrates that the SHH gradient is still sensed at a distance beyond 15 lambda (SHH signalling represses PAX7 expression at the dorsal end of the neural tube) [Dessaud et al., Nature, 2007], where an exponential concentration has dropped more than 3-million-fold.

      As the referee correctly recites, we state that “morphogen gradients remain highly accurate over very long distances, providing precise positional information even far away from the morphogen source”. This statement is restricted to the positional information that the gradients convey, and does not touch potentially precision-enhancing or -deteriorating readout effects, nor does it concern the absolute number of morphogen molecules.

      Positional information goes through several steps. The gradients themselves convey a first level of positional information, by being variable in patterning direction, as quantified by the positional error. This is what we draw our conclusion about. This positional information from the gradients can then be translated into positional information further downstream, by specific readout mechanisms, inter-cellular processes, temporal averaging, etc. About these further levels of positional information, we make no statement.

      We therefore disagree that our conclusion is unjustified. In fact, we have phrased it exactly having the limited scope of our study in mind, making sure that we restrict the conclusion to the gradients themselves.

      MINOR COMMENTS

      - p. 1: "and find that positional accuracy is the higher, the narrower the cells".

      (This sentence, however, should be anyhow revised in view of major comment 5 above.)

      We have added “the”.

      - p. 4: "... with an even slightly smaller prefactor."

      We have removed “even”.

      Reviewer #3 (Significance (Required)):

      I believe that this work is significant to the community working on the theoretical foundations of morphogen gradient precision in developmental systems. The main interesting findings are that small cell diameters lead to smaller positional error (although the relevant measure should be clarified according to my comment no. 5), and that the gradient shape fluctuations are surprisingly robust with respect to the readout mechanism.

      Its limitations consist of the fact that the impact of small copy numbers on the readout and associated timescales are neglected, such that the findings of the authors on gradient robustness cannot be simply transferred by simple conversion formulas to readout robustness / positional information. Comment 5 goes hand in hand with this, as a different conclusion may emerge depending on how the relevant positional error measure is defined. This should be fixed by the authors as indicated in the main part of the report.

      Thank you for your assessment.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In their mansucript "Impact of cell size on morphogen gradient precision" the authors Adelmann, Vetter and Iber numerically analyse a one-dimensional PDE-based model of morphogen gradient formation in tissues in which the cell sizes and cell-specific parameters locally affecting the gradient properties are varied according to predefined distributions. They find that the average cell size has the largest impact on the variance of the gradient shape and the read-out precision downstream, while other factors such as details of the readout mechanism have markedly less influence on these properties. In addition they demonstrate that averaging gradient concentrations over typical cell areas induces a shift of the readout position, which however appears to be insignificant (~1% of the cell diameter) for typical parameters.

      Overall this manuscript is in very good shape already and tackles an interesting topic. I still would like the authors to address the comments below before I would recommend any publication. My main criticism pertains to some of the authors' derivations which, as I find, partly do deserve more detail, and to their conclusions about gradient readout precision.

      MAJOR COMMENTS

      1 - p. 1, left column: The positional error of the readout position does not only depend on the variation of the gradient parameters, as suggested by the first part of the introduction. A very important factor is also the fluctuations due to random arrival of molecules to the promoters that perform the readout due to the limited (and typically low) molecule number. In fact, for positions very distant to the source of the gradient, this noise source is expected to be dominant over gradient shape fluctuations. Importantly, these fluctuations also arise for non-fluctuating, "perfect" gradient inputs if copy numbers of the morphogen molecules are limited (which they always are). This important contribution to the noise is neglected in the work of the authors. This is OK if the purpose is focusing on the origin and influence of the gradient shape fluctuations, but that focus should be clearly highlighted in the introduction, saying explicitly that noise due to diffusive arrival of transcription factors is not taken into account in the given work (see, e.g., Tkacik, Gregor, Bialek, PLoS ONE 3, 2008)

      2 - p.1, right column: Why exactly are the parameters p, d, D assumed to follow a log-normal distribution? Such distribution has been verified for cell size, but the rationale behind choosing it also for the named parameters should be explained, in particular for D. Why would D depend on local properties of the cell? Which diffusion / transport mechanism precisely is assumed here?

      Moreover, is there any relationship between A_i and p_i, d_i and D_i, or are these parameters varied completely independently? If yes, is there a justification for that?

      3 - p. 2, right column, section on "Spatial averaging": First of all, how is "averaging" exactly defined here? Do the authors assume that the cells can perfectly integrate over their surface in the dimensions perpendicular to their height? If yes, then this should be briefly mentioned here. Secondly, the shift \Delta x calculated by the authors ultimately seems to trace back to the fact that the cells average over an exponential gradient, whose derivative also is exponential, such that levels further to the anterior from the cell center are higher (on average) than levels to the posterior of it. I suppose, therefore, that a similar calculation for linear gradients would not lead to any shift. If these things are true they deserve being mentioned in this part of the manuscript because they provide an intuitive explanation for the shift. Thirdly, in Fig. 2A the cell sizes seem exaggerated with respect to the gradient length. This seems fine for illustrative purposes, but if it is the case it should be mentioned. Also, I believe that this figure panel would benefit from showing another readout case where the average concentration e.g. in cell 1 maps to its corresponding readout position, in order to show that this process repeats in every cell. Moreover, it could be indicated that in the shown case C_\theta matches the average concentration in cell 2 at the indicated position.

      As for the significance of the magnitude of the shift for typical parameters as calculated by the authors: I believe that it could be said more explicitly and clearly that under biological conditions the calculated shift overall seems insignificant, as it amounts to a small fraction of the cell diameter.

      Finally, and most importantly: The term "spatial averaging" can have a different meaning in developmental biology than the one employed by the authors. While the authors mean by it that individual cells average the gradient concentration over their area, in other works "spatial averaging" typically means that individual cells sense "their" gradient value (by whatever mechanism) and then exchange molecules activated by it, which encode the read-out gradient value downstream, between neighboring cells, in order to average out the gradient values "measured" under noisy conditions. The noise reduction effect of such spatial averaging can be very significant, as evidenced by this (incomplete) list of works which the authors can refer to:

      • Erdmann, Howard, ten Wolde, PRL 103, 2009
      • Sokolowski & Tkacik, PRE 91, 2015
      • Ellison et al., PNAS 113, 2016
      • Mugler, Levchenko, Nemenman, PNAS 113, 2016

      The main point, however, is that this is a different mechanism as the one described by the authors, and this should be clearly mentioned in order to distinguished them. I would therefore also advise the authors to make the section title more precise here, by changing "Spatial averaging barely affects ..." to "Spatial averaging across the cell area barely affects ..." for clarity.

      4 - p. 3, Figure 3: It would be good to highlight the fact that the colours in panel A correspond to the bullet colors in the other panels also in the main text.

      As to the comparison of different readout strategies: How exactly were the different readout mechanisms compared on the mathematical side? More precisely: How was the readout by the whole area matched (in terms of fluxes) to the readout at a single point, be it in the center of the cell or a ranomly chosen point? How was it ensured that the comparison is done at equal footing?

      p. 3, right column: "... similar gradient variabilities, and thus readout precision": Linking to comment 1 above, this is strictly speaking only the case when the only source of fluctuations in the readout is the gradient fluctuations. I would therefore leave this statement out.

      5 - p. 3, section on positional error (right column): In this part I had most troubles following the thoughts of the authors.

      First of all, the measure that the authors use for the positional error is sigma_x / mu_lambda, i.e. the standard deviation of the readout position relative to the gradient length. The question is whether this is the correct measure. It should be specified what the motivation for normalizing by mu_lambda is. In the end, one could argue, what the cells really do care about would be that the developmental process can assign cell fates with single cell precision, for which the other measure shown in Eq. (6) is the representative one. Now in contrast to the former measure, the latter actually increases with decreasing cell diameter.

      Secondly, even when the former measure (sigma_x / mu_lambda) is employed, Fig. 3(D) shows that while it decreases with decreasing cell diameters, in the regime of small diameters the std. dev. of the readout position becomes larger than the average cell diameter, which actually would mean that cell fates cannot be assigned with single-cell precision. While the authors later report both quantities for specific gradients, it should be clarified beforehand which of the measures is the relevant one.

      Moreover, in the following derivations, mu_x is not properly introduced. What exactly is the definition of that quantity? Is it the mean readout position? If yes, it is not clear why exactly it would be interesting and relevant to the cell. This should be properly explained in a way that does not require the reader to look up further details in another publication.

      At the end of this section the authors come back to the sigma_x / mu_delta measure again and indeed point out that it increases with decreasing mu_delta, which causes a bit of confusion because the initial part of the section only talks about the increase of the pos. error with mu_delta. Overall I find that this section should be rewritten more clearly. Right now it leaves the reader with the "take home message" that small cells are good because they lead to smaller pos. error, but when the--in my opinion--relevant measure (sigma_x/mu_delta) is employed the opposite is the case. This is confusing and unclear about the authors' intentions in that part.

      Finally, the authors could also supplement the numbers that they name for the FGF8 and SHH gradients by the known numbers for the Bcd gradient in Drosophila, which has been studied excessively and constitutes a paradigm of developmental biology. Here mu_delta ~= 6.5 um, while mu_lambda ~= 100 um, such that mu_delta/mu_lambda < 1/15, which defines yet another regime than the other two gradients. It would be interesting to compare their respective numbers altogether, and also discuss the ones for Drosophila in view of the fact that in experiments sigma_x ~= mu_delta for this species.

      6 - p. 4, section on the effect of spatial correlation: Here the authors chose to order the kinetic parameters in ascending or descending order. Is there any biological motivation for that particular choice? Other types of correlations seem possible, e.g. imposing the rule that successive parameter values are sampled starting from the previous value, p_i+1 = o_i +- delta_i+1 where delta_i+1 are random numbers with a defined variance.

      7 - p.5, Discussion: "..., but with nuclei much wider than the average cell diameter". To be honest, I could not completely imagine what is meant with this sentence. Intuitively, it seems that the nuclei cannot be larger than the cells, but I suppose that some kind of special anisotropy is considered here? In any case, this should be made precise.

      Moreover, I find that the conclusion that morphogen gradients "provide precise positional information even far away from the morphogen source" goes to far based on the authors' work, precisely for the fact input fluctuations due to limited morphogen copy number, which can become detrimentally low far away from the source, are not considered, neither the timescales needed to both establish and sample such low concentrations far away from the source. While thus, according to the work of the authors, the fluctuations in the morphogen signal may be favorably small, these other factors are supposed to exert a strong limit on positional information. This conclusion therefore seems unjustified and should be toned down, or even better taken out and replaced by a more accurate one, which only focuses on the gradient shape fluctuations, not on the conveyed positional information.

      MINOR COMMENTS

      • p. 1: "and find that positional accuracy is the higher, the narrower the cells".

      (This sentence, however, should be anyhow revised in view of major comment 5 above.)

      • p. 4: "... with an even slightly smaller prefactor."

      Significance

      I believe that this work is significant to the community working on the theoretical foundations of morphogen gradient precision in developmental systems. The main interesting findings are that small cell diameters lead to smaller positional error (although the relevant measure should be clarified according to my comment no. 5), and that the gradient shape fluctuations are surprisingly robust with respect to the readout mechanism.

      Its limitations consist of the fact that the impact of small copy numbers on the readout and associated timescales are neglected, such that the findings of the authors on gradient robustness cannot be simply transferred by simple conversion formulas to readout robustness / positional information. Comment 5 goes hand in hand with this, as a different conclusion may emerge depending on how the relevant positional error measure is defined. This should be fixed by the authors as indicated in the main part of the report.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      How morphogen gradients yield to precise patterning outputs is an important problem in developmental biology. In this manuscript, Adelmann et al. study the impact of cell size in the precision of morphogen gradients and use a theoretical framework to show that positional error is proportional to the square root of cell diameter, suggesting that the smaller the cells in a patterning field, the more precise patterns can be established against morphogen gradient variability. This result remains true even when cells average the morphogen signal across their surface or spatial correlations between cells are introduced. Thus, the authors suggest that epithelial tissues patterned by morphogen gradients buffer morphogen variability by reducing apical cell areas and support their hypothesis by examining several experimental examples of gradient-based vs. non-gradient-based patterning systems.

      Major comments:

      1. While the idea that smaller cells yield to more precise morphogen gradient outputs is attractive, it is unclear whether patterning systems use this strategy to make patterns more precise, as there are several mechanisms that could achieve precision. Do actual developmental systems use it as a mechanism to increase precision? Or precision is achieved through other mechanisms (for example, cell sorting as in the zebrafish neural tube; Xiong et al. Cell, 2013). Indeed, classical patterning work on Drosophila embryo suggest that segmentation patterns are of an absolute size rather by an absolute number of cells (Sullivan, Nature, 1987). According to the authors, the patterning stripes should be more precise when embryos have higher cell densities than in the wild-type, but stripes are remarkably precise in wild-type embryos. This is likely due to other precision-ensuring mechanisms (such as downstream transcriptional repressors, in this case).

      2. Their modeling approach is based on exponential gradients formed by diffusion and linear degradation, but in reality, actual morphogen gradients are affected by receptor and proteoglycan binding and are likely not simply exponential and/or interpreted at the steady state. Do the main results of the manuscript hold even for non-exponential gradients or before they reach a steady state?

      3. In their Discussion section, the authors note that several patterning systems, such as the Drosophila wing and eye discs, show smaller cells near the morphogen source relative to other regions in the tissue. This observation suggests a prediction of the authors' hypothesis that can be tested experimentally. In the Drosophila wing and eye discs genetic mosaics of ectopic morphogen sources (such as Dpp) can (and have) been made. Therefore, one could predict that the patterning outputs in a region of larger cross-sectional areas will be more imprecise than in the endogenous source. Since this is a theoretical paper, it is understandable that authors are not going to make this experiment themselves, but I wonder if they can use published data to test this prediction or at least mention it in the manuscript to offer the experimental biology reader an idea of how their hypothesis can be tested experimentally.

      Other comments:

      • The Methods section should be expanded and should include more details about how authors consider cell size in their simulations. As presented, I believe that experimental biologists will not be able to grasp how the analysis was done.

      • Authors use adjectives such as 'little' as 'small' without a comparative reference. For example in the abstract, the authors say that apical areas "are indeed small in developmental tissues." What does "small" mean? This should be avoided throughout the text.

      Significance

      Overall, I believe that the manuscript is well written and deserves consideration for publication. However, authors should consider the points outlined above in order to make their manuscript more accessible and relevant to the developmental biology community.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      • In developing systems, morphogens gradients pattern tissues such that cells along the patterning length sense varying levels of the morphogen. This process has a low positional error even in the presence of biological noise in numerous tissues including the early embryo of the Drosophila melanogaster. The authors of this manuscript developed a mathematical model to test the effect of noise and mean cell diameter on gradient variability and the positional error they convey.

      • They solved the 1D reaction-diffusion equation for N cells with diameters and kinetic parameters sampled from a physiologically relevant mean and coefficient of variation (CV). They fit the resulting morphogen gradients to a hyperbolic cosine profile and determined the decay length (DL) and amplitude (A) for a thousand independent runs and reported the CV in DL and A.

      • The authors found that CV in DL and A increases with increase in mean cell diameter. They propose a mathematical relationship between CV in DL scales as an inverse-square-root of N. Whereas the CV in DL and A is a weak function of CV of cell surface area (CVa) if CVa < 1.

      • They further looked at the shift in readout boundaries and compared four different readout metrics: spatial averaging, centroid readout, random readout and readout along the length of the cilium. Their results show that spatial averaging and centroid have a high readout precision.

      • They finally showed that the positional error (PE) increases along the patterning length of the tissue and increases with increasing mean cell diameter.

      • The authors also supported their theoretical and simulated results by looking at mean cell areas reported for in patterning tissues in literature which also have a higher readout precision with smaller cell diameters.

      Major comments:

      • Most of the key conclusions are convincing. However, there are four major points that should be addressed. First, the authors conclude the section titled, "The positional error scales with the square root of the average cell diameter," by saying that morphogen systems with small cells can have high precision in absolute length scales, but not on the scale of one cell diameter. They state this would result in salt and pepper patterns in the transition zones. The authors should either support this with biological examples or explain why this is not observed experimentally.

      • Second, perhaps the main conclusion of the paper is that morphogen gradients pattern best when the average cell diameter is small. The authors support this by reviewing the apical cell area of epithelial systems that are known to be patterned by morphogens and those that are not (presumably taking apical cell area as a proxy for cell diameter). However, the key parameter is not absolute cell diameter, but the cell diameter relative to the morphogen length scale. The authors should report the ratio of these two quantities in their literature analysis.

      • Third, as part of their literature analysis, the authors state that in the Drosophila syncytium, there are morphogen gradients, but they imply that because these gradients operate prior to cellularization, one cannot use the large distances between nuclei as counter evidence to their main conclusion. Rather than simply dismissing the case of the Drosophila syncytium, the authors should explain why this case does not apply, using reasoning based on their model assumptions.

      • Fourth, related to the above: the authors then state that there are no morphogen gradients known during cellularization. Unless I am misunderstanding their point, this is untrue. The Dpp gradient acts during the process of cellularization and specifies at least three distinct spatial domains of gene expression. Furthermore, not long after gastrulation, EGFR signaling patterns the ventral ectoderm into at least two distinct domains of gene expression. What are the cell areas in that case?

      Minor comments:

      • Figs 1cd:

      The way the system is set-up: (DL = 20 micron, Patterning Length (LP) = 250 micron, Nominal cell diameter (D) = 5 micron) the DL/L ~ 0.08 which makes the exponential profile far to a small value around 100 micron. This means in all these simulations, the LP was only around 100 micron, cells beyond that saw nearly zero concentration. Because of this, when diameters were varied from 0.2 - 40 micron, there could be as few as 2.5 cells in the "patterning region" which could be responsible for higher variability in DL and A.

      Would any of the results change if DL/L was higher, around 0.2?

      The source region is 25 microns in length and all cell diameters above 25 micron get defaulted back to 25 micron which explains the flatness lines in the region beyond mu_delta/mu_DL> 1

      Results:

      Pg 2 (bottom left): In the git repository code, the morphogen gradients are fit to a hyperbolic cosines function (described in reference 19) which is not described in the main text. Having this in the main text would help readers understand why fig 1c has variation in d only, D only and all k parameters whereas fig 1d has variation with all individual parameters p, d and D and all k.

      • Figure 3b:

      In figures where markers are overlapping perhaps the authors can use a "dot" to identify one set of simulations and a "o" to identify the ones under it. The way the plots are set up currently makes it hard for the reader to understand where certain points on the plot are.

      Methods:

      The Methods can be more descriptive to include certain aspects of the simulations such as adjusted lambda which is only described in the code and not the main text or supplementary.

      Git code:

      The git code function handles do not represent figure numbers and should be updated to make it easier for readers to find the right code

      Significance

      This manuscript contributes certain key aspects to the patterning domain. The three most important contributions of this work to the current literature are: (1) the scaling relationships developed here are important, (2) the idea that PE increases at the tail-end of the morphogen profile is nicely shown and (3) Comparison of various readout strategies.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Reviewer #1 (Evidence, reproducibility and clarity):

      Summary:<br /> The Authors report on the synthesis and characterization of a class of small molecules, the tanshinone mimics (TMs), which interfere with binding of the RNA binding protein (RBP) HuR to its mRNA targets. HuR is an important regulator of mRNA stability and translation of genes involved in key homeostatic (cell cycle, stress response) and pathologic process (inflammation, carcinogenesis). In particular, the first part of the study describes the compounds' chemical synthesis and some pharmacokinetic parameters (i.e., definition of molecular binding, solubility, bioavailability, prodrug generation etc). The second part undertakes, in in vitro and ex-vivo model of LPS-induced mouse macrophage activation, the identification of HuR-bound mRNA targets, which is then evaluated within the global LPS-induced transcriptome; finally, the study evaluates the ability of TMs to inhibit HuR-mediated proinflammatory gene regulation, indicating their use and potential value as therapeutic anti-inflammatory strategy.<br /> Major comments:<br /> The manuscript contains a wealth of data generated from different experimental systems, spanning from synthetic chemistry to preclinical models of gene regulation, requiring cultural backgrounds in chemistry and biology as well. The key conclusions are well supported by the data, but it takes a great effort to get to the core results and thus critically read and evaluate their interpretation. Although the complexity and sheer size of data sets generated lends itself to a hard read, this is further complicated by data presentation, which especially in the second part needs to be significantly improved to gain clarity and focus. For ease of referral, specific comments will be addressed related to Figures whenever possible.<br /> 1.1 • Page 15: To measure TM7nox disrupting ability of HuR:mRNA complex for the HTRF assay (Figure 2G) and for biotin pull down assay (Figure 5C), it was chosen a biotinylated probe containing the AU rich elements of the TNFα, as known HuR target. Please comment on the rationale, and whether could it be relevant reevaluate these parameters post-hoc, based on the sequences identified in HuR targets more susceptible of modulation by TM compound (listed in table 1, Figure 5 A/B) and based on the absence of regulation of TNFa (Figures 3D, 4D, 7A) found in the tested systems.

      R1.1 - We thank the reviewer for this observation. We have been using the biotinylated probe containing the AU-rich elements of TNFα as a representative probe for HuR for biochemical assays in several articles (PMID: 29313684, PMID: 26553968, PMID: 23951323). As the reviewer suggests, a posteriori, it is worth reevaluating the representative probe to be used for evaluating the disrupting ability of TMs based on the data we present here. Indeed, we will tackle this problem in our following efforts, as it is a meaningful although time-consuming task which is outside of the scope of this manuscript.

      1.2 • Page 16-18: Description of the RNAseq data shown in Figure 3 should be more centered around the main findings regarding the effect of TMnox that are further pursued in the study: that is, (Figure 3B), the 249 downregulated DEGs found modulated by TM7nox in presence of LPS, where was observed a strong enrichment of categories related to the inflammatory response: cytokines (Il1b, Cxcl10, Il10, Il19, Il33), immune cell chemotaxis (Ccl12, Ccl22, Ccl17, Ccl6) and innate immune response.

      The description of the GO for the remaining data should be shortened to main points, perhaps reporting what described in the results with each section of the Venn in a table, while referring to the whole list in the supplements as already done. This could replace Figures 3D, E which do not add substantially to what provided in the supplementary table 2 and to which they can be added as further visualization.

      R1.2 - We thank the reviewer for this suggestion, accordingly, we simplified the text keeping only the description of the genes modulated by TM7nox during LPS treatment. The other information originally there was moved to Supplementary table 2. Revised figures 3E and 3F now focus only on the 249 downregulated genes of this group.

      1.3 • Page 18-19: Description of the results of the RIP-seq shown in Figure 4 set is very confusing: onward from the line "477 HuR-bound transcripts (log2 FC > 3) were also upregulated by LPS at the transcriptional level..." the numbers do not match or reconcile with those shown in the Venn diagram (Fig. 4B) nor with those listed in the figure legend of Figure S8.

      R1.3 - We agree with the reviewer, we apologize for having reported the wrong numbers, and we clarified this point in general by deeply revising the text. A more precise explanation of the selection procedure for the genes of interest is now reported and better explained also by adding a scheme (Fig 4D in the revised manuscript).

      1.4 Moreover, as previously remarked for Figure 3 (and even more for this dataset in which initial description of Venn in 4B is unclear), panel 4E does not add as much to the info provided in Table 1/supplementary Table 1, where they can eventually be added as further data visualization; Instead, Figure S8 displays very informative data merging together the results obtained in RNAseq (Fig. 3) and RIP-Seq (Fig.4) and should be displayed in Figure 4, as in the result section they are indeed presented together.

      R1.4 - We agree with this remark, thus we have removed the old panels 3E in S8C and 4E in S9B, and we now provide the information previously contained in old S8 in the main figure 4E of the revised manuscript.

      1.5 • Page 19-20: Description of the modulation by TM7nox of HuR binding to specific consensus sequences is summarized at the end of the relative paragraph as follows: "TM7nox reshapes HuR binding to target genes in presence of LPS by disrupting the binding of HuR towards target genes containing a lower number of HuR consensus sequences than the average observed in the HuR-bound transcripts". Understanding of these data through the provided text and the Supplementary Figure 9 is very laborious and referring of an entire dataset to a supplementary figure makes it even harder. It would be best to show this as main figure, not supplemental, either adding a Venn diagram as in 3B/4B showing to which dataset each part of the analysis refers, or even more efficaciously, extrapolate a representative gene set for the main analyses showing TM7nox activity in target genes with higher vs lower consensus sequences; same approach for the analysis in Figure 9B, where the effect on a gene with sequence #1 or #10 could be compared with one bearing sequence #3 for example.

      R1.5 - We agree with the reviewer, thus we moved the information of old S9 in figure 4C of the revised manuscript. We deeply revised the information provided also by taking into account the request to compare this experiment to the one in Lal et al. NAR 2017 (please see also R2.4). We made an effort to identify a subset of genes that follow a coherent modulation, identifying 82 genes highlighted in Supplementary Table 1. All such genes show increased expression during LPS or LPS/TMnox vs DMSO conditions, and decreased association to HuR during LPS/TMnox vs LPS. As 47 of these, i.e. more that 50%, contain less AU rich sequences than the average (highlighted in Supplementary Table 1), we can consider them as a representative gene ensemble modulated in accordance with the presence of AU rich sequences.

      1.6 • Page 21: Description of the effect of three TMs (TM6, TM7nox and TM7nred) on LPS response in macrophages at the single gene level (Figure 5 and Figure 6): TM6 and TM7nox were used in exps in Fig. 5 A and E, while only TM7nred was used for CXCL10 secretion analysis (fig.5 D and F): please describe the compound choices' rationale (as done for experiments in Figure 6).

      R1.6 – Following the reviewer suggestion, we now explain our rationale in choosing the small molecules, that is the use of the ones bearing the active quinone species. We have performed additional experiments, and now we report TM6n, TM7nox, and the control DHTS activity in decreasing the secretion of Cxcl10 (figure 5E in the revised manuscript). All compounds behave similarly in this experiment. TM7nred is now used to show its equivalence to TM7nox in figure 5E and in figure 6 of the revised manuscript.

      1.7 • Page 21-22: The effect on HuR expression of siRNA silencing and, more importantly, of TMs shown in Figure 6A, first panel, should be visualized at protein level by western blot. This is an important point as for CXCL10 and iL1b there seems to be an additive effect between decreased HuR levels and pharmacological blocking.

      R1.7 - Following the reviewer suggestion, we now show the protein level as measured by intracellular Elisa; as we were not able to detect the proteins by western blot. The protein level is in general agreement with the gene expression level. We do not observe an additive effect by pharmacological inhibition during HuR silencing, but we rather observe a slight increase in the protein level during HuR silencing. We do not have an explanation for this effect, which may depend on several reasons - for example, an aspecific effect of the TMs when their molecular target HuR is absent.

      1.8 • Page 24: please rephrase the statement 'These observations suggest the utilization of TMs in autoinflammatory and autoimmune diseases' as 'These observations suggest the evaluation of TMs in specific preclinical models for autoinflammatory and autoimmune diseases'.

      R1.8 - We fully agree with the reviewer, and we changed the text in the revised manuscript accordingly.

      1.9 • In the discussion, please include a paragraph with study limitation and possible biases (for example, the choice of RNP-IP without crosslinking has pros and cons).

      R1.9 – Thank you for the good suggestion, we added a paragraph in the discussion which describes study limitations due to the utilization of RNP-IP vs crosslinking.

      The data and the methods are correctly presented for reproducibility, replicates and statistical analysis are adequate. Minor comments: 1.10 • At least in the single gene validation experiments (Fig.5), a negative control (such as recombinant HuR with mutated RRMs in trans-, or ARE-less/non-HuR targetable sequence in cis, or inactive TM) would be advisable.

      R1.10- We thank the reviewer for the suggestion. Accordingly, we used an ARE-less/non-HuR targetable gene as RPLP0 for validation.

      1.11 • Figure 6B/C: for immunofluorescence panels, zooming on a smaller number of cells will render more visible HuR and NFkB nucleocytoplasmic shuttling, given that quantification and statistics are provided by imaging software. Negative control stainings (secondary Abs only) should be included.

      R1.11 – In accordance with this suggestion, we now report a higher magnification of the immunofluorescence images. We also report the standard DHTS effect, showing a difference vs TMnox activity which may suggest its impact on NFkB shuttling.

      1.12 • Figure 7A: in the X axis LPS+8n is indicated: is it a typo for LPD+6n or was compound TM8n indeed used?

      R1.12 – Thanks for your spotting our mistake, the prodrug 8 described in figure 1 was used.

      1.13 • In the Methods section please include protocols and materials for immunofluorescence (results shown in Fig. 6B/C).

      R1.13 – As for your suggestion, protocols and materials for immunofluorescence were added to the methods.

      1.14 • There are some typos and repetition in figure legends (legend Figure S9).

      R1.14- Thank you for this, we revised all the figure legends.

      Prior studies are referenced appropriately. Review Cross-commenting I fully agree with the Reviewer's remarks. I would add that a general concern expressed is that this manuscript in its present form has a readership issue: the first part is for chemistry/pharmacology audience, the second is biology-based. Splitting the work has been suggested; or, the Authors may decide which part is more impactful and present the other in a streamlined version.

      Reviewer #1 (Significance):

      This is a large study reporting progress in the development of synthetic antagonists of HuR function, which is the Authors' well-established line of research. The TM compounds are small molecules with anti-inflammatory effects with strong potential for therapeutic use due to selected inhibition of HuR-mediated upregulation of proinflammatory molecules. The physicochemical and early biological characterization done in this study will allow further testing of their efficacy and of the overall role of HuR-mediated regulation as targetable mechanism in several preclinical human disease models. Targeting of the RNA-binding protein HuR has been tackled as therapeutic approach in cancer, less in chronic immune and inflammatory diseases despite many common mechanisms and mediators. This study could be well received by researchers involved in basic science and drug development (chemistry, biochemistry/biophysics, pharmacology, computational modeling) and biologists/physician scientists interested in testing these compounds in translational research settings where HuR-driven functions can be relevant (cancer, chronic inflammation), though the chemical part would be less accessible to the latter audience. Reviewer's background is in preclinical human models of chronic inflammation with interest in posttranscriptional gene regulation with familiarity with RNAseq and RIPseq dataset and analysis. For the part of the manuscript regarding the synthesis and physicochemical characterization of the TN compound I requested assistance to a faculty from the chemistry department with expertise in that field, who did not request any specific clarification or addendum.

      Reviewer #2 (Evidence, reproducibility and clarity):

      In the manuscript entitled "HuR modulation with tanshinone mimics impairs LPS response in murine macrophages" the authors have described the synthesis and application of small molecule mimics of the naturally occurring compound tanshinone, which is known to inhibit the binding of the RBP HuR to a class of its mRNA targets. The authors have shown that the tanshinone mimics (TMs) used by them block the binding of RRM1-2 of HuR to ARE-containing RNA in vitro, and reduce the interaction of HuR with a set of ARE-containing mRNAs in LPS-treated mouse macrophage cells. This reduction of interaction of HuR with some of these mRNAs correlates with the reduction in their level in the cells treated with the TMs, and in the secreted level of their proteins in the serum of animals with LPS-induced peritonitis. Together, the study demonstrates the role of these TMs as modulators of the LPS-induced inflammatory response by blocking the binding of HuR to a subset of LPS-induced inflammatory mRNAs and thereby downregulating their mRNA and protein levels in inflammatory cells. The manuscript describes a comprehensive study, ranging from chemical synthesis of TMs, MD simulations to demonstrate the binding site of the TMs to the cleft formed by the RRM1-linker-RRM2 domains of HuR, which has been shown in crystal structure to be the main binding site of A/U-rich RNA molecules, in vitro studies showing the ability of the TMs to hinder ARE-containing RNA binding to HuR RRM1-2, whole transcriptome analysis to show the effect of the TMs on LPS-induced differential gene expression in murine macrophages, and on HuR binding to target mRNAs, and animal studies to show the effect of the TMs on the level of some inflammatory mediators in the serum of mice with LPS-induced peritonitis. The results are quite convincing and is in line with what is generally known about the effect of HuR on the expression of a large number of genes encoding pro-inflammatory proteins, and the ability of tanshinone derivatives/mimics in inhibiting HuR binding to target mRNAs. The authors put these two information together in this study and the results are on expected lines. While the results are convincing and quite comprehensive, I would suggest the following in order to substantiate and strengthen the results: 2.1. The experiments do not have any "positive control", such that the performance of the TMs can be compared with that of a molecule with known HuR binding inhibition activity, such as DHTS. It would be good to have such a comparison, to understand whether the TMs work similar to DHTS or differently, both qualitatively in terms of the mRNA targets which they affect and the extent of their anti-inflammatory activity.

      R2.1- We added DHTS as a comparison to TMs, following the reviewer’s comment. In this model, the net effect of DHTS is partially overlapping with TMs, at least for the parameters that we checked (see Figure 5, 6 and 7), showing some differences in the modulation of NF-kB shuttling upon LPS stimulation. Therefore, we suggest that DHTS and TMs show partially different effects on mRNA targets and in terms of anti-inflammatory activities.

      2.2. It is not clear to me whether the mRNAs which show differential expression in the RNAseq analysis of cells treated with LPS and TMs are exactly the ones which show difference in binding with HuR in the RIPseq analysis in presence of the TMs. This analysis is important for a number of reasons: all the mRNA binding targets of HuR are not affected by HuR at the level of mRNA stability, many are affected at the level of translation, without change in mRNA level. These mRNAs should therefore show change in binding of HuR in the RIPseq assay in presence of TM, but not show change in expression. Secondly, there may be mRNAs which show a change in expression in presence of TMs, but do not show binding of HuR, suggesting pleiotropic roles of the TMs. Therefore, instead of an overall correlation between differential expression and change in HuR binding of mRNAs, a table comparing the RIPseq status of individual mRNAs with that of their differential expression status, in presence and absence of LPS/TMs would be useful, further designating the different groups of mRNAs based on these differential status (change in HuR binding/change in expression, change in HuR binding/no change in expression etc.).

      R2.2 – We tried to rationalize the data following the reviewer’ suggestion, however, we could not fully adopt this strategy due to the complexity of the experiment design. Indeed, we have focused our attention on the effect of TMs during LPS stimulus, which induces a strong transcriptional response, rather than in steady state conditions. This is why we reported the overall correlation of LPS vs DMSO and TM7nox/LPS vs DMSO. Then, we evaluated whether the observed difference in the correlation may be reflected on a change of HuR binding, and we checked the RIPseq status during co-treatment vs LPS. This was the case for a subset of genes that are reported in Supplementary Table 1. Nevertheless, to be fully compliant with the reviewer’s request we now report a Supplementary Table 1 containing the entire gene list, so that the reader can immediately filter out the subsets according only to the comparison TM7nox/LPS vs LPS.

      2.3. Nuclear/cytoplasmic localization of HuR and NFkb is impossible to discern at the magnification of the immunofluorescence images in Fig 6 B and C. Higher magnification images are required to understand changes in localization.

      R2.3 – In accordance with this suggestion, we now report higher magnification, please see also R1.11. We do not observe any change in nuclear/cytoplasmic localization of HuR and NFkb due to TMs treatment. We rather observe LPS-induced NFkB nuclear accumulation, ActD-induced HuR cytoplasmic shuttling and inhibition of NFkB translocation, during LPS and DHTS treatment.

      2.4. It has been shown that DHTS-I increases the binding of HuR to the mRNAs with longer 3'UTR and with higher density of U/AU-rich elements, whereas it reduces the interaction of HuR with the mRNAs having shorter 3'UTR and with low density of U/AU-rich elements (Lal et al., NAR, 2017). It is not clear if the same is observed in case of the TMs or not, and such a comparative analysis would be useful to address this point.

      R2.4 – We re-analysed the data, checking the density of U/AU rich elements and the length of the 3’UTR of the displaced mRNA as in Lal et al. NAR 2017. Although we could not compare DHTS and TMs within the same biological system, it appears that the rules dictating their mechanism of action are similar.

      I think that the above suggested points are feasible as most of them really involve re-analysis of existing data. Only the suggestion to add DHTS or tanshinone as a positive/comparison control will require experimentation and addition of new data.

      Review Cross-commenting

      I think most of the reviewers' comments coincide in the evaluation of the manuscript. I would especially like to draw attention to the fact that all three reviewers found that the content and form of data presented in the paper is very dense and bogs down the reader and distracts from the overall focus of the manuscript.

      Reviewer #2 (Significance):

      The work described in the manuscript is comprehensive as it ranges from chemical synthesis and in vitro evaluation of the TMs to the characterization of their effects in vivo. Although the effect of tanshinone derivatives on HuR mRNA target binding is known, and the effect of HuR on inflammatory gene expression is also known, the manuscript is significant as it brings these two information together and tests the effect of these TMs on HuR-mediated regulation of inflammatory gene expression.<br /> However the extensiveness of the work also makes it quite dense, and I feel that the focus of the paper is often lost in the details. Also, the text of the manuscript is dense and verbose and uses many irregular grammatical and phraseological usages, for eg "their<br /> modulation or mis-localization lead to the insurgence of complex phenotypes and diseases". It appears to me that it would be ideal if the chemical synthesis, MD simulation studies and in vitro studies are presented in an independent manuscript. Also, that would allow a more exhaustive referencing of the known studies in literature where tanshinone derivatives, and other small molecules, have been used to modulate HuR binding to mRNA targets.<br /> This work would be of interest to molecular cell biologists in general and RNA biologists in particular, especially those who are studying RNA-protein interactions, and scientists who are interested in drug development using RNA-protein interactions as drug targets.<br /> My interest in the work lies in my expertise in studying RNA-protein interactions, especially of RNA-binding proteins such as HuR involved in regulating the translation of mRNAs encoded by inflammatory genes. I do not have expertise in chemical synthesis and am therefore not qualified to evaluate the first set of results describing the chemical synthesis of TMs.

      Reviewer #3 (Evidence, reproducibility and clarity):

      In this study, the authors investigated the modulation of HuR by tanshinone mimics and how it mitigates LPS response in murine macrophages. This represents a nice integration of synthetic chemistry, molecular simulations, and in vitro as well as in vivo experimental validations. Overall, this is an interesting study, and will add to the growing interest in HuR in inflammatory-mediated disease. The paper contains a lot of data (actually several papers in one) which may bog down the reader and distract from the overall message. it is suggested that they condense the data and simplify the figures and use more supplemental figures.<br /> Major Comments:<br /> 3.1. The authors have shown the dose response and cytotoxicity effect of tanshinone mimics; The authors show that TMs affect the overall HuR mRNA but they don't show protein levels.

      R3.1 – In accordance with the reviewer’s comment, we now show also protein levels, as we performed intracellular ELISA (Figure 6 in the revised manuscript); please see also R1.7.

      3.2. It is unclear the timing of certain experiments for LPS vs TMs (whether macrophages were pre-treated with TMs before LPS)-e.g fig 5. The authors should clarify for all experiments as the long-term clinical paradigm would be treatment after inflammation has been established.

      R3.2 – In most experiments TMs are co-administered with LPS. Only in one of the two protocols used for Actinomycin D chase experiment TMs are added after LPS with Act D, as we wanted to discriminate between transcriptional and post-transcriptional effects of TMs (see also R3.3).

      3.3. They have also identified differentially expressed genes which are RNA binding ligands of HuR by RIP-Seq. However, it would be necessary to check whether TM7nox affects the stability of those targets before conclusions can be made that TMs don't inhibit the primary transcriptional response (as mentioned in the Discussion section). Transcriptional effects of HUR chemical inhibition or genetic silencing has been reported previously in other cell systems.

      R3.3 – The reviewer is entirely correct, and we accordingly amended our conclusions. Indeed, TMs have an impact on gene transcription during co-administration with LPS as now suggested by Actinomycin D chase experiments reported in Figure 6C in the revised data and discussion in the manuscript.

      3.4. HuR competes with many RBPs (e.g. TTP and KSRP) as well as microRNAs (including miR-21 and miR-122) to regulate the stability/translational efficiency of several AU-rich transcripts. Does TM binding to HuR lead to increase access of these RBPs/microRNA to the transcripts? This could be addressed by RNA IP with antibodies to TTP or KSRP.

      R3.4 – The reviewer is suggesting an important experiment that requires multiple controls and significant efforts. Indeed, we are planning to study the specificity of TMs, and we prefer to tackle and report this point in a later publication.

      3.5. Another aspect of HuR functioning is the dimerization of HuR. HuR dimerization has been linked with many pathophysiologic conditions. The authors should show the effect of TM7nox on HuR dimerization. In figure 2, for example, there is a suggestion of this in the representative EMSAs where an intermediate shifted band is seen with the addition of TMs. Also, the legend should make clear which ligand is being tested in the modeling (purple structure) versus the RNA probe in the EMSAs. It would help the reader to identify the RNA probe used-e.g. "5′-DY681-labeled ARE RNA probe.

      R3.5 – We agree with the reviewer’s suggestion, and we investigated whether TM7nox influences HuR dimerization in the absence of RNA as performed in PMID 17632515 (Meisner et al 2007). We used MS-444 as a positive control, and we did not observe inhibition of dimerization by TMs at least at the used dosages. Data are reported in Supplementary Figure S6B of the revised manuscript.

      3.6. HuR does alter M2-associated targets like IL-10 and this should be addressed more directly. Fig. 3 suggests that IL-10 is reduced by TM7nox but the variance is so high that the statistics show NS. HuR regulates IL-10 in other cellular contexts and this would be important to determine for TM7 in the long run.

      R3.6 – Although we acknowledge its relevance, however, we did not investigate this gene directly. The variance becomes significant in the RIP-seq experiment (Supplementary Figure 9D). Therefore, we confirm that Il10 is among the 47/82 genes that show the same behavior as Cxcl10, Il1b and many other cytokines as Ccl12, Ccl7, Fas, Il1a, Il33; in conclusion, it is among the restricted list of genes modulated by TM7nox according to the presence of less AU rich sequences than average.

      3.7. Fig. 5-10 um of the TM used here produces significant toxicity to BMDM according to fig. S7. This may distort the ELISA/qPCR results as the RNA levels may be lower due to toxicity. The authors should address this or use a lower dose that is not toxic.

      R3.7 – The viability curves mentioned by the reviewer are run at 24-48 hours, and no toxic effects have been observed using TMs after 6 hours of treatment.

      3.8. In Fig 6 the immunocytochemistry is difficult to interpret as the magnification is too small to appreciate the N/C ratio. The investigators should provide higher magnification. A nuclear/cytoplasmic western blot is recommended as well to confirm that TM does not impair HuR shuttling (or NFkb shifts). This is an important area as there is a suggestion that TM blocks dimerization (Fig. 2) which does impair shuttling. Also, the modeling data suggest that TMs appear to sit in a similar groove between RRM1 and 2 as other HuR inhbitors that block shuttling.

      R3.8 – This point has also been raised by other reviewers, and we replied in R2.3 and R1.11. We understand the reviewer’s points, and we agree with the observation. However, we do not observe a change in HuR nuclear/cytoplasmic shuttling by immunofluorescence, neither we see an effect on HuR dimerization.

      3.9. IL-6 does not appear to be affected by TM treatment after LPS stimulation in vivo or in vitro -either mRNA or protein. However, DHTS did suppress this cytokine. The authors should address this discrepancy. Likewise, TNFa data here show no change and possibly a trend upward (Fig 3,4 and 7). This is in contrast to the effect of DHTS on TNF-a reported by the authors in a prior publication (D'Agnistino et al). The authors should address this discrepancy. There are reports suggesting that HuR is a translational inhibitor of TNFa in macrophages--Katsanou V, Papadaki O, Milatos S, Blackshear PJ, Anderson P, Kollias G, Kontoyiannis DL. HuR as a negative posttranscriptional modulator in inflammation (PMID 16168373)

      R3.9 – The reviewer’s comments are correct, but we do not have an explanation for this. In theory, there could be several possibilities such as 1) a DHTS effect on NFkB, 2) the fact that previously mentioned experiments with DHTS are not run with the same cells-at the same doses and timing as our current TM experiments, and 3) that HuR silencing is only partially overlapping with TMs treatment also in our recent experiments. Irrespective of specific transcripts, we think we have shown that TMs’ mechanism of action involves the modulation of HuR binding at the transcriptional level in our experimental condition.

      Review Cross-commenting

      I think the other reviewers' comments are pertinent and well thought out. I have no further suggestions.

      Reviewer #3 (Significance):

      The characterization of novel HuR inhibitors derived from tanshinones is an important advancement to the field which is rapidly growing. This complements other work with small molecule inhibitors and will allow the field to better understand the role of HuR in different disease contexts (cancer, neuroinflammatory etc) and cell types (e.g. macrophages, microglia, astrocytes). The ultimate significance is the clinical application of the inhibitors and the more options the better, particularly if there are toxic effects of some. My expertise is in post-trasnscriptional regulation of cytokines and we have already characterized some potent effects in cancer.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this study, the authors investigated the modulation of HuR by tanshinone mimics and how it mitigates LPS response in murine macrophages. This represents a nice integration of synthetic chemistry, molecular simulations, and in vitro as well as in vivo experimental validations. Overall, this is an interesting study, and will add to the growing interest in HuR in inflammatory-mediated disease. The paper contains a lot of data (actually several papers in one) which may bog down the reader and distract from the overall message. it is suggested that they condense the data and simplify the figures and use more supplemental figures.

      Major Comments:

      1. The authors have shown the dose response and cytotoxicity effect of tanshinone mimics; The authors show that TMs affect the overall HuR mRNA but they don't show protein levels.
      2. It is unclear the timing of certain experiments for LPS vs TMs (whether macrophages were pre-treated with TMs before LPS)-e.g fig 5. The authors should clarify for all experiments as the long-term clinical paradigm would be treatment after inflammation has been established.
      3. They have also identified differentially expressed genes which are RNA binding ligands of HuR by RIP-Seq. However, it would be necessary to check whether TM7nox affects the stability of those targets before conclusions can be made that TMs don't inhibit the primary transcriptional response (as mentioned in the Discussion section). Transcriptional effects of HUR chemical inhbiition or genetic silencing has been reported previously inother cell systems.
      4. HuR competes with many RBPs (e.g. TTP and KSRP) as well as microRNAs (including miR-21 and miR-122) to regulate the stability/translational efficiency of several AU-rich transcripts. Does TM binding to HuR lead to increase access of these RBPs/microRNA to the transcripts? This could be addressed by RNA IP with antibodies to TTP or KSRP.
      5. Another aspect of HuR functioning is the dimerization of HuR. HuR dimerization has been linked with many pathophysiologic conditions. The authors should show the effect of TM7nox on HuR dimerization. In figure 2, for example, there is a suggestion of this in the representative EMSAs where an intermediate shifted band is seen with the addition of TMs. Also, the legend should make clear which ligand is being tested in the modeling (purple structure) versus the RNA probe in the EMSAs. It would help the reader to identify the RNA probe used-e.g. "5′-DY681-labeled ARE RNA probe.
      6. HuR does alter M2-associated targets like IL-10 and this should be addressed more directly. Fig. 3 suggests that IL-10 is reduced by TM7nox but the variance is so high that the statistics show NS. HuR regulates IL-10 in other cellular contexts and this would be important to determine for TM7 in the long run.
      7. Fig. 5-10 um of the TM used here produces significant toxicity to BMDM according to fig. S7. This may distort the ELISA/qPCR results as the RNA levels may be lower due to toxicity.The authors should address this or use a lower dose that is not toxic.
      8. In Fig 6 the immunocytochemistry is difficult to interpret as the magnification is too small to appreciate the N/C ratio. The investigators should provide higher magnification and provide examples of ActD, LPS and LPS + drug. A nuclear/cytoplasmic western blot is recommended as well to confirm that TM does not impair HuR shuttling (or NFkb shifts). This is an important area as there is a suggestion that TM blocks dimerization (Fig. 2) which does impair shuttling. Also, the modeling data suggest that TMs appear to sit in a similar groove between RRM1 and 2 as other HuR inhbitors that block shuttling.
      9. IL-6 does not appear to be affected by TM treatment after LPS stimulation in vivo or in vitro -either mRNA or protein. However, DHTS did suppress this cytokine. The authors should address this discrepancy. Llikewise, TNFa data here show no change and possibly a trend upward (Fig 3,4 and 7). This is in contrast to the effect of DHTS on TNF-a reported by the authors in a prior publication (D'Agnistino et al). The authors should address this discrepancy. There are reports suggesting that HuR is a translational inhbitor of TNFa in macrophages--Katsanou V, Papadaki O, Milatos S, Blackshear PJ, Anderson P, Kollias G, Kontoyiannis DL. HuR as a negative posttranscriptional modulator in inflammation (PMID 16168373)

      Review Cross-commenting

      I think the other reviewers' comments are pertinent and well thought out. I have no further suggestions.

      Significance

      The characterization of novel HuR inhibitors derived from tanshinones is an important advancement to the field which is rapidly growing. This complements other work with small molecule inhibitors and will allow the field to better understand the role of HuR in different disease contexts (cancer, neuroinflammatory etc) and cell types (e.g. macrophages, microglia, astrocytes). The ultimate significance is the clinical application of the inhibitors and the more options the better, particularly if there are toxic effects of some. My expertise is in post-trasnscriptional regulation of cytokines and we have already characterized some potent effects in cancer.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In the manuscript entitled "HuR modulation with tanshinone mimics impairs LPS response in murine<br /> macrophages" the authors have described the synthesis and application of small molecule mimics of the naturally occurring compound tanshinone, which is known to inhibit the binding of the RBP HuR to a class of its mRNA targets. The authors have shown that the tanshinone mimics (TMs) used by them block the binding of RRM1-2 of HuR to ARE-containing RNA in vitro, and reduce the interaction of HuR with a set of ARE-containing mRNAs in LPS-treated mouse macrophage cells. This reduction of interaction of HuR with some of these mRNAs correlates with the reduction in their level in the cells treated with the TMs, and in the secreted level of their proteins in the serum of animals with LPS-induced peritonitis. Together, the study demonstrates the role of these TMs as modulators of the LPS-induced inflammatory response by blocking the binding of HuR to a subset of LPS-induced inflammatory mRNAs and thereby downregulating their mRNA and protein levels in inflammatory cells.

      The manuscript describes a comprehensive study, ranging from chemical synthesis of TMs, MD simulations to demonstrate the binding site of the TMs to the cleft formed by the RRM1-linker-RRM2 domains of HuR, which has been shown in crystal structure to be the main binding site of A/U-rich RNA molecules, in vitro studies showing the ability of the TMs to hinder ARE-containing RNA binding to HuR RRM1-2, whole transcriptome analysis to show the effect of the TMs on LPS-induced differential gene expression in murine macrophages, and on HuR binding to target mRNAs, and animal studies to show the effect of the TMs on the level of some inflammatory mediators in the serum of mice with LPS-induced peritonitis. The results are quite convincing and is in line with what is generally known about the effect of HuR on the expression of a large number of genes encoding pro-inflammatory proteins, and the ability of tanshinone derivatives/mimics in inhibiting HuR binding to target mRNAs. The authors put these two information together in this study and the results are on expected lines. While the results are convincing and quite comprehensive, I would suggest the following in order to substantiate and strengthen the results:

      1. The experiments do not have any "positive control", such that the performance of the TMs can be compared with that of a molecule with known HuR binding inhibition activity, such as DHTS. It would be good to have such a comparison, to understand whether the TMs work similar to DHTS or differently, both qualitatively in terms of the mRNA targets which they affect and the extent of their anti-inflammatory activity.
      2. It is not clear to me whether the mRNAs which show differential expression in the RNAseq analysis of cells treated with LPS and TMs are exactly the ones which show difference in binding with HuR in the RIPseq analysis in presence of the TMs. This analysis is important for a number of reasons: all the mRNA binding targets of HuR are not affected by HuR at the level of mRNA stability, many are affected at the level of translation, without change in mRNA level. These mRNAs should therefore show change in binding of HuR in the RIPseq assay in presence of TM, but not show change in expression. Secondly, there may be mRNAs which show a change in expression in presence of TMs, but do not show binding of HuR, suggesting pleiotropic roles of the TMs. Therefore, instead of an overall correlation between differential expression and change in HuR binding of mRNAs, a table comparing the RIPseq status of individual mRNAs with that of their differential expression status, in presence and absence of LPS/TMs would be useful, further designating the different groups of mRNAs based on these differential status (change in HuR binding/change in expression, change in HuR binding/no change in expression etc.).
      3. Nuclear/cytoplasmic localization of HuR and NFkb is impossible to discern at the magnification of the immunofluorescence images in Fig 6 B and C. Higher magnification images are required to understand changes in localization.
      4. It has been shown that DHTS-I increases the binding of HuR to the mRNAs with longer 3'UTR and with higher density of U/AU-rich elements, whereas it reduces the interaction of HuR with the mRNAs having shorter 3'UTR and with low density of U/AU-rich elements (Lal et al., NAR, 2017). It is not clear if the same is observed in case of the TMs or not, and such a comparative analysis would be useful to address this point.

      I think that the above suggested points are feasible as most of them really involve re-analysis of existing data. Only the suggestion to add DHTS or tanshinone as a positive/comparison control will require experimentation and addition of new data.

      Review Cross-commenting

      I think most of the reviewers' comments coincide in the evaluation of the manuscript. I would especially like to draw attention to the fact that all three reviewers found that the content and form of data presented in the paper is very dense and bogs down the reader and distracts from the overall focus of the manuscript.

      Significance

      The work described in the manuscript is comprehensive as it ranges from chemical synthesis and in vitro evaluation of the TMs to the characterization of their effects in vivo. Although the effect of tanshinone derivatives on HuR mRNA target binding is known, and the effect of HuR on inflammatory gene expression is also known, the manuscript is significant as it brings these two information together and tests the effect of these TMs on HuR-mediated regulation of inflammatory gene expression.<br /> However the extensiveness of the work also makes it quite dense, and I feel that the focus of the paper is often lost in the details. Also, the text of the manuscript is dense and verbose and uses many irregular grammatical and phraseological usages, for eg "their<br /> modulation or mis-localization lead to the insurgence of complex phenotypes and diseases". It appears to me that it would be ideal if the chemical synthesis, MD simulation studies and in vitro studies are presented in an independent manuscript. Also, that would allow a more exhaustive referencing of the known studies in literature where tanshinone derivatives, and other small molecules, have been used to modulate HuR binding to mRNA targets.<br /> This work would be of interest to molecular cell biologists in general and RNA biologists in particular, especially those who are studying RNA-protein interactions, and scientists who are interested in drug development using RNA-protein interactions as drug targets.<br /> My interest in the work lies in my expertise in studying RNA-protein interactions, especially of RNA-binding proteins such as HuR involved in regulating the translation of mRNAs encoded by inflammatory genes. I do not have expertise in chemical synthesis and am therefore not qualified to evaluate the first set of results describing the chemical synthesis of TMs.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      The Authors report on the synthesis and characterization of a class of small molecules, the tanshinone mimics (TMs), which interfere with binding of the RNA binding protein (RBP) HuR to its mRNA targets. HuR is an important regulator of mRNA stability and translation of genes involved in key homeostatic (cell cycle, stress response) and pathologic process (inflammation, carcinogenesis). In particular, the first part of the study describes the compounds' chemical synthesis and some pharmacokinetic parameters (i.e., definition of molecular binding, solubility, bioavailability, prodrug generation etc). The second part undertakes, in in vitro and ex-vivo model of LPS-induced mouse macrophage activation, the identification of HuR-bound mRNA targets, which is then evaluated within the global LPS-induced transcriptome; finally, the study evaluates the ability of TMs to inhibit HuR-mediated proinflammatory gene regulation, indicating their use and potential value as therapeutic anti-inflammatory strategy.

      Major comments:

      The manuscript contains a wealth of data generated from different experimental systems, spanning from synthetic chemistry to preclinical models of gene regulation, requiring cultural backgrounds in chemistry and biology as well. The key conclusions are well supported by the data, but it takes a great effort to get to the core results and thus critically read and evaluate their interpretation. Although the complexity and sheer size of data sets generated lends itself to a hard read, this is further complicated by data presentation, which especially in the second part needs to be significantly improved to gain clarity and focus. For ease of referral, specific comments will be addressed related to Figures whenever possible.

      • Page 15: To measure TM7nox disrupting ability of HuR:mRNA complex for the HTRF assay (Figure 2G) and for biotin pull down assay (Figure 5C), it was chosen a biotinylated probe containing the AU rich elements of the TNFα, as known HuR target. Please comment on the rationale, and whether could it be relevant reevaluate these parameters post-hoc, based on the sequences identified in HuR targets more susceptible of modulation by TM compound (listed in table 1, Figure 5 A/B) and based on the absence of regulation of TNF (Figures 3D, 4D, 7A) found in the tested systems.
      • Page 16-18: Description of the RNAseq data shown in Figure 3 should be more centered around the main findings regarding the effect of TMnox that are further pursued in the study: that is, (Figure 3B), the 249 downregulated DEGs found modulated by TM7nox in presence of LPS, where was observed a strong enrichment of categories related to the inflammatory response: cytokines (Il1b, Cxcl10, Il10, Il19, Il33), immune cell chemotaxis (Ccl12, Ccl22, Ccl17, Ccl6) and innate immune response. The description of the GO for the remaining data should be shortened to main points, perhaps reporting what described in the results with each section of the Venn in a table, while referring to the whole list in the supplements as already done. This could replace Figures 3D, E which do not add substantially to what provided in the supplementary table 2 and to which they can be added as further visualization.
      • Page 18-19: Description of the results of the RIP-seq shown in Figure 4 set is very confusing: onward from the line "477 HuR-bound transcripts (log2 FC > 3) were also upregulated by LPS at the transcriptional level..." the numbers do not match or reconcile with those shown in the Venn diagram (Fig. 4B) nor with those listed in the figure legend of Figure S8. Moreover, as previously remarked for Figure 3 (and even more for this dataset in which initial description of Venn in 4B is unclear), panel 4E does not add as much to the info provided in Table 1/supplementary Table 1, where they can eventually be added as further data visualization; Instead, Figure S8 displays very informative data merging together the results obtained in RNAseq (Fig. 3) and RIP-Seq (Fig.4) and should be displayed in Figure 4, as in the result section they are indeed presented together.
      • Page 19-20: Description of the modulation by TM7nox of HuR binding to specific consensus sequences is summarized at the end of the relative paragraph as follows: "TM7nox reshapes HuR binding to target genes in presence of LPS by disrupting the binding of HuR towards target genes containing a lower number of HuR consensus sequences than the average observed in the HuR-bound transcripts". Understanding of these data through the provided text and the Supplementary Figure 9 is very laborious and referring of an entire dataset to a supplementary figure makes it even harder. It would be best to show this as main figure, not supplemental, either adding a Venn diagram as in 3B/4B showing to which dataset each part of the analysis refers, or even more efficaciously, extrapolate a representative gene set for the main analyses showing TM7nox activity in target genes with higher vs lower consensus sequences; same approach for the analysis in Figure 9B, where the effect on a gene with sequence #1 or #10 could be compared with one bearing sequence #3 for example.
      • Page 21: Description of the effect of three TMs (TM6, TM7nox and TM7nred) on LPS response in macrophages at the single gene level (Figure 5 and Figure 6): TM6 and TM7nox were used in exps in Fig. 5 A and E, while only TM7nred was used for CXCL10 secretion analysis (fig.5 D and F): please describe the compound choices' rationale (as done for experiments in Figure 6).
      • Page 21-22: The effect on HuR expression of siRNA silencing and, more importantly, of TMs shown in Figure 6A, first panel, should be visualized at protein level by western blot. This is an important point as for CXCL10 and iL1there seems to be an additive effect between decreased HuR levels and pharmacological blocking.
      • Page 24: please rephrase the statement 'These observations suggest the utilization of TMs in autoinflammatory and autoimmune diseases' as 'These observations suggest the evaluation of TMs in specific preclinical models for autoinflammatory and autoimmune diseases'.
      • In the discussion, please include a paragraph with study limitation and possible biases (for example, the choice of RNP-IP without crosslinking has pros and cons).
      • The data and the methods are correctly presented for reproducibility, replicates and statistical analysis are adequate.

      Minor comments:

      • At least in the single gene validation experiments (Fig.5), a negative control (such as recombinant HuR with mutated RRMs in trans-, or ARE-less/non-HuR targetable sequence in cis, or inactive TM) would be advisable.
      • Figure 6B/C: for immunofluorescence panels, zooming on a smaller number of cells will render more visible HuR and NFB nucleocytoplasmic shuttling, given that quantification and statistics are provided by imaging software. Negative control stainings (secondary Abs only) should be included.
      • Figure 7A: in the X axis LPS+8n is indicated: is it a typo for LPD+6n or was compound TM8n indeed used?
      • In the Methods section please include protocols and materials for immunofluorescence (results shown in Fig. 6B/C).
      • There are some typos and repetition in figure legends (legend Figure S9).
      • Prior studies are referenced appropriately.

      Review Cross-commenting

      I fully agree with the Reviewer's remarks. I would add that a general concern expressed is that this manuscript in its present form has a readership issue: the first part is for chemistry/pharmacology audience, the second is biology-based. Splitting the work has been suggested; or, the Authors may decide which part is more impactful and present the other in a streamlined version.

      Significance

      This is a large study reporting progress in the development of synthetic antagonists of HuR function, which is the Authors' well-established line of research. The TM compounds are small molecules with anti-inflammatory effects with strong potential for therapeutic use due to selected inhibition of HuR-mediated upregulation of proinflammatory molecules. The physicochemical and early biological characterization done in this study will allow further testing of their efficacy and of the overall role of HuR-mediated regulation as targetable mechanism in several preclinical human disease models.

      Targeting of the RNA-binding protein HuR has been tackled as therapeutic approach in cancer, less in chronic immune and inflammatory diseases despite many common mechanisms and mediators.

      This study could be well received by researchers involved in basic science and drug development (chemistry, biochemistry/biophysics, pharmacology, computational modeling) and biologists/physician scientists interested in testing these compounds in translational research settings where HuR-driven functions can be relevant (cancer, chronic inflammation), though the chemical part would be less accessible to the latter audience.

      Reviewer's background is in preclinical human models of chronic inflammation with interest in posttranscriptional gene regulation with familiarity with RNAseq and RIPseq dataset and analysis. For the part of the manuscript regarding the synthesis and physicochemical characterization of the TN compound I requested assistance to a faculty from the chemistry department with expertise in that field, who did not request any specific clarification or addendum.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      1. General Statements

      Imbalance in gut-derived AhR ligands has been shown to be involved in inflammatory bowel disease and in neuro-inflammation. The aim of our study was to address the role of dietary AhR ligands in a context that had not been previously explored. We decided to focus on allergy because AhR has broad functions in barrier tissues homeostasis, which is directly relevant to allergy.

      2. Description of the planned revisions

      Additional experiments in response to Reviewer 2

      "The authors make a strong claim that the epidermal barrier function is not affected by AhR poor diet conditions (claim made in abstract and last paragraph of the discussion). This should be experimentally validated."

      We already performed footpad histology and observed that the stratum corneum is not affected by the diet (fig1A and figS1E). We will provide a quantitative analysis by measuring stratum corneum thickness on the images, and add this data to figure 1. To strengthen this point, we will also perform ultra-structural analysis of the epidermis in the two diet groups using electron microscopy of the skin. This will provide a deeper characterization of the epidermal structure, including cornified layers and intercellular tight junctions.

      "Injection into the footpad as a route of administration is also physiologically distinct from epicutaneous sensitization given the natural barriers are artificially breached via needle puncture. Did the authors consider epicutaneous sensitization via the skin without additional barrier disruption? Does this yield the same response?"

      We will perform skin sensitization without barrier disruption by applying papain (or vehicle) on shaved flank skin. To minimize skin abrasion, mice will be shaved the day before the application. We will analyze dendritic cells migration to the draining lymph nodes after 48h by flow cytometry, and helper T cell responses in the draining lymph nodes after 6 days by measuring cytokine secretion.

      Text edits

      Comments from Reviewer 1

      • We will add appropriate references in response to comments from reviewer 1: " in several places they cited review articles instead of original articles for key findings. Ex. For the expression of Mucin 5 and CLCA1 a review is cited." and " the role of AHR in ILC2 (PMID: 30446384) and alveolar epithelial cells (PMID: 35935956) has been documented. The authors should add these references."
      • We will modify the figures legends according to reviewer 1's suggestions: " Although the authors mentioned treatment schedule and stimulants used in the method, a short description in the figure legend will be helpful for the readers".
      • We will address other comments from reviewer 1 by modifying the text where appropriate:

      "1. In the introduction section, the authors should explain adequately why they thought that AHR signaling is important for allergy.<br /> 2. Since IL-5, IL-13 production by skin draining lymph nodes and pulmonary lymph nodes was different, is this difference due to difference in AHR expression?<br /> 3. In Fig.3, the authors showed that intra-nasal stimulation does not induce eosinophil migration or IL-5, IL-13 induction in I3C diet group. These data and the data shown in figure-2 are in contrast. The authors should discuss this discrepancy."

      Comments from Reviewer 2

      • "How to explain the difference between IL4 (no effect between the two diets/or absence/presence LCs in Fig. 4D) and IL5/IL13 (small effect in Fig 1D and 4D). "

      This is an interesting point. It has been shown that IL4 is produced in lymph nodes by T cells distinct from those producing IL5 and IL13 (https://doi.org/10.1038/ni.2182). In addition, IL4 expression is regulated at the transcriptional level by distinct mechanisms from IL5 and IL13 expression (https://doi.org/10.1016/S1074-7613(00)80073-4, https://doi.org/10.1038/ni.1966).<br /> We speculate that IL4-producing T cells are not affected by Langerhans cells presence in the lymph nodes. We will add a point in the discussion section to discuss this. - We will tune down our conclusion regarding the different effects of diet-derived and microbiota-derived AhR ligands according to the comments of the reviewer: "This part seemed far-fetched. There are many more differences between germ free and specific pathogen free mice than only the presence/absence of AhR ligands. Hence, it seemed like a very big step to compare both conditions and draw the conclusion that microbiota-derived AhR ligands activate different sets of genes. It would also make more sense if Fig. 5 would be immediately followed by Fig. 7". We also propose to move Fig6 to the supplementary data.

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

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

      Comments from Reviewer 1

      "In Fig.4, the authors show there is no difference in total number, but difference in migration, was there a difference in expression of migratory markers?"

      We assume the reviewer is referring to the number of Langerhans cells in the epidermis in steady-state, which is not different between diets (fig4A). We actually already show in supplementary figure S3E classical cell surface markers that are upregulated upon dendritic cells migration (MHC class II and CD40). We found no difference in the expression of these markers between diet groups.

      Comments from Reviewer 2

      "Fig. 1D Cytokine production<br /> In AhR poor diet the spread between the individual data points is much larger and the difference between presence/absence of dietary ligands in IL5 and IL13 seems to be based merely on a few outliers (which especially in the case of IL13 appear to be completely out of range). Most other datapoints do not seem to be highly different from the ones in the AhR rich diet.<br /> Where does this high variation come from in AhR poor diet (and what is the reason for these high outliers)? Would the data have been significantly different without the outliers? "

      Throughout the manuscript, we have represented raw data and individual data points for transparency. We observed some variability between biological replicates, not just for cytokine secretion (fig1D) but in most assays (for instance cell counts in lymph nodes in fig1C or inflammatory cell counts in fig2A and fig3A or antibody production in fig2E), yet the reviewer focuses their comments on fig1D. In the case of fig1D, we have performed Kruskal-Wallis statistical tests to account for this variation, and the difference between diet groups was statistically significant. We do not understand how we could remove the so-called ‘outliers’ without data manipulation to perform an alternative statistical test. We also disagree with the reviewer that 4 out of 11 points can be considered ‘outliers’.

      "In general, increases of all canonical T-helper cytokine responses (except for IL4) can be noted in the LN and the difference in IL10, IL17 or IFNg production between AhR poor and rich diet appears even more pronounced than the difference in IL5/IL13 (Fig. S1F). Still the authors decide to focus the entire story on the allergic response after stating that a 'lack of dietary AhR ligands amplifies allergic responses'. Why was this choice made?"

      Imbalance in gut-derived AhR ligands has been shown to be involved in inflammatory bowel disease and in neuro-inflammation. The aim of the project was to address the role of dietary AhR ligands in a context that had not been previously explored. We decided to focus on allergy because AhR has broad functions in barrier tissues homeostasis, which is directly relevant to allergy. We will better explain this point in the introduction. In the course of the study, we analyzed IL10, IL17 and IFNg production by lymph node T cells to get a complete view of helper responses, and we provided this data in supplementary information for transparency. We believe this information might be useful for other groups studying other types of skin inflammation.

      "Would the authors expect other inflammatory models via the skin (e.g. bacterial, viral infection) to confer better/worse outcomes under an AhR poor diet?"

      This is an interesting question. Unfortunately, we do not have the means to analyze bacterial or viral skin infections for lack of adequate facilities (i.e. BSL2 animal facility) or ethics approval for this kind of experiments. We believe that our work may prompt in the future other groups to analyze the impact of dietary AhR ligands in other inflammatory skin contexts.

      "At a mechanistic level, how do LC suppress the activation of T cells in the LN, and how would this impact secretion of certain cytokines but not others?"

      "it remains a bit speculative how migration of LCs to the dLNs of the skin contributes to suppressing Th2 immunity in the airways. Several hypotheses have been put forward in the discussion. What is their thought about this and how to validate experimentally?"

      This is an important question. A regulatory role for Langerhans cells has been evidenced by other studies, but the molecular mechanisms involved remain elusive. This point is discussed in the discussion part of the manuscript. We believe that deciphering the mechanism of action of Langerhans cells is beyond the scope of the present study (and is unrelated to the direct effect of the diet), and would represent an entire project in itself.

      “Fig. 3 - Why would the alteration of diet pose a confounding factor to the model? Did the authors determine that such diet-associated changes are only important at the sensitization phase? The footpad (Fig. 1) and pulmonary (Fig 2) models were performed with the altered diets throughout the entire length of the experiment. If anything, wouldn't changing the diet after sensitization also provide an additional variable here? Is it known what happens (e.g. inflammatory state, genetic changes) when a normal diet is resumed after a period of adaptation? This reviewer does not understand the reason for all-of-a-sudden changing the diet after the sensitization phase.”

      Our goal with this experiment was to address the effect of the dietary AhR ligands during the skin sensitization phase only. This is why diets are different only in this phase of the protocol. We want to emphasize that the IC3 diet and the AhR-poor diet only differ in the presence of one molecule, which is I3C. The composition of the food is otherwise exactly the same, therefore we do not believe that a change between AhR-poor and I3C would represent a confounding factor. This is different to the adaptation period when we place the mice on I3C or AhR-poor diets instead of normal chow diet (which has a completely different formulation). We will make this point clearer in the text.

      "Fig. 7 Role of TGFb<br /> At first site, it seems counterintuitive that TGFb, which is a molecule generally associated with homeostasis and dampening of inflammation, is associated here with more profound inflammation. How to reconcile? At this point the data on TGFb are merely correlative. Did the authors directly test the contribution of TGFb to LC migration? In addition, did they check whether they could restore defects in LC migration in absence of AhR ligands by blocking the formation of active TGFb? "

      We agree with the reviewer that the role of TGFb seems counter-intuitive. However, multiple studies have shown that TGFb produced by keratinocytes retains Langerhans cells in the epidermis, using a variety of experimental approaches including genetic tools (https://doi.org/10.1073/pnas.1119178109, https://doi.org/10.1038/ni.3396,

      https://doi.org/10.4049/jimmunol.1000981, https://doi.org/10.1016/j.xjidi.2021.100028). We do not have any reason to doubt the validity of these studies. Therefore, we believe that demonstrating again the role of TGFb in Langerhans cells migration is not necessary.

      Using blocking antibodies against TGFb or its receptor, as suggested by the reviewer, would most probably not allow us to address whether it restores the defect in Langerhans cells migration. Indeed, results from the literature (cited above) indicate that such blocking would increase Langerhans cells migration in both diet groups, therefore it will most likely be impossible to conclude.

      In addition, we have provided several lines of evidence that AhR activation regulates the expression of Integrin-beta8 in keratinocytes and the release of bioactive TGFb, at transcriptomic and protein levels, in both mouse and human keratinocytes (fig7). Therefore, we believe that additional experiments to support the link between AhR ligands and TGFb are not necessary within the scope of the revision.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this paper Cros et al describe how the absence of dietary ligands of AhR exacerbate cutaneous papain-induced allergy. This was only observed when papain was applied topically, but not intranasally. However, lack of dietary AhR ligands also worsened allergic airway inflammation after cutaneous sensitization. At a mechanistic level, the authors found that the absence of dietary AhR ligands hampered migration of Langerhans cells (LC) to the lymph nodes, where they are claimed to be needed to suppress T cell activation. Complementary models that lead to loss of LCs gave a similar phenotype. The authors performed RNA-sequencing on epidermal cells derived from mice that were either fed an AhR ligand rich or poor diet to define differences in transcriptome signature. They uncovered increased expression of the integrin Itgb8 in absence of AhR ligands, which is needed for production of active TGF, a factor known from literature to contribute to LC retention in the skin.

      In general, the study is well done, and the different experimental conditions are well controlled for. The experiments are built up in a logical fashion, and most of the times, the interpretation is appropriate (except for a few claims, see further). The paper reads very fluently, and the key points are interesting.

      Major comments:

      • Fig. 1D Cytokine production
      • In AhR poor diet the spread between the individual data points is much larger and the difference between presence/absence of dietary ligands in IL5 and IL13 seems to be based merely on a few outliers (which especially in the case of IL13 appear to be completely out of range). Most other datapoints do not seem to be highly different from the ones in the AhR rich diet.<br /> Where does this high variation come from in AhR poor diet (and what is the reason for these high outliers)? Would the data have been significantly different without the outliers?<br /> How to explain the difference between IL4 (no effect between the two diets/or absence/presence LCs in Fig. 4D) and IL5/IL13 (small effect in Fig 1D and 4D).
      • In general, increases of all canonical T-helper cytokine responses (except for IL4) can be noted in the LN and the difference in IL10, IL17 or IFNg production between AhR poor and rich diet appears even more pronounced than the difference in IL5/IL13 (Fig. S1F). Still the authors decide to focus the entire story on the allergic response after stating that a 'lack of dietary AhR ligands amplifies allergic responses'. Why was this choice made?<br /> Would the authors expect other inflammatory models via the skin (e.g. bacterial, viral infection) to confer better/worse outcomes under an AhR poor diet?<br /> At a mechanistic level, how do LC suppress the activation of T cells in the LN, and how would this impact secretion of certain cytokines but not others?

      Fig. 3 - Why would the alteration of diet pose a confounding factor to the model? Did the authors determine that such diet-associated changes are only important at the sensitization phase? The footpad (Fig. 1) and pulmonary (Fig 2) models were performed with the altered diets throughout the entire length of the experiment. If anything, wouldn't changing the diet after sensitization also provide an additional variable here? Is it known what happens (e.g. inflammatory state, genetic changes) when a normal diet is resumed after a period of adaptation? This reviewer does not understand the reason for all-of-a-sudden changing the diet after the sensitization phase.<br /> - Fig. 6: Microbiota-derived and diet-derived AhR ligands modulate different sets of epidermal genes.<br /> This part seemed far-fetched. There are many more differences between germ free and specific pathogen free mice than only the presence/absence of AhR ligands. Hence, it seemed like a very big step to compare both conditions and draw the conclusion that microbiota-derived AhR ligands activate different sets of genes.<br /> It would also make more sense if Fig. 5 would be immediately followed by Fig. 7<br /> - The authors make a strong claim that the epidermal barrier function is not affected by AhR poor diet conditions (claim made in abstract and last paragraph of the discussion). This should be experimentally validated. Injection into the footpad as a route of administration is also physiologically distinct from epicutaneous sensitization given the natural barriers are artificially breached via needle puncture. Did the authors consider epicutaneous sensitization via the skin without additional barrier disruption? Does this yield the same response?

      Fig. 7 Role of TGFb<br /> - At first site, it seems counterintuitive that TGFb, which is a molecule generally associated with homeostasis and dampening of inflammation, is associated here with more profound inflammation. How to reconcile? At this point the data on TGFb are merely correlative. Did the authors directly test the contribution of TGFb to LC migration? In addition, did they check whether they could restore defects in LC migration in absence of AhR ligands by blocking the formation of active TGFb?

      Finally, also other steps of the proposed model by the authors are based on literature rather than direct experiments. In this regard, it remains a bit speculative how migration of LCs to the dLNs of the skin contributes to suppressing Th2 immunity in the airways. Several hypotheses have been put forward in the discussion. What is their thought about this and how to validate experimentally?

      Significance

      The major strength of the paper (and the most interesting finding) is the explanation of why the effect of the diet is only detectable after cutaneous but not intranasal sensitisation and the causal link to the LCs present in the skin.

      The major limitations of the paper is that many parts of the proposed model are not experimentally validated but based on literature suggestions (eg the claim that TGFb would prevent LC migration to LN, that LC would suppress T cell responses in the LN, that the suppression of T cell cytokines (with very limited effects on IL5 and IL13 but no effect on IL4) would be sufficient to explain improved allergy symptoms in the lung...). It is also unclear why the authors studied allergic symptoms while effects on other T cell cytokines appeared more prominent. There are a few questions on the change in model from figure 1-2 to figure 3.

      The key findings are interesting and the paper is nice to read.<br /> The findings will be interesting to specialised audience: LC biology, allergy and Th2 immunity people<br /> Own research field, dendritic cell biology and papain-induced models of allergy

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      The manuscript is well written. The authors mostly cited appropriate papers but in several places they cited review articles instead of original articles for key findings. Ex. For the expression of Mucin 5 and CLCA1 a review is cited.

      General comments

      1. the role of AHR in ILC2 (PMID: 30446384) and alveolar epithelial cells (PMID: 35935956) has been documented. The authors should add these references.
      2. Although the authors mentioned treatment schedule and stimulants used in the method, a short description in the figure legend will be helpful for the readers.

      Specific comments

      1. In the introduction section, the authors should explain adequately why they thought that AHR signaling is important for allergy.
      2. Since IL-5, IL-13 production by skin draining lymph nodes and pulmonary lymph nodes was different, is this difference due to difference in AHR expression?
      3. In Fig.3, the authors showed that intra-nasal stimulation does not induce eosinophil migration or IL-5, IL-13 induction in I3C diet group. These data and the data shown in figure-2 are in contrast. The authors should discuss this discrepancy.
      4. In Fig.4, the authors show there is no difference in total number, but difference in migration, was there a difference in expression of migratory markers?

      Minor points

      1. TGF-β, TCR-β and cytokine names should be written consistently across the manuscript.
      2. The authors should use "β" instead of beta

      Significance

      The work is significant and will impact the field

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Dear Editor,

      Please find below our detailed responses (in black font) to the Reviewer's comments (in blue). In addition, and to the request of Reviewer #1, we added a PDF file called “Reply to the reviewers' MS data” that shows MS/MS and quantification information of representative peptides which were selected based on their (different) caspase/control abundance ratios. We thank the reviewers for their time and valuable comments.

      NOTE: our original reply (that was uploaded to ReviewCommons) includes several tables and graphs that were not incorporated into our reply shown below

      Reviewer #1

      Page 4 - In contrast to the hindrance of N-terminal amine ionization by Nt-acetyl groups concluded by the authors, previous studies reported an improved MS-scoring if α-amino-acetylated (tryptic) peptides by the higher numbers of b and y fragment ions observed as compared to α-amino-free (tryptic) peptides (e.g. (Staes et al., 2008)). It is rather the lack of any N-/C-terminal charged residue in case of Lys-N type N-termini which makes LATE less suitable for studying N-terminal protein acetylation.

      We thank the reviewer for this comment. In the HYTANE and LATE workflows, only peptides with modified N-termini (by dimethylation or acetylation) are observed after negative selection, hence we argue that the important comparison here is between Nt-acetylated peptides and Nt-dimethylated peptides with (as in HYTANE) or without basic residue (as in LATE). While we are aware of the study by Staes et al 2008 (PMID: 18318009), we do not believe it contradicts our claim as it discusses the difference between Nt-acetylated peptides and peptides with free N-termini.

      As we indicated in the manuscript (page 5 in the last sentence of 1st paragraph), we observed less overall peptide identifications in LATE, which was expected due to the lack of basic C-term residue. The reduction of identification was more pronounced for Nt-acetylated peptides. However, this still does not exclude LATE as a useful tool for the identification of such peptides.

      Of note, the overall fragmentation coverage we obtained by LATE and HYTANE for Nt-acetylated and Nt-dimethylated peptides was similar. See the figure below.

      Hence, following Cho et al 2016 (PMID: 26889926), we suggest that the difference in ionization of Nt-dimethylated peptides vs Nt-acetylated peptides is the more dominant factor in peptide identifications.

      Figure 1:relative Ion coverage for modified peptides in LATE and HYTANE

      Page 4 - Besides indication the retained N-termini with high relative caspase-3/control abundance ratio's as putative caspase-3 proteolytic products, also indicate that unique peptides were retained, as many such singletons were reported in previous (caspase-focussed) degradomics studies making use of differential proteomics (e.g. (Van Damme et al., 2005)). Therefore the cut-off ratio of 2 rather seems unsubstantiated, unless the cellular proteomes of so-called control cells were affected by caspase activation. As such, showing some representative MS-spectra of neo-N-termini would be informative.

      We thank the reviewer for this comment. We agree that caspase-3 cleavage generates many singletons. This is indeed what we observed in the in vitro experiment as shown in Figure 2B by the long straight lines at Log2(caspase-3/control) >10. We also add here histograms of the obtained ratios that we hope will make this clearer. We chose a cut-off of 2 due to the basal activity of proteases (including caspase-3) as we did not add caspase-3 inhibitors to the cell lysate. In addition, peptides derived from the putative caspase-3 cleavages in the in vitro experiment were required to be identified only in the caspase-3-treated samples (i.e. to appear only with the heavy labeling). Minor changes to Figure 3 legend have been introduced accordingly. As can be seen in the table below, with a cut-off ratio of 2 (Log2=1) and selection of cleavage sites after D or E we identified >98% of the cleavage sites that were identified only in the caspase-3 treated samples (column text in blue). This rate did not change when the cut-off was set to 8 (Log2=3). Therefore, we have chosen to maintain our selection criteria and cut-off ratio as used before for both experiments.

      Figure 2: Histograms of Log2(Caspase/control) ratios indicating the large number of singleton peptides (marked with arrows)

      Table 1: In vitro experiment selection ratio

      Method

      Cutoff

      Time

      Sites

      Sites identified only caspase-3 treated samples

      % of caspase-treated only sites (singleton)

      Sites D/E with light

      Sites after D/E no light

      % of singleton

      LATE

      Log2=1

      18H

      936

      906

      96.8%

      798

      786

      98.5%

      LATE

      Log2=2

      18H

      884

      866

      98.0%

      767

      759

      99.0%

      LATE

      Log2=3

      18H

      819

      810

      98.9%

      722

      716

      99.2%

      HYTANE

      Log2=1

      18H

      1186

      1159

      97.7%

      1037

      1032

      99.5%

      HYTANE

      Log2=2

      18H

      1128

      1110

      98.4%

      998

      993

      99.5%

      HYTANE

      Log2=3

      18H

      1035

      1025

      99.0%

      924

      919

      99.5%

      LATE

      Log2=1

      6H

      755

      732

      97.0%

      656

      645

      98.3%

      LATE

      Log2=2

      6H

      711

      700

      98.5%

      630

      623

      98.9%

      LATE

      Log2=3

      6H

      671

      666

      99.3%

      601

      597

      99.3%

      HYTANE

      Log2=1

      6H

      1028

      988

      96.1%

      899

      890

      99.0%

      HYTANE

      Log2=2

      6H

      955

      931

      97.5%

      851

      844

      99.2%

      HYTANE

      Log2=3

      6H

      882

      871

      98.8%

      795

      791

      99.5%

      LATE

      Log2=1

      1H

      445

      423

      95.1%

      380

      372

      97.9%

      LATE

      Log2=2

      1H

      411

      402

      97.8%

      361

      355

      98.3%

      LATE

      Log2=3

      1H

      386

      380

      98.4%

      344

      338

      98.3%

      HYTANE

      Log2=1

      1H

      593

      559

      94.3%

      513

      506

      98.6%

      HYTANE

      Log2=2

      1H

      544

      532

      97.8%

      488

      482

      98.8%

      HYTANE

      Log2=3

      1H

      508

      498

      98.0%

      461

      455

      98.7%

      In the cell-based experiments of caspase-3, we induced apoptosis on both cell types (over-expressing caspase-3 and the control). Therefore, in this case, as the reviewer has also mentioned, a cut-off of 2 is appropriate because the control cells are also affected by caspase activation. Following the reviewer’s request we have added (in a separate PDF file) several representative MS/MS spectra of neo-N-term peptides and their corresponding quantification data.

      Page 4 - replace 'without labelling of lysine residues (epsilon-amines)' to 'without notable labelling of lysine residues (epsilon-amines)', as residual labelling of lysine side-chains was observed. Also in case of the latter, do note that reduced MS-ionization potential might impact labelling efficiency calculation, and chromatographic detection of labelling efficiency should be considered to conclusify this finding.

      We thank the reviewer for this comment. We have changed the sentence as requested (Page 4 marked in red). Regarding the labeling efficiency calculations, it is true that ionization potential might affect them. We used a common way to test this aspect (see e.g. Hurtado Silva et al 2019 (PMID: 30934878)) and we are not aware of any reduction in ionization potential following lysine dimethylation. Although we did not study this aspect thoroughly, we frequently observe the opposite: that dimethylation improves MS detections.

      Page 6 - The experimental setup comparing caspase-3 overexpressing and ABT-199 induced versus ABT-199 induced cells will be highly biased as it will not be able to detect efficient caspase-3 cleavages (Plasman et al., 2011), as such cleavage events are complete and thus do not require any additional overexpressed capase-3. I see this as an important flaw and the authors should demonstrate that the list also includes efficient caspase-3 cleavages.

      We thank the reviewer for highlighting this important aspect. We agree that with our setup, we can miss some efficient cleavages of caspases-3. We acknowledged this caveat in the original text (page 6), but chose to perform our experiments this way in order to highlight cleavages that are affected by caspase-3 expression. To address the reviewer’s comment we have added new experiment and data on caspase cleavages that occur following ABT-199 treatment in HCT116 cells without overexpression of caspase-3. The focus of this experiment was on the relatively short time points following the ABT-199 treatment when no cell death is observed based on XTT assay (see Supplement Figure 6B). This experiment was used to prove that neo-Nt-acetylation of NACA is an early event in apoptosis (Figure 5 E-F page 12). We also used this experiment as an indication of the appearance of efficient cleavages. As can be seen from Supplement Table S10, if we consider all 3 time points of the ABT-199 treatment, we quantified 106 cleavages with free neo-Nt that were cleavages after D and were identified only in the ABT-treated samples. We refer to such cleavages, which appeared prior to noticeable cell death, as "efficient cleavages". Out of these efficient cleavages, 82 were also identified and quantified in the cell-based experiment with overexpression of caspase-3. Twenty efficient cleavages show a high ratio (≥2) in both experiments. Fifty six efficient cleavages had a high ratio in the new experiment and a ratio below 2 in the cell-based experiment with overexpression of caspase-3. This supports our original claim regarding efficient cleavages and addresses the reviewer’s concern regarding our ability to identify efficient caspase-3 cleavages with the experimental setup of HCT116 cells overexpressing caspase-3.

      Page 12 - The setup doesn't permit ORF N-terminal stability per se, rather the cleavage susceptibly of N-termini holding (a) putative caspase-3 cleavage site(s). Please adjust accordingly. Again since the setup might have missed efficient cleavages, the assessment might be biased.

      Thanks for the comment. As requested, the word “stability” has been deleted. As discussed above, we demonstrate that our setup does allow the identification of efficient cleavages and hence our basis for believing that the assessment is not biased. Please also refer to our reply to the next comment.

      The claim that Nt-acetylation is protective for caspase-3 cleavage should be validated by monitoring cleavage efficiency of an Nt-acetylated versus an Nt-free variant (e.g. by introducing a Pro residue at AA position 2, or comparing cleavage efficiencies in corresponding NAT knockdown versus control cells) of an identified caspase substrate (i.e. a substrate holding a caspase-3 cleavage site in its N-terminal sequence) versus its Nt-free counterpart

      Thanks for raising this point. The reviewer's suggestions have some caveats: a mutation at a protein’s N-terminus in order to generate an Nt-free variant can alter its stability or function and NAT knockdown might have a profound biological impact on the cells. Therefore we chose a different way to study this aspect by selecting from our data ORF N-terminal peptides that were identified with both free N-termini and acetylated N-termini (i.e. the same peptide was identified in some PSMs as acetylated and in other as dimethylated). We managed to find 136 ORF N-terminal peptides that were quantified in both forms, and out of these, 122 contained Asp or Glu residues (the putative caspase cleavage motifs). We added the comparison of the abundance ratios of these peptides in Figure 4C (see also below). It shows a remarkable difference between the groups when the Nt-acetylated peptides ratios did not change as a result of caspase-3 overexpression while the peptides with free Nt were identified mostly in the control cells (negative Log2(caspase-3/control)). A comparison of the 14 ORF Nt-peptides that do not have Glu or Asp in their sequence shows no difference (see below).

      Figure 3: Abundance ratio distributions of the ORF Nt peptides identified with both Nt-acetylated and free Nt in HCT116 cells overexpressing caspase-3 and in the control. A. Comparison of peptides that contain putative caspase cleavage in their sequence (D or E) B. comparison of peptides without putative caspase cleavage

      These results provide additional support for the notion of the protective or shielding effect of Nt-acetylation against caspase-3 cleavage.

      Page 12 - Since post-translational Nt-acetylation of neo-N-termini could be reproduced in vitro in the non-dialyzed sample, enzymatic over chemical Nt-acetylation should be demonstrated (e.g. by the use of a (bisubstrate) NAT inhibitor).

      We think this is an interesting idea for future work. Yet, in our opinion, the fact that only very few neo-Nt-acetylated peptides were affected in vitro and that a similar trend of few selected neo-Nt-acetylation targets was shown in the cell-based experiments indicates that this process is enzymatic and not chemical in nature.

      Other concerns:

      Abstract - The abstracts holds complex/incorrect sentence constructions (e.g. simply indicate 'Protein N-termini', '... undergo ... processing by proteases' (currently: 'not be processed by proteases').

      Thanks for pointing this out. We have changed the abstract accordingly.

      Abstract - 'To expand the coverage of the N-terminome' only applies when this is used in conjunction with other negative enrichment strategies as by itself, LATE doesn't intrinsically provide a better coverage of the N-terminome (this is also noted at page 2).

      We thank the reviewer for pointing this out. We have changed the abstract accordingly.

      Change 'that cannot be identified by other methods' to 'that cannot be identified by other negative selection methods'

      Thanks for pointing this out. We believe that our description here is appropriate as we explicitly state “some of which cannot be identified by other methods”.

      Page 1 - Suggestion to change 'Proteases are typically described as degradative enzymes' to 'Proteases used to be described as degradative enzymes'

      Changed as suggested.

      Page 1 - Not really correct how written; 'N-terminomics methods highlight the N-terminal fragment of every protein (N-terminome)'

      Changed as suggested.

      Page 2 - Positive selection techniques .... Enrichment of unblocked (or Nt-free) N-termini

      We are not sure what the reviewer had in mind here but have added the text in the brackets

      Page 2 - Besides altering charge, Nt-acetylation also alters hydrophobicity ...

      Changed as suggested.

      Page 2 - remove 'to better chart'

      Changed as suggested.

      Page 2 etc. - Do note that caspase-3 can potentially activate downstream caspases in vitro

      Following this comment, we have added a sentence on Page 5 with this reservation

      Page 3 - functional crosstalk between proteolysis and neo-Nt-acetylation has already been demonstrated in the case of co-translational acting methionine aminopeptidases and chloroplast N-terminal acetyltransferases. Adjust accordingly.

      We thank the reviewer for highlighting this aspect, although we used the term “neo-Nt-acetylation” which we used to mark that this is not the common (co-translational) acetylation. To assure that this is more clear we have added the words “post-translational” to better define the novelty of our findings.

      Page 3 - when discussing the identification of ORF N-termini, note that some of the strategies of which note when used to enrich for in vivo blocked N-termini, can also be used without blocking/labelling of Lys residues, and thus trypsin will also result in Lys-ending peptides. This is important to consider in this context.

      Following the reviewer's remark we have changed the sentence so it now states: “Many of these N-terminomics methods……”

      Page 3 - remove the following sentence part; '... or run individually which can be useful for quantifying naturally modified N-termini.', since also a differential/labelled proteomics setup enables such assessment. Related to this, the authors should comment on the observation that much fewer (i.e. less than 40%) Nt-acetylated N-termini were identified by LATE as compared to HYTANE. How is this reflected in the number of PSMs? Probably these difference are further intensified when considering PSMs.

      We have changed the sentence as suggested.

      Regarding the reduction of Nt-acetylation, we thank the reviewer for this question as it led us to find typos in the numbers in Figure 1E which are now corrected. These typos did not change the overall observation that with LATE we identify fewer Nt-acetylated peptides than Nt-free (dimethylated) peptides. As the reviewer anticipated (see below), the reduction in LATE-based “contribution” to the identification of Nt-acetylated peptides as opposed to the identification of dimethylated peptides, is pronounced when considering PSMs but this is not much different than the peptide-based data. Therefore, we prefer to keep the current presentation of Figure 1E.

      Figure 4: Comparison of HACAT cells N-terminal peptides identification with LATE and HYTANE when considering peptide sequences and PSMs. Peptides identified with both methods are in green and those that are unique to one method are in blue. Shared peptides were determined based on the sequence of the first 7 amino acids of the identified peptides. A. comparison for peptides with dimethylated N-terminal (free Nt) B. comparison for Nt-acetylated peptides.

      Page 6 - Informative to indicate how many of the in silico predicted putative DEVD P4-P1 cleavages were actually present in the list of 2049 putative cleavages identified.

      In our dataset, we identified 17 cleavages after DEVD motif. 11 were identified only with HYTANE, 3 were identified by both methods, and 3 more were identified only with LATE. Of note, it seems that in large-scale proteomic studies of apoptosis, the number of caspase cleavages after DEVD motif is quite low. For example, in the CASBAH database (PMID: 17273173__) __there are 10 reports of such cleavage out of 391 reported sites, and in DegraBase (PMID: 23264352) that combined many different apoptotic experiments there are 64 reported DEVD sites out of a total of 6896 P1-Asp sites.

      Page 6 - Unclear if any of the of 2049 putative cleavages, included non-canonical P1 cleavages besides the P1 Asp and Glu cleavages identified.

      These are 2049 putative cleavage sites with P1 Asp or Glu. We have changed the text to make it clearer.

      Page 6 - Were the 'regular' cells mock transfected?

      No. The control cells used for the cell-based experiments were the non-transfected cells from the same culture of HCT166. We chose this option to guarantee that exactly the same cells that were grown in the same dish went through the same FACS sorting as a control.

      Page 6 -Important to note that an ORF can have multiple N-termini besides neo-N-termini (e.g. in the case of alternative translation initiation)

      Thanks for the great point. We have added an indication if the neo-N-termini site has been reported as an alternative translation initiation site to all of the results of the cell-based experiments (Supplementary Tables S4, S5, S6, S9). We also changed the Figures and text accordingly. Our analysis of reported/unreported neo-N-temini is based on the TopFind database which includes information about alternative translation initiation sites from TISdb. Of note, since our focus is on caspase cleavages and we further select putative cleavages based on D/E in P1 and fold change, out of 973 peptides that we reported as putative caspase cleavage (Table S6) only one is in the vicinity of an alternative initiation site.

      Page 6 - The authors should be more careful with generalization when comparing LATE and HYTANE (and other degradomics approaches) as in this study LATE was only applied for the identification of caspase-3 neo-N-termini, which by its extended substrate specificity might hold specific features enabling the preferred detection by one technique over the other. Also note that as compared to less recent studies, evidently the MS instrument used is a key factor in the increase in cleavages reported in the current study.

      It is conceivable that caspase cleavage may differ from other proteases and thus theoretically work better with LATE, but we fail to see why this would also be the case for other N-terminomics method (like TAILS, Subtiligase, CoFRADIC, ChaFRADIC etc). We showed that LATE provides additional ORF Nt peptides identifications and demonstrated its effectiveness in E. coli (Supplement Figure S2) also, which has a proteome with a different amino acid composition to the human proteome. Furthermore, using LATE in the cell-based experiment led to the identification of many neo-Nt-peptides that do not match caspase cleavage patterns (as indicated for both HYATNE and LATE in Figures 3E and 3F). We reviewed the text again, and believe that we have used a fair description of the results especially when we compared them to previous studies.

      Page 9 - The authors should provide some info/supporting statistics in the text regarding the new putative substrates showing GO-enrichments (compared to which control?) similar to previously reported caspase-3 substrates.

      The results of the GO enrichment analysis are presented in Fig. S8 and details about how the test was performed are provided in the Materials & Methods. In the revised version, we are including the numerical data that include results of the statistical tests per GO term as Table S12. The enrichment analysis was performed with respect to the whole human proteome.

      Page 11 - Indicate that the 11 neo-N-terminal peptides of which note are the neo-Nt-peptides matching (signal peptide) cleavages indicated in the Uniprot database. Were any corresponding di-methylated neo-N-termini of these cleavages identified? In case of the 'other' proteolytic cleavages of which note, refer to these as not-annotated in UniProt.

      We thank the reviewer for pointing this out. We have added an indication that this analysis is based on UniProt annotations. Yes, all of the reported 11 neo-Nt-Acet peptides shown in Figure 4 were also found as neo-Nt-DiMet peptides.

      Page 11 - post-translational Nt-acetylation is abundant in plant and the responsible NAT has been identified, please reference these studies as well.

      We thank the reviewer for pointing this out regarding page 11. A relevant reference has been added in Page 11. In the discussion, we already referenced Nt-acetylation in plants in the discussion as well (see page 14).

      Page 12 - Define 'undoubtedly dependent on caspase-3 cleavage'

      We thank the reviewer for pointing this out. The word ‘undoubtedly’ has been deleted.

      Page 14 - The NAA30 discussion is not really relevant for the discussion of the post-translational Nt-acetylation of mitochondrial neo-N-termini.

      We thank the reviewer for pointing this out. This sentence has been deleted.

      Viewing the harsh in vitro caspase-3 cleavage condition used, namely 1 µg caspase 3 over 20 µg protein, the P1 specificities of all identified neo-N-termini should clearly be shown.

      The P1 specificities of all neo-N-termini found in the in vitro experiment are listed in the supplementary tables S2 and S3. For the reviewer’s convenience, we are providing the table with the P1 specificities below:

      Since acetylation of serine and threonine residues are reported forms of post-translational modification, and many so-called past-translational Nt-acetylated neo-N-termini harbour such AA residues in their N-terminal sequence, b-ion coverage for these neo-N-termini should be provided/inspected.

      We are not sure that we understand this comment. O-Acetylation of amino acids refers to their side chain. Since we are using Di-methylation labeling in both HYTANE and LATE, if we have a peptide with O-acetylated Ser or Thr at its first position, it is possible to distinguish it from the same peptide with Nt-acetylation by MS1 as illustrated in the following table for a hypothetical peptide SAAANPELKR (mass is MH+1)

      Regardless we include in the manuscript MS/MS spectra of NACA Neo-Nt-acetylated peptide by HYTANE and LATE

      Reviewer #2

      Major suggestions:

      • The LATE method relies on digestion with LysN. Can the authors comment on the digestion efficiency of the samples where the LATE workflow was applied?

      The LysN digestion details that we used were based on vendor (Promega) instructions combined with details from the Nature Protocol paper by Giansanti et al 2016 (PMID: 27123950__)__ that describes optimized digestion protocol for LysN. We tested LysN efficiency in terms of the identification of missed cleavage and found that it performed very well with a missed-cleavage rate of

      • The authors state that the number of peptides with acetylated N-termini was lower compared with HYTANE. Yet, the Nt-acetylation can occur co-translationally in approximately 85% of human proteins.

      Did the authors consider optimizing the method (e.g. by fractionating the sample) for better identification of such peptides?

      We thank the reviewer for this important comment. We are certain that it is possible to improve the output of LATE by fractionation and/or optimization by changes to the LC gradient as it is well established for most, if not all, bottom-up proteomics methods. In this work, we concentrated more on the proof of concept of the methodology and hence chose to work without fractionation. We performed one attempt to optimize the LC gradient but found that the results were not significantly different, and we thus used the same LC-MS methods that have been optimized for trypsin.

      Regarding the reduced identification of Nt-acetylated peptides, as we state in the manuscript following Cho et al 2016 (PMID: 26889926), we believe that this is mainly due to the reduced ionization efficiency of Nt-acetylated peptides compared to Nt-dimethylated peptides which is more pronounced when a C-terminal positive charge is missing (due to LysN digestion).

      Also, were the results of the study compared with searches done using other proteomic pipelines (e.g. FragPipe)?

      Unfortunately, when we started this project, MS-Fragger did not support LysN as the digesting enzyme. At the time TPP also provided better visualization and quantification support than FragPipe. Recently, we found that MSFragger is faster while providing similar identifications but we are not convinced of the quantification output via FragPipe. In addition, we performed comparisons of Comet to X!Tandem and while the searches took longer than with Comet, the total number of IDs did not improve significantly.

      Can the authors provide details on the settings used for searches done in COMET, especially for the samples treated with LysN?

      The settings are provided in Table S10 in the supplementary information (Page 14 of the PDF file).

      "Fractions containing relatively pure caspase-3 were pooled together and dialyzed against 20 mM HEPES 7.5, and 80 mM NaCl. Aliquots of the protein were stored at -80{degree sign}C"

      o What exactly is meant by 'relatively pure'?

      We apologize for the inaccurate description. The relevant text has been updated (Page 17) and now explains that this was based on Coommasse stain SDS-PAGE.

      Minor suggestions:

      • Please check the link for the Github as this reviewer could not open it.

      We thank the reviewer for pointing this out. We corrected the link. In any case, the relevant scripts can be found here: https://github.com/OKLAB2016

      • Please correct the spelling.

      The manuscript was proofread.

      Comments regarding figures:

      • Figure 2:

      o All figures comparing LATE and HYTANE utilize color green for LATE. Yet, in figure 2G, HYTANE is depicted in green-like color. Please consider staying consistent with the color scheme.

      We thank the reviewer for this comment. Done as suggested.

      Reviewer #2 (Significance (Required)):

      Significance:

      • The LATE method provides an excellent way to study proteases in vitro or in cell-based experiments. It enables deep investigation of N-terminome based on a simple and cost-effective workflow that utilizes digestion with LysN followed by chemical derivatization of α-amines. This approach allows for the identification of N-terminal peptides that may escape detection by other N-terminomics methods. With LATE, proteases' cleavage sites that might not so far be reporter due to technical limitations, can be studied and characterized. Hence, LATE is a useful addition to the N-terminomic toolbox.

      We thank the reviewer for the positive comments and general assessment of LATE.

      Reviewer #3

      In this manuscript, Hanna et al. report LATE, an N terminomics method similar to N-TAILS and HYTANE, with modifications that enhance or change coverages of the N-terminal proteome in proteomics datasets. LATE relies on selective N-terminal modification of protease-treated, LysN digested samples, enabling internal peptides to be depleted based on the presence of the unblocked lysine epsilon amine. Using LATE in comparison with HYTANE, the authors identified a large number of both known and unknown caspase-3 cleavage sites, both in vitro and in vivo. Because LATE enables identification of both proteolytic neo-N termini and natively blocked N termini such as those that are acetylated, the authors were able to discover a number of post-translationally acetylated proteolytic neo-N termini. This finding points to potential functional cross talk between apoptotic proteolysis and Nt-acetylation. Overall, this is a very nice manuscript that adds a valuable new tool to the N-terminal proteomics toolbox. However, the manuscript could be improved by addressing the following questions and comments.

      We thank the reviewer for this assessment.

      1. One of the benchmark points used to describe the need for a new technology such as LATE is the idea that there are 134 putative caspase-3 substrates in the human proteome, of which only about half can be identified based on ArgC cleavages. However, the 134 substrates seem to include only those that have the exact canonical DEVD motif. Many more substrates than this are already known for caspase-3. For example, >900 caspase-3 substrates were identified by Araya et al. alone. It might make more sense to apply a position-specific scoring matric to the human proteome to predict a maximum number of possible caspase-2 cleavage sites and how many would be expected to be identified using other technologies. Otherwise, please provide a rationale for why these 134 putative caspase-3 sites are representative.

      The reviewer is correct. Indeed, most of the identified caspase-3 cleavage are not exact matches to the DEVD motif. We used the DEVD as an example to illustrate the added value of using lysine-based digestion together with ArgC. We obtained a similar trend with some variations when we tested the feasibility of the identification of the human ORF Nt-peptides, E. coli ORF Nt-peptides and more. We are quite confident that any prediction will show a relatively similar distribution. To demonstrate this, we show here the relative contribution of each method for the identification of any peptide that begins after Asp in the human proteome.

      While the distributions are not identical, they are very similar, and the specific additions from LATE (LysN) are between 20% to 22% out of the total and it can help to expand the coverage by 42% to 45%.

      It is definitely plausible and have been previously demonstrated that selective N-terminal demethylation can be achieved under the right reaction conditions, and I do not doubt that it has been achieved here. However, I do not understand how the authors were able to conclude that alpha-amines are blocked with 95% efficiency and lysines are blocked at

      This is a very good point. The reviewer is correct and indeed we don’t have a way to establish if the dimethylation is on the side chain amine of lysine or on its N-terminal amine. A partial support for our claim is from labeling experiments that we (and others) conducted with tryptic and LysC peptides that clearly demonstrate that under the specified labeling conditions, 95% of the N-terminal amines are labeled and not the lysine side chain amines. However, at the end of the day, this does not change the outcome of LATE.

      Related to the above comment, Table S10 seems to indicate that MS/MS data from LATE were searched with dimethylation as a fixed modification at the N terminus. Were LATE samples searched with different parameters to generate Figure 1C? Are the dimethylated Ks identified mostly from missed cleavages and therefore not at the N terminus?

      We thank the reviewer for pointing this out. The search parameters used for the generation of Figure 1C have been added to Table S10. The reviewer is correct, the few dimethylated Ks identified in the search used for Figure 1C are mostly from missed cleavages.

      For both the in vitro and in vivo experiments, how many of the new caspase-3 cleavage sites occurred in proteins that were not previously known to be caspase substrates?

      In the in vitro experiments, we identified cleavages of 372 proteins that were not reported as caspase-3 substrates based on the databases we used as references. A line specifying this number was also added to text on page 7. In the cell-based experiment, we identified putative caspase-3 cleavages of 67 proteins that were not reported so far as caspase-3 substrates. This information has been added to the main text on page 10. We have added columns indicating the known/unreported protein substrates to Tables S2, S3, S4, S5, and S6.

      For the experiment in cells, can the authors explain the rationale for comparing cells in which apoptosis is induced with ABT-199 to ABT-199-treated cells with caspase-3 overexpression? What is the advantage over comparing ABT-199 treated cells to untreated cells

      Great question. An N-terminomics study of “common” apoptosis would lead mainly to the identification of effector caspases (caspase-3 and -7) substrates. Our aim was to focus mostly on the caspase-3 cleavages that occur in the cell during apoptosis. In choosing this gain-of-function approach we were motivated by the idea that it couldprovide new insights that would otherwise go undetected when using knockout or other loss-of-function approaches. The advantage of this system over comparing ABT-199 treated to non-treated cells (which we have now added as well) is that it can enhance the identification of caspase-3 specific cleavages.

      Can the authors discuss the timescale of cell death in ABT-199 treated cells vs. ABT-199 treated caspase-3-overexpressing cells. Ideally, data showing cell viability over time (e.g. Cell Titer Glo or MTT assays) would be presented, but if the authors could at least describe whether apoptosis is accelerated in the caspase-3 overexpressing cells, it would be helpful.

      Great suggestion. Following the reviewer’s suggestion we have characterized the effect of caspase-3 overexpression of the cells by XTT assay, and indeed caspase-3 overexpressing cells do show accelerated cell-death in response to ABT199 compared to non-transfected cells. These results are now presented as Supplement Figure S6B and are mentioned in the results section.

      The authors say that in their experimental design, they expect to see no difference between ABT-199 only and ABT-199/caspase-3 overexpression for substrates that are cleaved efficiently by endogenous caspases. If the new caspase-3 substrates are not cleaved efficiently by endogenous caspase-3, this seems to call into question their physiological relevance. Can the authors explain more thoroughly how these new substrates fit into the apoptotic program?

      We thank the reviewer for raising this issue. We are aware that our original cell-based experimental design may have some limitations, yet we chose this gain-of-function setup in order to identify caspase-3 substrates in a cell-based system. We believe that this setup does allow identification of substrates that are efficiently cleaved by endogenous caspase-3, such as cleavage and acetylation of NACA at Ser34 (and neo-Nt-acetylation after caspase-3 cleavage in general). To study the physiological relevance of the neo-Nt-acetylation, we have added to the revised manuscript a time-course N-terminomics characterization of early apoptosis events conducted in HCT116 cells (without caspase-3 overexpression). The results of these experiments are now shown in Figure 5C and also in the Supplementary Table

      The authors convincingly show that cleaved NACA is a neo-substrate for Nt-acetylation, suggesting functional crosstalk between proteolysis and acetylation. However, it is not clear if this acetylation event has a functional consequence, so it seems inaccurate to say at the top of page 3 that "This is the first demonstration of functional crosstalk between neo-Nt-acetylation and proteolytic pathways."

      The author is correct. We changed the text accordingly.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this manuscript, Hanna et al. report LATE, an N terminomics method similar to N-TAILS and HYTANE, with modifications that enhance or change coverages of the N-terminal proteome in proteomics datasets. LATE relies on selective N-terminal modification of protease-treated, LysN digested samples, enabling internal peptides to be depleted based on the presence of the unblocked lysine epsilon amine. Using LATE in comparison with HYTANE, the authors identified a large number of both known and unknown caspase-3 cleavage sites, both in vitro and in vivo. Because LATE enables identification of both proteolytic neo-N termini and natively blocked N termini such as those that are acetylated, the authors were able to discover a number of post-translationally acetylated proteolytic neo-N termini. This finding points to potential functional cross talk between apoptotic proteolysis and Nt-acetylation. Overall, this is a very nice manuscript that adds a valuable new tool to the N-terminal proteomics toolbox. However, the manuscript could be improved by addressing the following questions and comments.

      1. One of the benchmark points used to describe the need for a new technology such as LATE is the idea that there are 134 putative caspase-3 substrates in the human proteome, of which only about half can be identified based on ArgC cleavages. However, the 134 substrates seem to include only those that have the exact canonical DEVD motif. Many more substrates than this are already known for caspase-3. For example, >900 caspase-3 substrates were identified by Araya et al. alone. It might make more sense to apply a position-specific scoring matric to the human proteome to predict a maximum number of possible caspase-2 cleavage sites and how many would be expected to be identified using other technologies. Otherwise, please provide a rationale for why these 134 putative caspase-3 sites are representative.
      2. It is definitely plausible and have been previously demonstrated that selective N-terminal demethylation can be achieved under the right reaction conditions, and I do not doubt that it has been achieved here. However, I do not understand how the authors were able to conclude that alpha-amines are blocked with 95% efficiency and lysines are blocked at <5%. This claim seems to be based on PSMs for each type of modification. However, in a LysN digested sample, we would expect the vast majority of peptides to begin with K and the vast majority of Ks to be found at the N terminus of a peptide. In this situation, how is it possible to distinguish whether demethylation has occurred on the alpha-amine or the epsilon-amine? With N-terminal K, all of the MS2 fragments containing the N-terminal a-amine would also contain the lysine epsilon-amine. The m/z values for the y-ions, b-ions, and a-ions containing this residue would be the same. I may be misunderstanding, so it would be helpful if the authors could explain how they are able to distinguish these.
      3. Related to the above comment, Table S10 seems to indicate that MS/MS data from LATE were searched with dimethylation as a fixed modification at the N terminus. Were LATE samples searched with different parameters to generate Figure 1C? Are the dimethylated Ks identified mostly from missed cleavages and therefore not at the N terminus?
      4. For both the in vitro and in vivo experiments, how many of the new caspase-3 cleavage sites occurred in proteins that were not previously known to be caspase substrates?
      5. For the experiment in cells, can the authors explain the rationale for comparing cells in which apoptosis is induced with ABT-199 to ABT-199-treated cells with caspase-3 overexpression? What is the advantage over comparing ABT-199 treated cells to untreated cells
      6. Can the authors discuss the timescale of cell death in ABT-199 treated cells vs. ABT-199 treated caspase-3-overexpressing cells. Ideally, data showing cell viability over time (e.g. Cell Titer Glo or MTT assays) would be presented, but if the authors could at least describe whether apoptosis is accelerated in the caspase-3 overexpressing cells, it would be helpful.
      7. The authors say that in their experimental design, they expect to see no difference between ABT-199 only and ABT-199/caspase-3 overexpression for substrates that are cleaved efficiently by endogenous caspases. If the new caspase-3 substrates are not cleaved efficiently by endogenous caspase-3, this seems to call into question their physiological relevance. Can the authors explain more thoroughly how these new substrates fit into the apoptotic program?
      8. The authors convincingly show that cleaved NACA is a neo-substrate for Nt-acetylation, suggesting functional crosstalk between proteolysis and acetylation. However, it is not clear if this acetylation event has a functional consequence, so it seems inaccurate to say at the top of page 3 that "This is the first demonstration of functional crosstalk between neo-Nt-acetylation and proteolytic pathways."

      Significance

      See above.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      The authors present a novel N-terminal enrichment method named LATE (LysN Amino Terminal Enrichment) that utilizes chemical derivatization of α-amines that enables characterization of the N-terminome. Using LATE as well as the already established HYTANE method, Hanna et al conducted a study of caspase-3 mediated proteolysis both in vitro and in cell-based apoptosis experiments, which led to the discovery of new potential caspase-3 cleavages. The results are well presented and nicely highlight that LATE is an efficient and inexpensive method that can be used to identify cleavage sites that cannot be found by other N-terminomics workflows.

      Major suggestions:

      • The LATE method relies on digestion with LysN. Can the authors comment on the digestion efficiency of the samples where the LATE workflow was applied?
      • The authors state that the number of peptides with acetylated N-termini was lower compared with HYTANE. Yet, the Nt-acetylation can occur co-translationally in approximately 85% of human proteins. Did the authors consider optimizing the method (e.g. by fractionating the sample) for better identification of such peptides? Also, were the results of the study compared with searches done using other proteomic pipelines (e.g. FragPipe)?
      • Can the authors provide details on the settings used for searches done in COMET, especially for the samples treated with LysN?
      • "Fractions containing relatively pure caspase-3 were pooled together and dialyzed against 20 mM HEPES 7.5, and 80 mM NaCl. Aliquots of the protein were stored at -80{degree sign}C"
        • What exactly is meant by 'relatively pure'?

      Minor suggestions:

      • Please check the link for the Github as this reviewer could not open it.
      • Please correct the spelling. Comments regarding figures:
      • Figure 2:
        • All figures comparing LATE and HYTANE utilize color green for LATE. Yet, in figure 2G, HYTANE is depicted in green-like color. Please consider staying consistent with the color scheme.

      Significance

      • The LATE method provides an excellent way to study proteases in vitro or in cell-based experiments. It enables deep investigation of N-terminome based on a simple and cost-effective workflow that utilizes digestion with LysN followed by chemical derivatization of α-amines. This approach allows for the identification of N-terminal peptides that may escape detection by other N-terminomics methods. With LATE, proteases' cleavage sites that might not so far be reporter due to technical limitations, can be studied and characterized. Hence, LATE is a useful addition to the N-terminomic toolbox.
    4. Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Manuscript Reference: RC-2022-01676

      TITLE: In-depth characterization of apoptosis N-terminome reveals a link between caspase-3 cleavage and post-translational N-terminal acetylation By Rawad Hanna, Andrey Rozenberg, Daniel Ben-Yosef, Tali Lavy, and Oded Kleifeld

      Summary of key results:

      The manuscript "In-depth characterization of apoptosis N-terminome reveals a link between caspase-3 cleavage and post-translational N-terminal acetylation" by Rawad and co-authors reports on a negative enrichment strategy, named LysN Amino Terminal Enrichment (LATE) to perform N-terminome analysis, a strategy which complements the cohort of existing negative enrichment strategies thereby jointly permitting a more comprehensive capture of the (neo-)N-terminome by additionally enabling the capture of (neo-)N-termini with (semi-)Lys-N specificity. The authors provide preliminary evidence that Nt-acetylation is protective for a proteins' N-terminus to be cleaved by caspase-3 besides the occurence of putative post-translational Nt-acetylation occurring on neo-N-termini generated upon caspase-3 cleavage.

      Concerns:

      Page 4 - In contrast to the hindrance of N-terminal amine ionization by Nt-acetyl groups concluded by the authors, previous studies reported an improved MS-scoring if α-amino-acetylated (tryptic) peptides by the higher numbers of b and y fragment ions observed as compared to α-amino-free (tryptic) peptides (e.g. (Staes et al., 2008)). It is rather the lack of any N-/C-terminal charged residue in case of Lys-N type N-termini which makes LATE less suitable for studying N-terminal protein acetylation.

      Page 4 - Besides indication the retained N-termini with high relative caspase-3/control abundance ratio's as putative caspase-3 proteolytic products, also indicate that unique peptides were retained, as many such singletons were reported in previous (caspase-focussed) degradomics studies making use of differential proteomics (e.g. (Van Damme et al., 2005)). Therefore the cut-off ratio of 2 rather seems unsubstantiated, unless the cellular proteomes of so-called control cells were affected by caspase activation. As such, showing some representative MS-spectra of neo-N-termini would be informative.

      Page 4 - replace 'without labelling of lysine residues (epsilon-amines)' to 'without notable labelling of lysine residues (epsilon-amines)', as residual labelling of lysine side-chains was observed. Also in case of the latter, do note that reduced MS-ionization potential might impact labelling efficiency calculation, and chromatographic detection of labelling efficiency should be considered to conclusify this finding.

      Page 6 - The experimental setup comparing caspase-3 overexpressing and ABT-199 induced versus ABT-199 induced cells will be highly biased as it will not be able to detect efficient caspase-3 cleavages (Plasman et al., 2011), as such cleavage events are complete and thus do not require any additional overexpressed capase-3. I see this as an important flaw and the authors should demonstrate that the list also includes efficient caspase-3 cleavages.

      Page 12 - The setup doesn't permit ORF N-terminal stability per se, rather the cleavage susceptibly of N-termini holding (a) putative caspase-3 cleavage site(s). Please adjust accordingly. Again since the setup might have missed efficient cleavages, the assessment might be biased.

      The claim that Nt-acetylation is protective for caspase-3 cleavage should be validated by monitoring cleavage efficiency of an Nt-acetylated versus an Nt-free variant (e.g. by introducing a Pro residue at AA position 2, or comparing cleavage efficiencies in corresponding NAT knockdown versus control cells) of an identified caspase substrate (i.e. a substrate holding a caspase-3 cleavage site in its N-terminal sequence) versus its Nt-free counterpart

      Page 12 - Since post-translational Nt-acetylation of neo-N-termini could be reproduced in vitro in the non-dialyzed sample, enzymatic over chemical Nt-acetylation should be demonstrated (e.g. by the use of a (bisubstrate) NAT inhibitor).

      Other concerns:

      Abstract - The abstracts holds complex/incorrect sentence constructions (e.g. simply indicate 'Protein N-termini', '... undergo ... processing by proteases' (currently: 'not be processed by proteases').

      Abstract - 'To expand the coverage of the N-terminome' only applies when this is used in conjunction with other negative enrichment strategies as by itself, LATE doesn't intrinsically provide a better coverage of the N-terminome (this is also noted at page 2).

      Change 'that cannot be identified by other methods' to 'that cannot be identified by other negative selection methods'

      Page 1 - Suggestion to change 'Proteases are typically described as degradative enzymes' to 'Proteases used to be described as degradative enzymes'

      Page 1 - Not really correct how written; 'N-terminomics methods highlight the N-terminal fragment of every protein (N-terminome)'

      Page 2 - Positive selection techniques .... Enrichment of unblocked (or Nt-free) N-termini

      Page 2 - Besides altering charge, Nt-acetylation also alters hydrophobicity ...

      Page 2 - remove 'to better chart'

      Page 2 etc. - Do note that caspase-3 can potentially activate downstream caspases in vitro

      Page 3 - functional crosstalk between proteolysis and neo-Nt-acetylation has already been demonstrated in the case of co-translational acting methionine aminopeptidases and chloroplast N-terminal acetyltransferases. Adjust accordingly.

      Page 3 - when discussing the identification of ORF N-termini, note that some of the strategies of which note when used to enrich for in vivo blocked N-termini, can also be used without blocking/labelling of Lys residues, and thus trypsin will also result in Lys-ending peptides. This is important to consider in this context.

      Page 3 - remove the following sentence part; '... or run individually which can be useful for quantifying naturally modified N-termini.', since also a differential/labelled proteomics setup enables such assessment. Related to this, the authors should comment on the observation that much fewer (i.e. less than 40%) Nt-acetylated N-termini were identified by LATE as compared to HYTANE. How is this reflected in the number of PSMs? Probably these difference are further intensified when considering PSMs.

      Page 6 - Informative to indicate how many of the in silico predicted putative DEVD P4-P1 cleavages were actually present in the list of 2049 putative cleavages identified.

      Page 6 - Unclear if any of the of 2049 putative cleavages, included non-canonical P1 cleavages besides the P1 Asp and Glu cleavages identified.

      Page 6 - Were the 'regular' cells mock transfected?

      Page 6 -Important to note that an ORF can have multiple N-termini besides neo-N-termini (e.g. in the case of alternative translation initiation)

      Page 6 - The authors should be more careful with generalization when comparing LATE and HYTANE (and other degradomics approaches) as in this study LATE was only applied for the identification of caspase-3 neo-N-termini, which by its extended substrate specificity might hold specific features enabling the preferred detection by one technique over the other. Also note that as compared to less recent studies, evidently the MS instrument used is a key factor in the increase in cleavages reported in the current study.

      Page 9 - The authors should provide some info/supporting statistics in the text regarding the new putative substrates showing GO-enrichments (compared to which control?) similar to previously reported caspase-3 substrates.

      Page 11 - Indicate that the 11 neo-N-terminal peptides of which note are the neo-Nt-peptides matching (signal peptide) cleavages indicated in the Uniprot database. Were any corresponding di-methylated neo-N-termini of these cleavages identified? In case of the 'other' proteolytic cleavages of which note, refer to these as not-annotated in UniProt.

      Page 11 - post-translational Nt-acetylation is abundant in plant and the responsible NAT has been identified, please reference these studies as well.

      Page 12 - Define 'undoubtedly dependent on caspase-3 cleavage'

      Page 14 - The NAA30 discussion is not really relevant for the discussion of the post-translational Nt-acetylation of mitochondrial neo-N-termini.

      Viewing the harsh in vitro caspase-3 cleavage condition used, namely 1 µg caspase 3 over 20 µg protein, the P1 specificities of all identified neo-N-termini should clearly be shown.

      Since acetylation of serine and threonine residues are reported forms of post-translational modification, and many so-called past-translational Nt-acetylated neo-N-termini harbour such AA residues in their N-terminal sequence, b-ion coverage for these neo-N-termini should be provided/inspected.

      References

      Plasman, K., Van Damme, P., Kaiserman, D., Impens, F., Demeyer, K., Helsens, K., . . . Gevaert, K. (2011). Probing the efficiency of proteolytic events by positional proteomics. Mol Cell Proteomics, 10(2), M110 003301. doi:M110.003301 [pii] 10.1074/mcp.M110.003301

      Staes, A., Van Damme, P., Helsens, K., Demol, H., Vandekerckhove, J., & Gevaert, K. (2008). Improved recovery of proteome-informative, protein N-terminal peptides by combined fractional diagonal chromatography (COFRADIC). Proteomics, 8(7), 1362-1370. doi:10.1002/pmic.200700950

      Van Damme, P., Martens, L., Van Damme, J., Hugelier, K., Staes, A., Vandekerckhove, J., & Gevaert, K. (2005). Caspase-specific and nonspecific in vivo protein processing during Fas-induced apoptosis. Nat Methods, 2(10), 771-777. doi:nmeth792 [pii] 10.1038/nmeth792

      Significance

      The manuscript "In-depth characterization of apoptosis N-terminome reveals a link between caspase-3 cleavage and post-translational N-terminal acetylation" by Rawad and co-authors reports on a negative enrichment strategy, named LysN Amino Terminal Enrichment (LATE) to perform N-terminome analysis, a strategy which complements the cohort of existing negative enrichment strategies thereby jointly permitting a more comprehensive capture of the (neo-)N-terminome by additionally enabling the capture of (neo-)N-termini with (semi-)Lys-N specificity. The authors provide preliminary evidence that Nt-acetylation is protective for a proteins' N-terminus to be cleaved by caspase-3 besides the occurence of putative post-translational Nt-acetylation occurring on neo-N-termini generated upon caspase-3 cleavage.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Manuscript number: RC-2022-01682

      Corresponding author(s): Peter Keyel

      1. General Statements

      We thank the reviewers for their thorough and critical analysis of our manuscript. We have addressed most of the concerns and questions with our revised version. To address the remaining concerns, we plan to perform two lines of experiments— aerolysin sensitivity of dysferlin null C2C12 muscle cells and aerolysin sensitivity of ESCRT-impaired cells. When these experiments are complete, we believe the revised contribution will provides important novel insights into membrane repair that will appeal to a broad audience.

      Reviewer comments below are in italics.

      Description of the planned revisions

      Reviewer 1

      Major

      In order to show that patch repair is indeed protecting cells against aerolysin, the authors should disrupt patch repair of the cells under study and observe and increased toxicity.

      Reviewer 2

      Major

      *1. The effect of dysferlin overexpression does not indicate that patch repair is a protective mechanism or that dysferlin plays a significant role in aerolysin resistance. The authors should knock out dysferlin and assess cell resistance to lysis. *

      Reviewer 3

      Significance

      The work presents a foundation to further investigate into the mechanism of aerolysin function, following the discovery of the role of extracellular Ca2+ in its activity. As aforementioned, the role of dysferlin in resisting aerolysin also has potential, but the limitations of this work were discussed including the absence of performing a dysferlin knockout, although performing this experiment may help to strengthen the current finding.

      We agree with all 3 reviewers that a dysferlin knockout will complement our gain-of-function studies and this will strengthen the manuscript. We plan to challenge C2C12 myocytes that express control shRNA or dysferlin shRNA with toxin and determine their sensitivity.

      We chose this system instead of targeting a patch repair protein in HeLa cells for 3 reasons. First, it will provide the corresponding loss-of-function experiment to match the gain-of-function experiments we have already done. Second, other patch repair proteins work redundantly with other proteins, complicating their knockdown and/or their disruption may interfere with lipid/protein transport. Finally, dysferlin null C2C12 cells are commercially available, so other groups will have an easier time replicating our results.

      Reviewer 1

      Significance

      *and in the statement that a cellular process that has been artificially introduced in the experimental system is the cellular protection mechanism against aerolysin attack. In order to prove that this process is a bona fide protection mechanism, the authors should show that it is present without the need of overexpressing a protein that is not expressed at all either in the used cell line (HeLa), or in the natural cellular target of aerolysin (epithelial cells). The significance of the proposed protection mechanism is therefore questionable. *

      We plan to address this concern by using C2C12 muscle cells that have and do not have dysferlin. Muscle cells are natural cellular targets of Aeromonas during necrotizing soft-tissue infections.

      Reviewer 2

      Major

      *2. ESCRT complex was shown to play a role in plasma membrane repair following mechanical damage or perforin treatment of cells (Jimenez 2014, and Ritter, 2022). Whether ESCRT is important in aerolysin pore repair can be assessed by knocking out the Chmp4b gene or overexpressing dominant-negative mutant of VPS4a, E228Q. *

      We plan to use a previously characterized (Lin 2005 PMID: 15632132) inducible system (TRex cells) to express the dominant negative VPS4b E235Q in cells. We plan to pulse cells for 2 h with 1 ug/mL doxycycline one day prior to the assay. This pulse time and dose strikes a balance between cell death due to non-functional ESCRT, and compromising ESCRT function. Then we will challenge parental cells (TRex) or TRex cells expressing VPS4b E235Q with toxin and measure lysis. We also plan to compare plus/minus doxycycline as a further control. We will also use fluorescent toxins to compare binding across cell types.

      One caveat on the ESCRT work is that ESCRT has an essential role in MVB formation, and ESCRT effects might be due to perturbation of protein/lipid flux through this system in addition to their recruitment to the plasma membrane. Even with knockdowns and overexpression, it can be challenging to interpret some of the pleiotropic effects of altering the ESCRT complex. While we do not contest the role for ESCRT in plasma membrane repair, we suspect the role for ESCRT will be more complicated than previously appreciated. Digging deeper into these possibilities beyond our proposed experiment is beyond the scope of this manuscript.

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

      Reviewer 1

      *Major: The authors conclusions contradict established results, which they cite. Yet experimental conditions are not similar in two ways: toxin concentration-wise and toxin treatment duration-wise. *

      We agree with the reviewer that there were differences in experimental design between our study and the other cited studies. Due to the cited differences, our results, Gonzalez et al and Larpin et al are not necessarily contradictory on most points. Our conclusions differ from Gonzalez et al in that we do not think K+ efflux drives repair in the first hour, and differ from Larpin et al in that we observe Ca2+ flux after aerolysin challenge. Along with the toxin variables discussed below, we also discussed the potential cell type differences between the studies that may account for the discrepancy. We have now included these additional differences in our manuscript on line 435 for Larpin et al and lines 423-425 for Gonzalez.

      Our study set out to do something distinct from the prior studies. The prior studies did not compare the efficacy of distinct membrane repair mechanisms to the same toxin because that was not their study aim. Hence, our goal is not to prove the prior literature wrong, but contribute to a better understanding of the immediate membrane repair events triggered by aerolysin. We argue that the significance of our contribution is this comparative approach to membrane repair, which has not previously been done, and our finding that aerolysin engages distinct, but overlapping mechanisms compared to CDCs. We have updated our significance to better convey our advance, which is explained on lines 99-102, 128, 519-525.

      *While we appreciate the efforts of the authors to standardize the concentration of toxins used based on hemolytic units, we note that the concentrations used are very much higher than in the other studies cited. Indeed, based on table 1, materials and methods, and the various experiments, aerolysin has a LC50 of approximately 200 HU/ml, which corresponds to about 2 ug/ml. This is approximately 200x more concentrated than for example in Gonzalez et al 2011 and Larpin et al. 2021. It makes the validity of direct comparison with those studies questionable. *

      We agree with the reviewer that the toxin concentrations are different from prior studies. This is why we argue hemolytic activity needs to be reported along with toxin mass.

      One potential explanation for this difference is purification method. We do nickel NTA purification from whole bacterial lysates, instead of from the periplasm. It is possible that the most active aerolysin precipitates early or is otherwise lost in our purification process, which accounts for both the lower toxin specific activity and lack of toxin precipitation during trypsin activation that we observe. To control for impurities, we purified two preps of our aerolysin to >90% purity after nickel beads. However, we did not observe a significant change in specific activity or cytotoxic activity. We interpret this finding to suggest there was a trade-off between improved specific activity due to increased purity and loss of specific activity due to toxin inactivation during the extended purification process.

      We have included a new figure (Fig S10) showing our toxin purification and activity.

      *We noticed that the authors activate pro-aerolysin at high concentration (in the range of 1 to 5 mg/ml) and at room temperature. In our experience, under these concentration, activation leads to immediate oligomerization and massive precipitation. The final concentration of active toxin is thus unknown. *

      When we titrated the trypsin to determine the optimal concentration of trypsin to use, we did not observe oligomerization/precipitation (Fig S10B). If there was precipitation of aerolysin after trypsin treatment, we would expect a difference in cytotoxicity between pro-aerolysin and aerolysin treatment. We did not observe significant differences in cytotoxicity between pro-aerolysin and activated aerolysin (see Figs 1-2). Finally, we measured hemolytic activity on trypsin-activated toxin, so any precipitation would be expected to occur prior to assessing hemolytic activity. Thus, we argue our use of hemolytic activity measured after trypsin activation mitigates this risk.

      * The authors keep their cells in toxin-containing medium for the whole duration of the experiments, typically 45 minutes. This is in stark contrast with 45 seconds to 3 minutes transient exposure to toxin in Huffman et al 2004. *

      We agree this is one of the differences. We also note Huffman et al examined cells at 6 or 28 h later. While we ruled out the impact of MAP kinases on membrane repair occurring within 30 min of toxin challenge, we make no claims about their ability to promote cell survival at later time points. We have clarified these differences in the manuscript (line 461).*

      The authors do not report binding and oligomerization assays of the toxins. The only figure showing a western blot (fig. 7) is of low quality and shows unexpected observations. Aerolysin Y221G mutant is expected to bind and oligomerize. Yet, no band is present at about 250 kDa (expected oligomer) or at about 47 kDa (monomer). In addition, in aerolysin lanes (1 and 2) the oligomer is saturated, seems to be covering three lanes, indicating a possible spill-over. *

      We performed binding studies in Fig S3C and Fig S5. For Fig 7, in the original blot, the cell lysate is a wider band than the MV band, but there are only two bands, that remained in their respective lanes. We have now included another independent biological replicate of the aerolysin blot as Supplementary Fig S7D which shows clear demarcation between cell lysate and MV pellet. This blot was not included in the main figure because in the process of stripping and reprobing for all of the targets, we lost detection of our penultimate targets. We agree with the reviewer that oligomer bands for the Y221G were very faint, and we expected them to be stronger. In the new blot (Fig S7D), some oligomer can be detected. As a result, we are hesitant to risk over-interpreting these findings.*

      Finally, while the patch repair hypothesis is interesting, it is unclear why the authors decided to overexpress dysferlin in cell lines that normally do not express it. Sure, there is a repair phenotype but this phenotype is artificially introduced. Dysferlin is not expressed at all in HeLa cells. *

      One challenge with membrane repair is the difficulty perturbing the system due to redundancies. While loss-of-function experiments are important, gain-of-function experiments also add confidence to the system. The simplest way to perform a gain-of-function experiment is to add a well-known patch repair protein to a well-characterized cell line lacking it. Thus, exogenous expression of dysferlin enables us to test the hypothesis that increasing patch repair enhances repair against the toxins.

      We have included this rationale now in the manuscript, lines 366-369

      *Furthermore, dysferlin is not expressed in epithelial cells, which are the prime target of aerolysin. Why then focus on this protein? *

      We chose dysferlin because it is well-characterized as a patch repair protein, whose defect causes Limb-Girdle Muscular Dystrophy 2B and Miyoshi Myopathy. Additionally, setting up this assay enables future work to probe the role of individual dysferlin domains in patch repair.*

      Minor: The graphic legends should be boxed out to be clearly separated from the data. In Figure 4A, it is mixed up with the data. *

      This has been corrected.*

      Some western blots are saturated, e.g. B-actin in figure 4B. Full blots should be provided. *

      We have added full western blots as requested as Supplementary Figs S11-12.*

      In the methods, aerolysin sublytic dose for HeLa cells is specified at 62 HU/ml. In figure 5C and D, 31 HU/ml kills more than 50% of HeLa cells. This is not compatible. *

      Even when controlling by hemolytic activity, and toxin prep, we find some variability in toxin activity between assays. For the live cell experiments, 62 HU/mL remained sublytic despite the higher activity in the flow cytometry assays. We controlled for death in our live cell imaging experiments, by including TO-PRO. This confirmed the toxin was at a sublytic dose in those experiments.

      We included a new figure S10C to show the variation in LC50 per assay as a function of toxin specific activity. We have clarified that the sublytic dose was for live cell imaging experiments, lines 640-641.

      *Figure 2A and B have quite different LC50 for starting conditions ({plus minus} 200 HU/ml in A, 600-700 HU/ml in B). Why is it so different? Y-axis has a linear scale in A and a logarithmic scale in B. It would make comparison easier to have the same scale in both panels. *

      We agree there is variability between assays. We note that toxin doses change vary in other manuscripts that report toxin mass. For example, aerolysin varies by 10-fold (2 – 20 ng/mL) between figures in Gonzalez et al 2011. We interpret this variation as a common challenge for toxin studies. We mitigate this challenge by including controls for each assay so the relative change can be assessed. We provide additional transparency by including Fig S10 to show batch-to-batch variability of both our toxin preps and assays.

      We have changed the scale to linear in Fig 2.*

      The letters detonating statistically significant groups are sometimes unclear. For example in Figure 1A and B, PFO belongs to group a and b simultaneously. What does this mean? *

      Samples that share letters are not statistically distinct from each other. In the example cited, PFO is not statistically significant compared to all other bars with an a and is not statistically significant compared to all other bars with a b. While confusing at first, the alternative is a mess of stars and bars.

      This has been explained in lines 981-985.*

      In Figure 8, aerolysin hat a LC50 in cells overexpressing GFP-Dysferin of approximately 1700 HU/ml in A and of approximately 400 HU/ml in B. Why is it so different? *

      This is due to intra-assay variation. We include controls for each assay to ensure the trend remains consistent.*

      In Figure S1, it is unclear what the plots « all events » vs « single cells » mean. *

      We have clarified these plots.*

      In the discussion, the authors write « First, survival did not correlate with overexpression, which would be expected if dysferlin acted as Ca2+ sink ». What is meant? GFP-dysferlin overexpression does correlate with survival in Figure 1A. *

      We meant that the extent of Dysferlin expression did not correlate with survival. If Dysferlin acted as a calcium sink, cells expressing 100x dysferlin levels should be more resistant than cells expressing 1x dysferlin levels. If Dysferlin needs to serve a cellular function, the brightest cells may not be more resistant (or even be less resistant due to aggregates, etc). We checked to see if the brightest Dysf+ cells had better survival than the dimmest Dysf+ cells. They did not. However, all Dysf+ cells had better survival than Dysf- cells.

      We have updated the manuscript (lines 496-498) to reflect these changes.

      Significance

      *General assessment: The study strength lies in the several possible protection mechanisms that are tested. The weaknesses lie in the contradictions of the results reported here with established mechanisms, *

      We disagree with the reviewer that findings that contradict previously proposed mechanisms are a weakness for significance. Instead, we argue this is a strength of our study’s significance. Replication of prior studies’ conclusions using distinct experimental conditions is critical for the reproducibility and rigor of the underlying science, and may give new insights into toxin biology. While we acknowledge the differences in approach, these differences narrow the prior mechanisms that may have been assumed to be widely applicable. The finding that they cannot be replicated in our system suggests one or more of the differences between the studies may drive a critical aspect of aerolysin biology. For example, the Ca2+ difference with Larpin et al could be due to a cellular Ca2+ channel present in HeLa cells that is absent in THP.1/U937 cells.

      This distinction is expected to spur additional research in the aerolysin field.

      * Advance: The study contradicts previously established results but the experimental conditions used here are quite different to those used in the earlier studies, which makes the comparison quite difficult. As such it does not really fill a gap. *

      We have rephrased the significance to better convey both the gap our study fills in membrane repair and the advance that it has made. See lines 99-102, 128, 519-525.*

      Audience: The study will be of interest of specialized audience. *

      Given the emerging broad importance of membrane repair in response to endogenous pore-forming toxins, and the large gaps in the field of membrane repair, we respectfully disagree with the reviewer. We have revised our significance statements to better convey this broad appeal. See lines 99-102, 128, 519-525.

      Reviewer 2

      Major

      *3. I find the optimisation of lysin concentrations and data presentation quite confusing. I eventually understood, what was done, but I feel that the authors should be able to transform the data and plots so these are more accessible to a reader, eg a simple dose/time-response curves would be very helpful in that respect. For example, in Figure S1E, why does aerolysin appear to be less cytotoxic after 24 hrs than after 1 hr. In principle, I would expect to observe an additive effect, i.e. cell death at 1, 3, 6, 12, and 24 hrs should add to 100%; however, if 100% cells die at 500HU/ml, how can more cells die after 24hrs? Or am I missing something in the experimental design/data presentation? *

      We agree that presenting the results from cytotoxicity can be challenging. We use LC50 in the main text because it is easiest to understand. However, we provide all dose-response curves underlying those numbers in the supplemental data. We recently published our approach to assays and data analysis (Haram et al PMID: 36373947) to make it easier to understand.

      In Fig S1E, each time point is a distinct assay. In contrast to the approach suggested by the reviewer, where we read the plate at different timepoints, we used different replicates to generate the time points. As a result, the % will not add to 100. Instead, we observe that the majority of cell death occurs in the first hour. We have clarified our discussion of Fig S1E, lines 154-155.

      At 24 h, it is possible that cell growth interfered with the assay. The plate has a finite surface area. If control cells are confluent near the start of the assay, but toxin-treated cells are not due to cell death by aerolysin, the growth rates may not be equal. Since our focus is on proximal membrane repair events, and not on late signaling events, pursuing this further is beyond the scope of the current manuscript.

      *I also wonder whether using haemolytic units is appropriate (it may well be, if justified), given that the toxins used here have various membrane-binding properties. Wouldn't it make more sense to compare the cytotoxicity using nucleated cells? *

      We agree with the reviewer on the need for standardization, and do compare cytotoxicity using nucleated cells (HeLa). Our first level of standardization is the use of hemolytic units instead of toxin mass. This normalizes toxin activity to the ability to kill human red blood cells, which are widely accepted as having minimal membrane repair mechanisms. This gives us a baseline activity, and allows us to control for toxin impurities/differences between toxin preps/toxins. We prefer cytotoxicity over membrane binding for our baseline because it is a functional assay.

      After this first level of standardization, we compare the cytotoxicity in HeLa cells. This is one reason why the majority of our assays are performed in HeLa cells—we know how they behave at different toxin doses in our hands, the cells are easy to use, and we can standardize assays in the lab. We included HeLa cells as a control in Fig 5 to show the standardization requested by the reviewer. We split Fig 1 up differently to better convey the results.*

      1. The authors use "sublytic" concentrations of aerolysin (64HU) throughout most of the paper, but according to Figure S1C, 50% cells died at that concentration after 1hr, suggesting that when the cells were investigated over a shorter period of time, they were already dying - it's almost like the cells had life support turned off, but still being investigated as though they survived aerolysin treatment. This needs to be clarified or reassessed. *

      We agree with the reviewer that we did not track cell survival beyond 45 min in our live cell imaging assays. We labeled cells as ‘surviving >45 min’ to acknowledge the fact that these cells could have died at 46, 47, 60, or 600 min after the experiment ended. We focused on time points earlier than 45 min because proximal membrane repair mechanisms are expected to have occurred in that time, and had time to complete. We have updated the manuscript on lines 214-215.

      We next considered the reviewer’s excellent point that the cells alive at 30-40 min could be executing a cell death program. If this were the case, then based on our FACS data (Fig S1C), we would predict ~50% of total cells would be dead by 1 h. From Fig 3A, ~35% of the cells died in the first 45 min. From the remaining 65%, we would predict another 15% dying from this programmed cell death pathway, which would be 15/65 = ~25% of the surviving cells. We did not notice 1/4 of the surviving cells behaving distinctly. For example, the large error bars in 3H is due to a range of cell behaviors that we could not easily subgroup. For individual cells (shown in Figs 6 and 7), there is similarly no clear demarcation of 1/4 of the cells. While we see a gap with pro-aerolysin, that is ~1/3 of the cells (not the expected 1/4), and it is not repeated with aerolysin. While we can’t rule out a cell death program contributing to the top or bottom 1/4 of our results, removing the top or bottom 25% of data points would not alter our major conclusions from the live cell imaging. If a programmed cell death pathway that occurs in the 30-90 min range is identified for aerolysin, it would be interesting to see how that pathway changes repair kinetics. However, that would require identification of the death pathway.

      *

      1. What effect does the addition of 150mM KCl have on the plasma membrane, trafficking/repair - wouldn't the plasma membrane be depolarised? There were a number of papers by John Cidlowski in mid 2000s, where his team explored the effect of potassium supplementation on apoptosis - this may be worth exploring. *

      We thank the reviewer for suggesting these interesting papers. We have explored these papers, and our understanding of them is as follows. Franco et al 2008 PMID: 18940791 shows that ferroptosis is independent of high extracellular K+. This contrasts with Fas-dependent apoptosis, which is suppressed by high extracellular K+. This is consistent with the Cidlowski group’s other work (eg Ajiro et al 2008 PMID: 18294629) and Cohen’s group (eg Cain et al 2001 PMID: 11553634) showing that apoptotic DNA degradation performs better at low K+, and extracellular K+ interferes with apoptosis. Similarly, other papers have shown that NLRP3-activated pyroptosis can be blocked by addition of extracellular K+. Depletion of intracellular K+ inhibits endocytosis and other vesicle trafficking pathways.

      While these are good papers, they do not directly relate to our K+ findings, which is that blocking K+ efflux via elevated extracellular K+ levels has no impact on aerolysin-mediated killing. Therefore, to stay focused on the repair pathways, we opted not to include these papers to avoid distracting the reader from our key points. *

      1. Figure 3 and accompanied text: it would be more informative to show all the data rather than breaking it down to 45 min. In my view, *

      We have added histograms to show when individual cells died during the assay as supplemental Fig S3E. We used the three bins for the exact reason articulated by the reviewer—we wanted to consider cells that died fast vs slow differently. However, in order to interpret the data, a cutoff of 5 min was chosen as optimal. While we agree with the reviewer that the 5 min death could be dismissed, we presented the data to avoid questions about why we omitted those data.*

      1. I am curious whether EGTA diffuses into the cytosol through aerolysin pores. If so, then unlike BAPTA-am it would affect Ca inside and outside the cell. *

      We agree with the reviewer this is an interesting question. While EGTA might diffuse into the cytosol, its binding properties suggest it would be unsuitable to block cytoplasmic Ca2+ transients (see Nakamura 2019 PMID: 31632263). BAPTA binds to Ca2+ ~40x faster than EGTA, which enables it to capture Ca2+ prior to Ca2+-binding proteins. In contrast, EGTA is thought to be too slow to sequester intracellular Ca2+ before Ca2+-binding proteins. While EGTA might perturb Ca2+ close (

      *Are the authors confident that in the absence of extracellular calcium (EGTA treatment), aerolysin formed the pores at all? Have they looked, for example, at intracellular Na/K, or have any other evidence of membrane disruption? *

      Prior structural studies suggest that Ca2+ is not required for aerolysin pore formation. For example, Iacovache et al (2011) PMC3136475 induce oligomerization with low salt and pH 2+. Cryo-EM from the same group (Iacovache et al 2016 PMID: 27405240), showed pore formation under similar conditions.

      In Fig S3, aerolysin kills in the presence of EGTA at higher concentrations, suggesting that it can form pores when EGTA is present. Also, in Fig 2D, we used Tyrode’s buffer, which was made without Ca2+ or EGTA. We added the indicated amounts of Ca2+ in, and observed a reduction in lysis at low [Ca2+]. This argues against EGTA interfering with toxin oligomerization/pore formation because EGTA was not present, and the toxin still failed to kill.

      We have updated the manuscript (lines 203-205) to emphasize this point.*

      1. Figure 6 (and some other): I find the designation of statistical significance (a-f) quite confusing, as it is unclear which comparisons are statistically different. Looking at Figure S5, there was no difference between the effect of Annexin depletion on the toxicity of the three lysins. *

      Samples sharing the same letter are NOT statistically significant. This is done to avoid a mess of stars and bars with multiple comparisons. This has now been explained in lines 981-985.

      For Fig 6/ Fig S5 (now S6), there was a statistically significant difference in LC50 between control siRNA and Annexin knockdowns for SLO. We agree that visually the dose-response curve in Fig S6B looks similar. However, we note that the x-axis is a log2 scale, and the control line is distinct over the 250-1000 region. When we calculate the LC50, these differences give different LC50 values. Over multiple reps, these differences were consistent enough to be statistically different.

      Significance

      *The paper attempts to address an interesting question of aerolysin pore repair, and it is interesting from the perspective of a potential difference between various pore-forming proteins. *

      We agree with the reviewer and thank the reviewer for this assessment.*

      The study will be potentially interesting to a broad audience of biochemists/cell biologists and microbiologists working in the field of pore-forming proteins/virulence factors. *

      We agree with the reviewer and thank the reviewer for this assessment.

      Reviewer 3

      *Major comments In the first instance, the authors use a method of assaying the specific lytic activity of aerolysin in comparison to a number of different CDCs. Whilst it is acknowledged that these methods have been published in peer-review papers previously (e.g. Ray et al., Toxins, 2018), it would be great to have more information of how the specific activity is derived. Currently there is a convoluted method that makes a number of assumptions such as, but not limited to, 1) the number of dead cells measured in the FACS experiments is proportional to the activity of the different classes of PFPs however the authors do not show how they account for PFPs leading to loss of cells into debris which would involve a total cell count and *

      We thank the reviewer for raising these concerns. We tested these assumptions in our previous papers. We compared the FACS assays to other assays that measure total cells (i.e. MTT assay), and found that the FACS assay corresponds with the MTT findings. These findings were published in Keyel et al 2011 PMID: 21693578 and Ray et al 2018.

      Loss of countable events to debris is detected in our assay as saturation of cell death at a number under 100%. Since we perform dose-response curves, we can determine when the killing saturates. This is why loss of countable events does not change our ability to accurately calculate LC50.

      2) how the inflection or linear point is identified on individual experiments (e.g. Supp. Fig. 1B, 2A, 2B, 3A, 3B to name a few) and how reliable these points are (e.g showing the data points with model sigmoidal (?) curve and corresponding R values).

      This had been calculated manually in the prior version of the manuscript. To address the reviewer’s concern and to improve data quality, we reanalyzed all of our data by fitting our dose-response curves to logistic models, and determining the LC50 using that model. An in-depth explanation of our approach was just published in Haram et al PMID: 36373947, which we now cite (line 821). *

      Furthermore, the batch-to-batch variability of protein samples presented in table 1 may be an issue where inactive but folded protein can affect the formation of homo-oligomer pores so more effort to reduce the effects of batch variation would be integral to the foundation of this paper. Given that aerolysin has a very different action on cells then this new characterisation should be provided regardless of what has been previously published by the authors on the activity of CDCs on the cells.*

      We agree with the reviewer that batch-to-batch variability is a key concern for pore-forming toxins. To address the concern of batch-to-batch variability and toxin purity, we have added Supplemental Fig S10. In Fig S10C, D, we plot the LC50 against specific activity of each toxin prep when used against control cells. We found a statistical difference in LC50 between two of our toxin preps, but not between any of the others. Notably, there was no association between increasing specific activity and LC50.

      Furthermore, we tested the impact of impurities on our toxin prep. While we purify most toxins only using His-beads (obtaining ~40% purity) (Fig S10B), we purified two toxin preps to higher purity (>90%) (Fig S10A). We did not observe differences in LC50 between these toxin preps. The specific activity for these toxins did not increase. We interpret that finding to indicate the gain in specific activity for purity was offset by the loss of specific activity due to prolonged toxin purification.*

      • Can the authors provide the raw data for the total FACS observations (scatterplot for all events) and show that there is no significant loss of cells? Or at least there is accountability of the cells? *

      Our stop conditions were to collect at least 10,000 gated events instead of running for a set period of time/set volume to determine cell density. We provide example scatterplots in Fig S1A.

      * - Can the authors provide more information about how the linear regression on Supp. Fig. 1B and other experiments showing the model sigmoidal curve performed such that this work is more reproducible? *

      We agree with the reviewer that using logistic modeling would strengthen the work. To address this concern, we reanalyzed all of our data and switched to logistic modeling. This improved reproducibility for many figures. Changes that add or remove statistical significance to results include Fig 4A, loss of significance between Ca2+/DMSO and BAPTA/DMSO, Fig 6C, loss of significance for siRNA knockdown of A6 vs scrambled for ILY, and Fig 8A/B, gain of statistical significance for GFP-Dysf protecting SLO. We have updated our results accordingly.*

      The SEMs of some data points (specific lysis LC50 scatterplots, for e.g. Fig. 2C, 4A, 4C, 8A and fMAX plots, for e.g. Fig. 3B) may not be apparently representative of the skew (e.g. and individual values (including outliers). A clarification of the statistical analysis behind the results may benefit in a clearer understanding of how the SEMs were calculated and presented in the main figures. Also, further elaboration on the meaning of the lettering in the scatterplots (denoted as a, b, c etc.) across the main figures may help improve the interpretation of the data. *

      The SEMs were calculated by Graphpad and graphs also generated by Graphpad. To address the reviewer concern, we have switched all places where we plotted individual data points to median with no error bars. This will enable the reader to judge skew, outliers, etc without reliance on error bars.

      We have now further elaborated on the lettering in the scatterplots. Samples sharing the same letter are NOT statistically significant. This is done to avoid a mess of stars and bars with multiple comparisons. This has now been explained in lines 981-985.*

      Secondly, the authors present interesting results on the significance of Ca2+ on aerolysin's mechanism behind lytic activity and introduces dysfurlin-mediated patch repair as the primary cellular resistance mechanism against aerolysin mediated lysis. Results from Figure 2-4, indicate that extracellular Ca2+ plays a role in aerolysin's function and cell lysis (aerolysin triggers influx of extracellular Ca2+). However, the results presented in figure 8 suggest an impairment of dysferlin translocation from the cytosol to the plasma membrane upon removal of extracellular Ca2+. If this were the case, wouldn't dysferlin impairment sensitise cells to aerolysin? Thus, in these sets of experiments it seems that Ca2+ is a confounding factor.*

      We agree that Ca2+ is a confounding factor, which is one reason we aimed to define better membrane repair mechanisms in response to different pore-forming toxins. Our interpretation is that Ca2+ triggers a death pathway that overcomes repair, and that aerolysin toxicity is due to the activation of this pathway. In this case, the impairment of Ca2+-dependent pathways does not reduce survival because the extent of damage is reduced/not present. Figuring out this death pathway is beyond the scope of the present manuscript, but a one future direction in which we are interested. This would also account for differences observed in different cell lines.*

      • Can the authors further elaborate on how the function of dysferlin in protecting cells against aerolysin contrasts to how aerolysin kills cells? *

      We have added the requested discussion to our manuscript, lines 519-525.

      *Finally, it is also interesting to see that cells deploy different resistance mechanisms between different families of pores. In saying that, the usage of CDCs seems to be inconsistent between each set of results. For example, intermedilysin (ILY) was used in the siRNA knockdown experiments but not in others such as Ca2+ influx assays, while PFO was only used for the initial set of results. A comment on this would benefit in understanding the rationale for selecting certain CDCs for each set of experiments. *

      We thank the reviewer for raising this point. We used SLO as the primary CDC in all the experiments because it is the CDC we have best characterized and have extensively published on. We included PFO in initial experiments to give readers a better idea of how multiple CDCs compare to aerolysin in target cells. However, since we’ve previously published on PFO, including it for later experiments would have increased cost and time of experiments without providing new knowledge.

      We used ILY because it binds to the GPI-anchored protein human CD59, so its binding determinant is more similar to aerolysin, which binds GPI-anchored proteins. We included it where practical to determine the extent to which targeting may change repair responses. Since ILY does not bind to murine cells, it was omitted from experiments using murine cells.

      We have added the rationale to the manuscript on lines 138-140.*

      Minor comments Results (Nucleated cells are more sensitive to aerolysin and CDCs) - A statement of the EC50 values of aerolysin and CDCs from the haemolytic assays would be beneficial to compare activities between the two pores. *

      The hemolytic activity is defined as the EC50 for the toxin in human red blood cells. The specific activity enables comparison of toxin activity, which is reported in Table 1. We have now added Supplementary Fig S10 which further plots the aerolysin and SLO specific activities against LC50 so that the reader can better assess batch-to-batch variability. In this study, we did not use enough batches of the other toxins to make this analysis useful for them.

      * - Figure 1A: As stated in the introduction, pro-aerolysin exists as a precursor that is functionally inactive unless activated by trypsin, furin or potentially other proteases. It would benefit the reader if an explicit statement were made about this activity and how it may come about in HeLa and 3T3 cells. Why is pro-aerolysin not shown in the Casp 1/11-/- BMDM cells? *

      The cell surface furin activity that activates aerolysin is not well-characterized across different cell types. We have revised the manuscript (line 76) to indicate these activities are present on the cell membrane.

      We omitted pro-aerolysin from the Casp1/11-/- BMDM because we performed those experiments earlier in the study before we started including pro-aerolysin. Based on the other results, we judged that the time and resource costs of adding pro-aerolysin in this system outweighed the gain to the story.

      * - Figure 1C: It was stated that "Casp 1/11 -/- Mo were ~100 fold more sensitive to pro-aerolysin and aerolysin compared to PFO and SLO" but did not show the activity for pro-aerolysin in these cells. *

      We thank the reviewer for catching this typo, and have corrected this statement (line 172).

      * - Supp fig 1E: Shouldn't 24 hr incubation of aerolysin to HeLa cells result in 100% specific lysis? *

      We agree with the reviewer that these results were surprising. At 24 h, it is possible that cell growth interfered with the assay. The assay well has a finite surface area. If control cells are confluent near the start of the assay, but toxin-treated cells are not due to cell death by aerolysin, the growth rates between control and experimental wells may not be equal. Since our focus is the proximal membrane repair events, and not the late signaling events, pursuing this further is beyond the scope of the current manuscript.

      * (Delayed calcium flux kills aerolysin-challenged cells) - What is the intracellular concentration of K+ normally in cells? Similarly, what is the intracellular concentration of Ca2+? *

      Intracellular K+ is ~140 mM (see Ajiro et al 2008 PMID: 18294629), while cytosolic Ca2+ is ~100 nM at rest.

      * - Figure 2C: Based on the description in the methods and results, both buffers are supplemented with 2 mM Ca2+ but one buffer (RPMI) shows more killing with SLO and ILY. Does this mean that both buffers contain 2 mM CaCl2? If so, what are the other potential reasons why one buffer enabled greater potency in CDCs? *

      RPMI has 0.4 mM Ca2+ prior to Ca2+ supplementation. However, the 2.4 mM Ca2+ did not provide improved protection compared to RPMI alone (See Fig 2 in Ray et al 2018).

      We suspect the various amino acids added to RPMI promote membrane integrity and account for the difference from Tyrode’s buffer. Glycine has previously been implicated in promoting membrane repair, but at higher concentrations than it is present in RPMI (0.133 mM in RPMI vs the mM concentrations used to protect cells). If other amino acids also protect, and/or why they protect is beyond the scope of the present work.

      * - Figure 3H: The data for aerolysin (WT) would greatly benefit for comparison to the inactive mutant (and indicate the sustained Ca2+ increase). *

      We have added this comparison, and updated the figure legend, line 1015.

      * - Supplementary Video V1: The addition of Triton X-100 permeabilises cells; however, this wasn't evident in (A). - Video V2: Similar to previous comment on Supplementary Video V1 (for B). *

      In V1A, the video was cut short to fit the play time with other videos. From addition, the triton takes a few minutes to diffuse to the cells and permeabilize them. In V2B, the cells do become permeabilized as shown by loss of the Ca dye. The cells are out of focus, which is why the nucleus TO-PRO is not detected.*

      (Calcium influx does not activate MEK-dependent repair) - Figure 4A: Effective ionic concentration inside and outside cell is increased (if intracellular Ca2+ becomes chelated); therefore, Ca2+ may enter the cell by passive diffusion or transport by other intrinsic Ca2+ channels. *

      There is already a very steep concentration gradient for Ca2+. The cytosolic Ca2+ is ~0.1 uM, compared with growth medium at 400 uM or assay buffer at 2400 uM. Chelation of the intracellular Ca2+ is not expected to increase Ca2+ import from outside the cell.*

      (Caveolar endocytosis does not protect cells from aerolysin) - Figure 5C: What is the purpose of using HeLa cells as a control? *

      We included HeLa cells to demonstrate the toxin was active and to rule out batch-to-batch variability as one interpretation of the reduced killing of differentiated 3T3-L1 cells.

      * - "..with Alexa Fluor 647 conjugated pro-aerolysin K244C" - this should be introduced earlier as it was initially mentioned in Supp. Figure 3C. *

      We have now introduced this earlier at line 190, instead of 300

      * - Murine fibroblasts were used earlier (Figure 1). Following from this result (where the WT can be used as a positive control), can MEFs be used instead of adipocytes to see whether caveolar endocytosis plays any role in cellular resistance? *

      The 3T3-L1 cells are murine fibroblasts prior to differentiation. Since they can also be differentiated into adipocytes, we used them instead of MEFs. The other reasons we used them include the availability of Cavin knockout cells, and the extensive caveolae present in adipocytes. We included analysis of 3T3-L1 prior to differentiation them in Fig 5B.

      * - Further comment on the increased resistance of K5 knockout would benefit on the mechanism of aerolysin-mediated cytolysis. *

      We agree further characterization of this line would be interesting in the future. At the present, however, any further comment would be speculative on our part. Since the resistance was not replicated in the second CRISPR line, we suspect it is either an unexpected mutation(s) in the cell line that arose during routine cell culture, or off-target effect(s) from the CRISPR used to generate the line.

      * (Annexins minimally resist aerolysin) - Supplementary video V3 - it seems that annexin A6 is recruited to the membrane, to a greater extent (and also quicker) than SLO. This suggests that annexin recruitment is a cellular response against aerolysin challenge. *

      We agree with the reviewer that annexins are recruited to the membrane during repair. However, individual knockdown did not enhance death. This is one reason we believe functional studies (i.e. cytotoxicity) are necessary when studying the cell biology of repair events. Recruitment of the protein, and it promoting repair may be two different things.

      In V3, three of the SLO-challenged cells have translocated by the time focus is restored. In contrast, the first aerolysin cells translocate ~10 min. One complicating factor is that A6 cycles back off the membrane with the SLO challenge.

      * o SLO also shows A6 recruitment (arrows pointed). However, supplementary figure 6B does not clearly illustrate this. *

      Given the 45 min time scale, the rapid initial membrane enrichment is hard to see on the graph.

      * - As annexin A1 is sensitive to calcium, further comment on the significance of intracellular/extracellular calcium in annexin A1 recruitment and aerolysin challenge would explain observations in Figure 4A. *

      We have updated the manuscript, line 242 to include annexins and dysferlin as Ca2+-binding proteins in our discussion of intracellular calcium.*

      (Patch repair protects cells from aerolysin) - Supplementary video V4 - the intensity decreases for the inactive mutant; is this due to lysis? *

      We included TO-PRO in the experiment to rule out lysis. Since the cells remain in focus, we interpret the lack of TO-PRO to indicate no cellular lysis.

      *- The next paragraph sounds like a contradiction: "GFP-dysferlin localized to the plasma membrane and vesicles independently of extracellular Ca2+ (Fig 8C D, Video V5) o Followed by "To study the Ca2+ dependency of dysferlin, we removed extracellular Ca2+ with 2 mM EGTA and challenged with sublytic toxin doses...found less depletion of dysferlin from cytosol". *

      We thank the reviewer for pointing out our unclear language. In the second section, we intended to refer to dysferlin positive vesicles. We have rephrased the manuscript (lines 388-395) to clarify that we are focused on Ca2+-dependence of vesicle fusion, not steady-state.*

      (Methods) - Table 1: The values presented in the methods section are, overall, confusing and require clarification. *

      We have added Fig S10, and discussion of toxin activity and purity in the methods (lines 634-641) to provide further clarity on toxin activity.

      * o 10-fold difference in SLO and PFO WT - do the authors think this might change the interpretation between different figures? *

      We do not. The reason is that we changed the membrane affinity between SLO and PFO (Ray 2018), and this switches the properties of the respective toxins without changing their yields.

      * o Understood how the haemolytic activity was calculated (referred to work in 2012), but how was the haemolytic unit originally derived? *

      It was derived as a measure of activity for toxins by determining the EC50 in RBCs for a given toxin. Since species type of RBC and other factors can change the reported activity, we have normalized to using human red blood cells. This lets us assay human-specific toxins like ILY along with other toxins.

      * o How were these values (from table 1) derived to toxin concentrations used for killing nucleated cells? *

      Full discussion of our assay was recently published in Haram et al 2022 PMID: 36373947. For the cytotoxicity assays, we use the hemolytic activity. Suppose from Table 1, the toxin stock is 1.5 x10^5 HU/mL. Then to prepare a 2x working toxin stock, we dilute the toxin to 4 x10^3 HU/mL (this is a 1 in 37.5 dilution). To get the range of concentrations used in the dose response curve, we perform a 2-fold serial dilution. Finally we mix equal volumes of toxin and cells, giving us the final 1x toxin activity (2 x10^3 HU/mL for the highest concentration in this example).

      * o Therefore, an EC50 haemolytic curve showing the activities for all toxins would greatly facilitate in understanding the derivation of values for table 1.*

      The hemolytic unit already incorporates the EC50 hemolytic curve. 1 HU is the EC50 of the toxin in the human RBCs.

      * - Flow cytometry assay: What is meant by gating out the debris? And would debris also contribute to the count in dead cells? *

      We illustrate our gating strategy in Fig S1. The debris falls in the front left corner of the plot, and includes electronic noise, non-cellular debris and cellular fragments. Since one cell could give rise to multiple pieces of debris, we exclude the debris from analysis.

      * o What was added as the high PI control? *

      In Fig S1A, the high dose of toxin was used for maximal killing. In our cell populations, there is a low level (2-5%) of dead cells that serve as a control for PI staining. In the past, we’ve used 0.01% triton to validate permeabilization of the cells. We have also compared PI uptake with MTT assays (Keyel et al 2011, Ray et al 2018) to confirm that the PIhigh cells are dead.

      *Elaborating reviewer #2's comment 7 regarding the addition of EDTA : with respect to measuring the binding if fluorescently labelled aerolysin, how can the authors differentiate between full functional pores versus prepores/incomplete pores? *

      This requires electron microscopy, which is the beyond the scope of our current study. However, prior work and Fig 2D show that aerolysin forms pores without the need for Ca2+ (see next point).

      How else can the authors validate whether aerolysin remains functional in the presence of EDTA?

      Prior structural studies suggest that Ca2+ is not required for aerolysin pore formation. For example, Iacovache et al (2011) PMC3136475 induce oligomerization with low salt and pH 2+. Cryo-EM from the same group (Iacovache et al 2016 PMID: 27405240), showed pore formation under similar conditions.

      In Fig S3, aerolysin kills in the presence of EGTA at higher concentrations, suggesting that it can form pores when EGTA is present. Also, in Fig 2D, we used Tyrode’s buffer, which was made without Ca2+ or EGTA. We added the indicated amounts of Ca2+ in, and observed a reduction in lysis at low [Ca2+]. This argues against EGTA interfering with toxin oligomerization/pore formation because EGTA was not present, and the toxin still failed to kill.

      We have updated the manuscript (lines 203-205) to emphasize this point.

      Significance

      *While the work has investigated in-depth cellular resistance mechanisms, the significance and benefits of this study are unclear. For example, the authors have used different human cell lines to dissect how these cells are affected by different pores but have not stated the significance and potential benefit of studying these cell lines. Further elaboration in this aspect may increase the relevance of the study, to an audience who is interested in the field of infection and disease. *

      We have updated our significance to better convey our advance, which is explained on lines 99-102, 128, 519-525. We also added benefits of testing the cell lines chosen on lines 167-168, and 277-278. We plan to add muscle cells to address the dysferlin points, which has relevance to necrotizing soft-tissue infections.

      Description of analyses that authors prefer not to carry out

      Not applicable

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Summary

      This body of work by Thapa & Keyel explores the differences in cellular resistance mechanisms between two different pore families (aerolysin versus CDCs). Herein, the authors were able to elucidate the toxin activities across a variety of different nucleated cells, using the haemolytic assay as a reference for normalising activity. Their findings revealed that, in general, aerolysins were relatively more potent than CDCs at damaging certain nucleated cell lines. Furthermore, the authors performed an exploration of different resistance mechanisms, including MEK-dependent repair, annexins, and patch repair by dysfurlin. The work provides some supporting evidence that patch repair is the main mechanism that cells deploy to prevent aerolysin-mediated cytotoxicity. Overall, the amount of work that was put in to craft the manuscript was impressive and the manuscript showed potential prospects in further investigating 1) mode of aerolysin killing in nucleated cells and 2) the role of patch repair and function of dysferlin in cellular resistance against aerolysin.

      Major comments

      In the first instance, the authors use a method of assaying the specific lytic activity of aerolysin in comparison to a number of different CDCs. Whilst it is acknowledged that these methods have been published in peer-review papers previously (e.g. Ray et al., Toxins, 2018), it would be great to have more information of how the specific activity is derived. Currently there is a convoluted method that makes a number of assumptions such as, but not limited to, 1) the number of dead cells measured in the FACS experiments is proportional to the activity of the different classes of PFPs however the authors do not show how they account for PFPs leading to loss of cells into debris which would involve a total cell count and 2) how the inflection or linear point is identified on individual experiments (e.g. Supp. Fig. 1B, 2A, 2B, 3A, 3B to name a few) and how reliable these points are (e.g showing the data points with model sigmoidal (?) curve and corresponding R values).

      Furthermore, the batch-to-batch variability of protein samples presented in table 1 may be an issue where inactive but folded protein can affect the formation of homo-oligomer pores so more effort to reduce the effects of batch variation would be integral to the foundation of this paper. Given that aerolysin has a very different action on cells then this new characterisation should be provided regardless of what has been previously published by the authors on the activity of CDCs on the cells.

      • Can the authors provide the raw data for the total FACS observations (scatterplot for all events) and show that there is no significant loss of cells? Or at least there is accountability of the cells?
      • Can the authors provide more information about how the linear regression on Supp. Fig. 1B and other experiments showing the model sigmoidal curve performed such that this work is more reproducible?

      The SEMs of some data points (specific lysis LC50 scatterplots, for e.g. Fig. 2C, 4A, 4C, 8A and fMAX plots, for e.g. Fig. 3B) may not be apparently representative of the skew (e.g. and individual values (including outliers). A clarification of the statistical analysis behind the results may benefit in a clearer understanding of how the SEMs were calculated and presented in the main figures. Also, further elaboration on the meaning of the lettering in the scatterplots (denoted as a, b, c etc.) across the main figures may help improve the interpretation of the data.

      Secondly, the authors present interesting results on the significance of Ca2+ on aerolysin's mechanism behind lytic activity and introduces dysfurlin-mediated patch repair as the primary cellular resistance mechanism against aerolysin mediated lysis. Results from Figure 2-4, indicate that extracellular Ca2+ plays a role in aerolysin's function and cell lysis (aerolysin triggers influx of extracellular Ca2+). However, the results presented in figure 8 suggest an impairment of dysferlin translocation from the cytosol to the plasma membrane upon removal of extracellular Ca2+. If this were the case, wouldn't dysferlin impairment sensitise cells to aerolysin? Thus, in these sets of experiments it seems that Ca2+ is a confounding factor.

      • Can the authors further elaborate on how the function of dysferlin in protecting cells against aerolysin contrasts to how aerolysin kills cells?

      Finally, it is also interesting to see that cells deploy different resistance mechanisms between different families of pores. In saying that, the usage of CDCs seems to be inconsistent between each set of results. For example, intermedilysin (ILY) was used in the siRNA knockdown experiments but not in others such as Ca2+ influx assays, while PFO was only used for the initial set of results. A comment on this would benefit in understanding the rationale for selecting certain CDCs for each set of experiments.

      Minor comments

      Results

      (Nucleated cells are more sensitive to aerolysin and CDCs)

      • A statement of the EC50 values of aerolysin and CDCs from the haemolytic assays would be beneficial to compare activities between the two pores.
      • Figure 1A: As stated in the introduction, pro-aerolysin exists as a precursor that is functionally inactive unless activated by trypsin, furin or potentially other proteases. It would benefit the reader if an explicit statement were made about this activity and how it may come about in HeLa and 3T3 cells. Why is pro-aerolysin not shown in the Casp 1/11-/- BMDM cells?
      • Figure 1C: It was stated that "Casp 1/11 -/- Mo were ~100 fold more sensitive to pro-aerolysin and aerolysin compared to PFO and SLO" but did not show the activity for pro-aerolysin in these cells.
      • Supp fig 1E: Shouldn't 24 hr incubation of aerolysin to HeLa cells result in 100% specific lysis?

      (Delayed calcium flux kills aerolysin-challenged cells)

      • What is the intracellular concentration of K+ normally in cells? Similarly, what is the intracellular concentration of Ca2+?
      • Figure 2C: Based on the description in the methods and results, both buffers are supplemented with 2 mM Ca2+ but one buffer (RPMI) shows more killing with SLO and ILY. Does this mean that both buffers contain 2 mM CaCl2? If so, what are the other potential reasons why one buffer enabled greater potency in CDCs?
      • Figure 3H: The data for aerolysin (WT) would greatly benefit for comparison to the inactive mutant (and indicate the sustained Ca2+ increase).
      • Supplementary Video V1: The addition of Triton X-100 permeabilises cells; however, this wasn't evident in (A).
      • Video V2: Similar to previous comment on Supplementary Video V1 (for B).

      (Calcium influx does not activate MEK-dependent repair)

      • Figure 4A: Effective ionic concentration inside and outside cell is increased (if intracellular Ca2+ becomes chelated); therefore, Ca2+ may enter the cell by passive diffusion or transport by other intrinsic Ca2+ channels.

      (Caveolar endocytosis does not protect cells from aerolysin) - Figure 5C: What is the purpose of using HeLa cells as a control? - "..with Alexa Fluor 647 conjugated pro-aerolysin K244C" - this should be introduced earlier as it was initially mentioned in Supp. Figure 3C. - Murine fibroblasts were used earlier (Figure 1). Following from this result (where the WT can be used as a positive control), can MEFs be used instead of adipocytes to see whether caveolar endocytosis plays any role in cellular resistance? - Further comment on the increased resistance of K5 knockout would benefit on the mechanism of aerolysin-mediated cytolysis.

      (Annexins minimally resist aerolysin)

      • Supplementary video V3 - it seems that annexin A6 is recruited to the membrane, to a greater extent (and also quicker) than SLO. This suggests that annexin recruitment is a cellular response against aerolysin challenge. o SLO also shows A6 recruitment (arrows pointed). However, supplementary figure 6B does not clearly illustrate this.
      • As annexin A1 is sensitive to calcium, further comment on the significance of intracellular/extracellular calcium in annexin A1 recruitment and aerolysin challenge would explain observations in Figure 4A.

      (Patch repair protects cells from aerolysin)

      • Supplementary video V4 - the intensity decreases for the inactive mutant; is this due to lysis?
      • The next paragraph sounds like a contradiction: "GFP-dysferlin localized to the plasma membrane and vesicles independently of extracellular Ca2+ (Fig 8C D, Video V5) o Followed by "To study the Ca2+ dependency of dysferlin, we removed extracellular Ca2+ with 2 mM EGTA and challenged with sublytic toxin doses...found less depletion of dysferlin from cytosol".

      (Methods)

      • Table 1: The values presented in the methods section are, overall, confusing and require clarification.
        • 10-fold difference in SLO and PFO WT - do the authors think this might change the interpretation between different figures?
        • Understood how the haemolytic activity was calculated (referred to work in 2012), but how was the haemolytic unit originally derived?
        • How were these values (from table 1) derived to toxin concentrations used for killing nucleated cells?
        • Therefore, an EC50 haemolytic curve showing the activities for all toxins would greatly facilitate in understanding the derivation of values for table 1.
      • Flow cytometry assay: What is meant by gating out the debris? And would debris also contribute to the count in dead cells?
        • What was added as the high PI control?

      Referees cross-commenting

      Elaborating reviewer #2's comment 7 regarding the addition of EDTA : with respect to measuring the binding if fluorescently labelled aerolysin, how can the authors differentiate between full functional pores versus prepores/incomplete pores? How else can the authors validate whether aerolysin remains functional in the presence of EDTA?

      Significance

      The work presents a foundation to further investigate into the mechanism of aerolysin function, following the discovery of the role of extracellular Ca2+ in its activity. As aforementioned, the role of dysferlin in resisting aerolysin also has potential, but the limitations of this work were discussed including the absence of performing a dysferlin knockout, although performing this experiment may help to strengthen the current finding.

      While the work has investigated in-depth cellular resistance mechanisms, the significance and benefits of this study are unclear. For example, the authors have used different human cell lines to dissect how these cells are affected by different pores but have not stated the significance and potential benefit of studying these cell lines. Further elaboration in this aspect may increase the relevance of the study, to an audience who is interested in the field of infection and disease.

      Section for special notes to the editor:

      My major area of expertise and contribution to this paper is in the analysis and interpretation of activity (lytic) assays.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      The current study explored an interesting question of aerolysin pore repair mechanism. An unusual feature of aerolysin is that unlike many other pore-forming proteins its pores remain "open" over a longer period of time, and this affects ion homeostasis influencing cell death and inflammatory response. Eventually, aerolysin pores are repaired, but what governs this process remains unknown.

      In my opinion, the paper has several unresolved issues, some of which the authors mentioned in their discussion.

      1. The effect of dysferlin overexpression does not indicate that patch repair is a protective mechanism or that dysferlin plays a significant role in aerolysin resistance. The authors should knock out dysferlin and assess cell resistance to lysis.
      2. ESCRT complex was shown to play a role in plasma membrane repair following mechanical damage or perforin treatment of cells (Jimenez 2014, and Ritter, 2022). Whether ESCRT is important in aerolysin pore repair can be assessed by knocking out the Chmp4b gene or overexpressing dominant-negative mutant of VPS4a, E228Q.
      3. I find the optimisation of lysin concentrations and data presentation quite confusing. I eventually understood, what was done, but I feel that the authors should be able to transform the data and plots so these are more accessible to a reader, eg a simple dose/time-response curves would be very helpful in that respect. For example, in Figure S1E, why does aerolysin appear to be less cytotoxic after 24 hrs than after 1 hr. In principle, I would expect to observe an additive effect, i.e. cell death at 1, 3, 6, 12, and 24 hrs should add to 100%; however, if 100% cells die at 500HU/ml, how can more cells die after 24hrs? Or am I missing something in the experimental design/data presentation? I also wonder whether using haemolytic units is appropriate (it may well be, if justified), given that the toxins used here have various membrane-binding properties. Wouldn't it make more sense to compare the cytotoxicity using nucleated cells?
      4. The authors use "sublytic" concentrations of aerolysin (64HU) throughout most of the paper, but according to Figure S1C, 50% cells died at that concentration after 1hr, suggesting that when the cells were investigated over a shorter period of time, they were already dying - it's almost like the cells had life support turned off, but still being investigated as though they survived aerolysin treatment. This needs to be clarified or reassessed.
      5. What effect does the addition of 150mM KCl have on the plasma membrane, trafficking/repair - wouldn't the plasma membrane be depolarised? There were a number of papers by John Cidlowski in mid 2000s, where his team explored the effect of potassium supplementation on apoptosis - this may be worth exploring.
      6. Figure 3 and accompanied text: it would be more informative to show all the data rather than breaking it down to <5, 5-45 and >45 min. In my view, <5 min is an acute death due to lysis, where the toxins overcame all the protective mechanisms (membrane repair). If anything, I would dismiss that acute cell death altogether, and focused on the cells that survived the initial onslaught.
      7. I am curious whether EGTA diffuses into the cytosol through aerolysin pores. If so, then unlike BAPTA-am it would affect Ca inside and outside the cell. Are the authors confident that in the absence of extracellular calcium (EGTA treatment), aerolysin formed the pores at all? Have they looked, for example, at intracellular Na/K, or have any other evidence of membrane disruption?
      8. Figure 6 (and some other): I find the designation of statistical significance (a-f) quite confusing, as it is unclear which comparisons are statistically different. Looking at Figure S5, there was no difference between the effect of Annexin depletion on the toxicity of the three lysins.

      Referees cross-commenting

      I agree with the critique raised by the other two reviewers.

      I am also happy to revise the time required to complete revisions to 3-6 months, but feel that even this may be optimistic considering substantial technical problems raised by Reviewer 1.

      Significance

      The paper attempts to address an interesting question of aerolysin pore repair, and it is interesting from the perspective of a potential difference between various pore-forming proteins.

      The study will be potentially interesting to a broad audience of biochemists/cell biologists and microbiologists working in the field of pore-forming proteins/virulence factors.

      My expertise is the biochemistry and cell biology of pore-forming proteins.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      Thapa et al studied cellular mechanisms of membrane repair following pore forming toxin insult, namely aerolysin and CDCs. Out of four mechanisms tested, they show that under their conditions patch repair is the only mechanism able to counter aerolysin injury. The study is interesting but raises some concerns.

      Major:

      The authors conclusions contradict established results, which they cite. Yet experimental conditions are not similar in two ways: toxin concentration-wise and toxin treatment duration-wise. While we appreciate the efforts of the authors to standardize the concentration of toxins used based on hemolytic units, we note that the concentrations used are very much higher than in the other studies cited. Indeed, based on table 1, materials and methods, and the various experiments, aerolysin has a LC50 of approximately 200 HU/ml, which corresponds to about 2 ug/ml. This is approximately 200x more concentrated than for example in Gonzalez et al 2011 and Larpin et al. 2021. It makes the validity of direct comparison with those studies questionable. We noticed that the authors activate pro-aerolysin at high concentration (in the range of 1 to 5 mg/ml) and at room temperature. In our experience, under these concentration, activation leads to immediate oligomerization and massive precipitation. The final concentration of active toxin is thus unknown. The authors keep their cells in toxin-containing medium for the whole duration of the experiments, typically 45 minutes. This is in stark contrast with 45 seconds to 3 minutes transient exposure to toxin in Huffman et al 2004.

      The authors do not report binding and oligomerization assays of the toxins. The only figure showing a western blot (fig. 7) is of low quality and shows unexpected observations. Aerolysin Y221G mutant is expected to bind and oligomerize. Yet, no band is present at about 250 kDa (expected oligomer) or at about 47 kDa (monomer). In addition, in aerolysin lanes (1 and 2) the oligomer is saturated, seems to be covering three lanes, indicating a possible spill-over.

      Finally, while the patch repair hypothesis is interesting, it is unclear why the authors decided to overexpress dysferlin in cell lines that normally do not express it. Sure, there is a repair phenotype but this phenotype is artificially introduced. Dysferlin is not expressed at all in HeLa cells. Furthermore, dysferlin is not expressed in epithelial cells, which are the prime target of aerolysin. Why then focus on this protein? In order to show that patch repair is indeed protecting cells against aerolysin, the authors should disrupt patch repair of the cells under study and observe and increased toxicity.

      Minor:

      The graphic legends should be boxed out to be clearly separated from the data. In Figure 4A, it is mixed up with the data.

      Some western blots are saturated, e.g. B-actin in figure 4B. Full blots should be provided.

      In the methods, aerolysin sublytic dose for HeLa cells is specified at 62 HU/ml. In figure 5C and D, 31 HU/ml kills more than 50% of HeLa cells. This is not compatible.

      Figure 2A and B have quite different LC50 for starting conditions ({plus minus} 200 HU/ml in A, 600-700 HU/ml in B). Why is it so different? Y-axis has a linear scale in A and a logarithmic scale in B. It would make comparison easier to have the same scale in both panels.

      The letters detonating statistically significant groups are sometimes unclear. For example in Figure 1A and B, PFO belongs to group a and b simultaneously. What does this mean?

      In Figure 8, aerolysin hat a LC50 in cells overexpressing GFP-Dysferin of approximately 1700 HU/ml in A and of approximately 400 HU/ml in B. Why is it so different?

      In Figure S1, it is unclear what the plots « all events » vs « single cells » mean.

      In the discussion, the authors write « First, survival did not correlate with overexpression, which would be expected if dysferlin acted as Ca2+ sink ». What is meant? GFP-dysferlin overexpression does correlate with survival in Figure 1A.

      Referees cross-commenting

      I notice quite a number of overlapping points between my comments and those of the other reviewers. In particular concerning the varying definition of sublytic concentrations and the need of a dysferlin-KO.

      Significance

      General assessment: The study strength lies in the several possible protection mechanisms that are tested. The weaknesses lie in the contradictions of the results reported here with established mechanisms, and in the statement that a cellular process that has been artificially introduced in the experimental system is the cellular protection mechanism against aerolysin attack. In order to prove that this process is a bona fide protection mechanism, the authors should show that it is present without the need of overexpressing a protein that is not expressed at all either in the used cell line (HeLa), or in the natural cellular target of aerolysin (epithelial cells). The significance of the proposed protection mechanism is therefore questionable.

      Advance: The study contradicts previously established results but the experimental conditions used here are quite different to those used in the earlier studies, which makes the comparison quite difficult. As such it does not really fill a gap.

      Audience: The study will be of interest of specialized audience.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

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

      Summary:

      The submitted manuscript is comparing the effect of individual chaperones and heat-resistant obscure (Hero) proteins on the overall folding of the TDP-43 LCD-domain and its relation to aggregation propensity. Therefore, the authors apply smFRET in order to deduce eventual morphological changes of the LCD-domain from FRET efficiencies. The authors observe that the LCD domain extends its structure upon binding of chaperone/Hero proteins whereas it is collapsed in the absence of those. Furthermore, immunoblotting of filter trap assays indicate that overexpression of chaperones and Hero proteins reduce aggregation of TDP-43 in vivo. Both, the morphological effects on the LCD-domain and the aggregation propensity are significantly enhanced for the TDP-43 A315T mutant. Moreover, the authors tested a charge depleted Hero protein version with reduced "chaperone-like" behaviour. Therefore, the authors conclude that the binding or chaperone activity of the Hero protein is based on its residue specific charges. Finally, the authors conclude that Hero proteins can act similar to chaperones in order to keep protein homeostasis under stress conditions.

      We thank the Reviewer for their insightful evaluation of our study.

      Major comments:

      The similar effect of chaperones and Hero proteins on the morphology of TDP-43 found by the authors is intriguing and the applied experimental procedures seem well described and conducted.

      However, the assumption of the authors that a change in morphology of the LCD-domain by the chaperones and Hero proteins is directly connected to the reduction of TDP-43 aggregation is not entirely clear. Whether an overexpression of individual chaperones and Hero proteins has a direct effect on TDP-43 aggregation cannot be tested in vivo, only. It cannot be excluded that inside the cell the here tested chaperones and Hero proteins control intermediate processes or work as co-factors for other proteins involved in protein homeostasis rather directly influencing the aggregation of TDP-43. Therefore, I recommend in vitro aggregation experiments, using ThT signal as a readout. By adding chaperones, Hero proteins and a negative (BSA or others) control individually, a direct effect on TDP-43 aggregation could be concluded. Those experiments have been extensively used in the field and are quick and straightforward to handle.

      As the Reviewer explains, indirect effects on TDP-43 aggregation in cells may be accounted for by conducting aggregation experiments in vitro, with recombinant proteins. We are currently designing such experiments based on a previously described full-length recombinant TDP-43 with a TEV-cleavable MBP tag (Wang 2018 EMBO J). This can be incubated with Hero/DNAJA2/Control, and aggregation induced by cleavage of the tag, after which aggregation can be measured via filter trap similar to the method described in our work. We will include these results in our revised manuscript.

      We thank the Reviewer for their advice. While we note that it is controversial whether ThT binds to aggregates formed from full-length TDP-43 (used in all our assays in the current manuscript), it is reasonable to apply this assay to the LCD fragment as in the paper referenced by the Reviewer below (Lu 2022 Nat Cell Biol). Such an assay is also a reasonable method for confirming effects of Hero protein and DNAJA2 in vitro, and we can conduct this assay as a back-up if the above does not work.

      In addition, focusing on the LCD-domain as a main driver for TDP-43 aggregation is limiting this study. In particular, recent studies [1] indicate that the RRM1 and RRM2 sites of TDP-43 have a major impact on TDP-43 gelation and maturation to solid aggregates. Unfortunately, those sites have not been studied in this manuscript.

      We thank the Reviewer for their insight. While we are keen to investigate the impact of other regions on the aggregation of TDP-43 in the future, we chose to focus on the LCD in our current study because our smFRET assay is particularly suitable to monitor the dynamic conformational nature of this flexible, unfolded region.

      However, we agree with the Reviewer that it is possible the RRMs have an effect on the activities of Hero11 and DNAJA2. We will create constructs for the RRM-depleted variant, TDP43ΔRRM1&2, and RNA-binding deficient variant, TDP435FL for use in our cell-based assay. This will allow us to investigate how this domain influences the effects of Hero and DNAJA2, and we will include this in our revised manuscript.

      As an optional alternative for using Hero11KR->G could be the alteration of buffer conditions and using higher number of salts to promote charge screening. It would be of interest whether the results with the Hero11KR->G could be reproduced with wild type Hero11.

      We will perform our assays with Hero11 in high salt conditions for charge screening. While we agree that it may be a great alternative experiment, we note that changing the salt concentration may directly affect the LCD conformation, possibly complicating interpretation of results.

      [1] Lu et al. Nat Cell Biol;24(9):1378-1393 (2022)

      Minor comments:

      Overall, the text is clearly written, and the figures are appropriate.

      Whether the activity of individual chaperones or Hero proteins on TDP-43 aggregation "may result in the overall fitness of the cell" or "reinforcing the conformational health of the proteome" is disputable without knowing how the overexpression of certain chaperones or Hero proteins alter the formation of toxic TDP-43 oligomers.

      We thank the Reviewer for their balanced critique. We will remove or weaken this point regarding how Hero proteins "may result in the overall fitness of the cell" or may be "reinforcing the conformational health of the proteome" from the discussion.

      Reviewer #1 (Significance (Required)):

      Studying the mechanistic effects of chaperones on aggregating proteins is of major interest for the field in order to understand aging related disbalance of protein homeostasis and the progression of neurological decline, such as seen for amyotrophic lateral sclerosis (ALS). Furthermore, finding homolog proteins, also being able to inhibit protein aggregation, can help to understand overall mechanisms of protein aggregation and processes preventing such fatal behaviour. However, the technique used in this manuscript are not very novel and have been used numerously times before. smFRET is a common technique to look at protein folding/unfolding and is used frequently as a molecular ruler. The manuscript is of interest for the field of protein aggregation and folding, smFRET and neurodegeneration.

      My expertise lies in the field of protein aggregation and inhibition due to chaperones, measuring molecular interactions and neurodegenerative diseases.

      We greatly appreciate the Reviewer’s expert opinion on our work. As the Reviewer explains, we believe our work will contribute to the fields of protein aggregation and folding, smFRET and neurodegeneration. While the smFRET method may not be novel on its own, to our knowledge this is the first observation of the TDP-43 LCD, with the effect of a pathogenic mutation, at the single-molecule level. In fact, the production, dye-labeling and isolation of individual molecules is extremely challenging for TDP-43. This was made possible by our technical advances using genetic code expansion to site-specifically introduce an unnatural amino acid in TDP-43, purifying and labeling the TDP-43 from HEK cells, and isolation on glass slides.

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

      In this manuscript, the authors build on their findings (Tsuboyama 2020) that electrostatically charged IDPs (Heros) can protect proteins from denaturation and aggregation. In their previous work, they demonstrate that these Hero proteins could decrease the fraction of insoluble GFP-TDP43∆NLS in mammalian cell lines and that this mode of action was related to the electrostatic charge of the proteins and not sequence dependent. Although this protective mode of action appears to be similar to that of canonical chaperones, it is unclear how the Hero proteins compare. In this study, the authors compare Hero11 to a panel of canonical chaperones in their cell-based assays and show that it prevents aggregation in a comparable way to DNJA2. It appears that Hero11 decreases the GFP-TDP43∆NLS aggregates better than some other chaperones. They then utilise their expertise in smFRET analysis (Tsuboyama, 2018) to compare what effect DNJA2 and Hero11 (along with Hero11KR-->G (non-charged control)) have on the dynamic structures of the GFP-TDP43∆NLS (labelled with complementary fluorophores in the LCD domain). Based on analysis of the WT GFP-TDP43∆NLS and the A315T GFP-TDP43∆NLS, the authors suggest that the presence of Hero11 and DNJA2 maintain the LCD-domain of TDP43 in an extended conformation and that by doing so, aggregation can be prevented (as assessed in the cell-based assay).

      Despite finding the results very interesting, I feel that the study is preliminary and the conclusions drawn are not fully substantiated by the presented experimental work. Many questions need addressing to validate these findings and conclusions (please see more in the "Significance" section). I have tried to list the main concerns below.

      We thank the Reviewer for their detailed and critical assessment of our current study.

      Questions/concerns:

      Authors used double transient transfections but have not shown quantification of protein levels of the chaperones versus TDP43 - western blots to confirm proper expression (and levels) of the chaperones/Hero protein is crucial without it, we cannot assume that the differences in TDP-43 aggregation are a result of effective chaperoning or due to a lack of expression of any of the chaperone proteins, or high expression of others.

      We agree with the Reviewer that this is an important and straightforward validation experiment. We will perform the Western Blotting to confirm the proper and comparable expression of the chaperones/Hero proteins.

      Authors used quite a high BSA concentration in the smFRET work; it would be useful to see what the TDP43 smFRET trace looks like without BSA incubation (to ensure it is not causing some effect). Also, is there a concentration dependence? The Authors mention they are unable to identify a Hero/TDP43 complex; but if the amount of Hero protein is high (given that it is single molecules tethered), the change in compaction may not relate to the levels/ratios found in the cells (where changes to aggregation are occurring). have the authors considered whether positively charged polymers (poly-Lys) have any impact on the TDP-43 smFRET distribution? Given that the smFRET trace is so heterogeneous, to understand what is happening here would require the comparison of more than 2 variants.

      As the Reviewer suggests, we will include additional smFRET experiments in our revision.

      First, we will perform the smFRET experiment of the TDP-43 alone in the PBS buffer. However, we would like to clarify the reason we used BSA incubation for comparison in the current experiment is to account for the possibility of non-specific macromolecular crowding effects on the conformation of the LCD (an effect reported for IDPs in general, for example in Banks 2018 Biophys. J.); we expected that it would be fair to compare Hero11 against another protein, rather than buffer alone. As the Reviewer suggested, we can also perform the same experiments at lower concentrations of Hero11 and DNAJA2, including equimolar concentrations (as suggested below). Moreover, we can also test poly-K peptides for comparison.

      Although the A315T variant has a very distinct smFRET profile, it is clear that the effects of Hero11KR-->G (that is proposed to have no effect on aggregation or on the smFRET of WT TDP43) has a clear impact on A315T. Why is this?

      We thank the Reviewer for raising this interesting point. We envision that the observed effect is due to weak interactions between the LCD domain of TDP-43 and Hero11KR->G; even without K and R, there many other functional amino acids that are fully accessible due to the extremely disordered nature of the protein. The effect is easier to be observed with the A315T mutant, compared to the WT TDP-43, presumably because the mutant tends to take more compact conformations on its own. Nonetheless, unlike WT Hero11, Hero11KR->G fails to accumulate the very extended form of the LCD (FRET signal of ~0; please see below for the explanation of this value), which appears to be associated with suppression of aggregation. We will include these in our discussion.

      The LCD region is prone to PMTs - as the tethered protein is taken from expression in mammalian cells, how can the authors be sure that it has no PMTs? Although a clear difference is observed between WT and A315T in terms of "compactness" of the LCD domain, we cannot assume that the effect of DNAJ2 and Hero11 are the same - in fact, the Hero11 KR-->G control for the A315T is clearly different from the negative control (BSA) and the effect that was seen in WT. As the LCD domain is well-known to be the site of post-translational modifications (ie. Phosphorylation - this would have an effect on an electrostatic Hero11), could the effects be related to changes in PMTs as well?

      We thank the Reviewer for their insight. We would like to clarify that we make no assumption that our dye-labeled TDP-43 is free of post-translational modifications. Indeed, the fact that it is derived from HEK293 cells suggests it should have post-translational modifications relevant to humans and may be even considered an advantage of our method. (Most structural methods require purification of a large amount of protein, often only possible through recombinant expression in E. coli, thus lacking human-relevant PTMs.) As the Reviewer points out, the LCD is known to have many phosphorylation sites, which may help explain how the positively charged Hero11 interacts with it. Thus, we will perform mass spectrometry of TDP-43 and the A315T variant expressed in HEK cells to identify what post-translational modifications are present.

      The authors mention other studies on DNJA proteins on TDP-43; is the mechanism by which they suppress aggregation known? If the authors want to compare the unknown effects of Hero11, it would be useful to know what DNJ2A is doing, otherwise, the results are still not conclusive, only that "change is similar" in two experiments. What is known about DNJ2A interactions with TDP-43? Did the authors do any pulldown assays to detect a complex in cellulo?

      While previous studies have identified various DNAJ (specifically J-domain protein B-subfamily) proteins that suppress aggregation of overexpressed TDP-43, not much is known of this specific interaction (Udan-Johns 2014 Hum Mol Genet, Chen 2016 Brain, Park 2017 PLOS Genet). To address the Reviewer’s questions, we will include experiments characterizing the effects of DNAJA2 on TDP-43. We will perform colocalization experiments, explaining effects of DNAJA2 and Hero11 on TDP-43 in the cell. As explained below, we will also perform Pulse Shape Analysis (PulSA), a flow cytometry-based method that can be used to study protein localization patterns in cell, which will also provide insight into the effects on the distribution of TDP-43 in cells. We can also perform co-IP of TDP-43 to detect if there is a detectable, stable complex with DNAJA2 and/or Hero11. Together, these will clarify the similarities and differences between DNAJA2 and Hero11.

      It is unclear how the findings of the smFRET relate to structural understanding of the LCD-domain of TDP43 (i.e. NMR studies?); is it known whether PTMs are more prominent with the A315T variant as this may explain it's more compact nature? As well, putative helical structure in the LCD domain may lend to the changes in compaction.

      The Reviewer brings up an interesting and careful discussion. Currently, it is unknown if PTMs actually cause more compaction, or if they are more prominent in the A315T variant, but we will perform mass spectrometry to detect PTMs.

      As the Reviewer mentions, it would be very interesting to compare our smFRET results to other studies of specific LCD structures. However, it is not trivial to deduce lengths (and structure) from smFRET data as various other factors, for example, dye orientation and local chemical environment, may affect FRET efficiency. Nonetheless, we can still cautiously provide a discussion of how our FRET results compare with previous studies.

      For the dye pair used in our study, Cy3 and ATTO647N, the low/no FRET signals promoted by DNAJA2 and Hero11 correspond to a range of end-to-end distances of 6.9 nm to 10.2 nm (FRET signals of 0.1 to 0.01, respectively). Assuming that the LCD behaves like a ~140 amino acid worm-like chain (WLC) with persistence length (Lp) = 0.8 nm, we expect a mean end-to-end distance of 7.35 nm. Thus, the low FRET peak can be well explained by promotion of an extended WLC behavior of the LCD by DNAJA2 and Hero11. On the other hand, the FRET peaks of WT LCD and the A315T mutant (in the absence of Hero11 or DNAJA2) correspond to ~4 and ~3.3 nm, respectively. We will include a careful discussion of how our results relate to known structural understanding of the LCD in the revised discussion.

      It is unclear how there can be such a prominent FRET ~0 peak and in fact negative values.

      We regret that we did not clearly explain this in the manuscript. Negative values arise when applying correction factors from the alternating laser scheme (ALEX) to FRET signals. FRET efficiency, E, is the ratio of acceptor signal intensity, IA, over the total signal intensity, ID+IA, (with the application of a correction factor, γ, but this doesn’t affect the negative values and won’t be discussed further here) and is given by the equation: E=IA/(γ×ID+IA). However, due to leakage of the donor signal into the acceptor channel and direct excitation of the acceptor dye by the donor laser, raw IA values, IA,raw, are erroneously higher than in reality. For example, the ~0 FRET peaks in question appear to be around 0.1–0.2 before correction. These are accounted for by applying the respective correction factors, Dleakage and Adirect, through the equation: IA=IA,rawDleakage×IDAdirect×IAA. (IAA is the acceptor signal during excitation of the acceptor dye.) These two correction factors are determined by observing the traces and choosing the mean values using iSMS software (2015 Preus Nat Methods) and applied uniformly to all traces in an experiment. When IA is especially low, such as when FRET is almost 0, the magnitude of the correction factor terms may be larger than IA,raw, resulting in negative values. This does not mean that values less than 0 are invalid, but merely that they have been overcompensated in the error application. For the dye pair in our study, FRET efficiencies less than 0.1 correspond to distances greater than 6.9 nm, meaning peaks around zero represent LCD behaviors with end-to-end distances greater than around 7 nm. Please also note that kernel density estimation often gives distributions with values beyond the (0,1) range just because of how these plots are constructed. This will be added to the methods in the revised manuscript.

      Conclusion is that Hero11 and DNJA2 maintain the TDP43 LCD-domain (soluble protein) in an extended form and that this is linked with the decrease in aggregates found in the cell; however, with the cell-based assay, no analysis to quantify the expression levels of the TDP43 and the chaperones/Hero are presented, and more importantly, no analysis on the complementary soluble fraction (to the filter assay) has been done to show that indeed, these biomolecules maintain the proteins in a soluble form. It is possible that the TDP-43 is being degraded?

      As described above, we plan to perform Western Blotting to examine the expression levels of these proteins. To address the concerns about solubility, we will perform Pulse Shape Analysis (PulSA) to quantitatively measure the expression and soluble/aggregated distribution GFP-tagged TDP43 in HEK293T cells. Measuring the soluble diffuse signals and the punctate aggregate signals will also tell us if there are differences in how GFP-TDP43 is aggregated between Hero11, DNAJA2 and controls. In addition, to support results from the FTA, we will provide sedimentation assays, where the soluble and aggregate fraction from cells is separated by centrifugation and analyzed (Krobitsch 2000 PNAS). These will provide information on TDP-43 in the soluble fraction.

      Reviewer #2 (Significance (Required)):

      Contextually, this study has novelty and potential value for basic research. Firstly, understanding the underlying mechanisms by which Hero protein prevent aggregation would be valuable towards understanding the players in protein homeostasis which can be imbalanced with respect to disease. Secondly, the use of smFRET as a tool in understanding the dynamics of TDP-43 and mutational variants can be powerful in defining structural attributes with pathological consequences in ALS. Although this work shows comparisons between the effect of a canonical chaperone (DNJA2) and Hero11 on the dynamics of monomeric protein and the effect on cellular aggregation, proposing a general mechanism on the data from two TDP-43 variants and a cell-based aggregation assay is premature and more experimental evidence is needed to define the critical link that prevents aggregation of TDP-43 within the cell. Mechanistically, the study does not give a lot of additional insight into the mode of action of Hero11 in the process of preventing aggregation (nor does it explain what DNJA2 is doing and therefore how Hero11 compares and contrasts). The proposed "extended versus collapsed" switch is simplistic and doesn't account for the complexity of TDP-43 structural dynamics. To support their proposed mechanism of action, the authors needs to examine TDP-43 mutational variants (specifically disease-related ones) using their smFRET to understand exactly what the "collapsed" and "extended" data is defining before making the leap that this effect is what is preventing aggregation. There are some structural studies about residual structure in this region (via NMR) that should be considered (https://doi.org/10.1016/j.str.2016.07.007). Although the A315T variant has a very distinct smFRET profile, it is clear that the effects of Hero11KR-->G (that is proposed to have no effect on aggregation or on the smFRET of WT TDP43) has a clear impact on A315T. Why is this? Have the authors considered that the LCD domain of TDP43 is prone to post-translational modifications? Is this variant more phosphorylated - a PMT like phosphorylation is surely to have an impact on interactions with Hero proteins as they are positively charged. Given that the protein is expressed in mammalian cells, it is likely that PMTs have occurred (but the authors should analyse for this).

      With regards to the cell-based aggregation assays, the authors again present a simplified relationship - however, a number of control experiments and additional questions arise. It appears that there is less aggregation with co-expression of some chaperones and the Hero11, but what about the soluble fraction? What is the impact of these biomolecules? Is this that it is maintaining soluble protein, enhancing degradation, propagating soluble oligomers? Equally, how do we know that the levels of the chaperones/Heros and the TDP-43 is the same in each cell - these are transient transfections, and no western blots are shown to confirm the levels of the proteins. In fact, the authors state that "co-transfection of HSP70 (HSPA8), HSP90 (HSP90AB1) or HOP all failed to suppress TDP-43 aggregation compared to GST" and mention that this is in contrast to other studies, but could this be a failure to express these in the cell models? Some western blot/lysate analysis is needed. Chaperones often form complexes with their client proteins, is there any evidence of complexes in these cell models (i.e. using immunoprecipitation)?

      We thank the Reviewer for their detailed evaluation and interest in our work. As the Reviewer describes, smFRET is a powerful tool for studying the conformational dynamics of TDP-43, and we hope that this study will contribute to our understanding of how Hero proteins and chaperones prevent aggregation.

      We are also grateful to the Reviewer for their constructive criticism of our current model, and we will revise it accordingly. We completely agree with the Reviewer that there are complex structural dynamics within the LCD that determine aggregation and phase separation behaviors. Our simple model was intended to explain how external factors that suppress aggregation, DNAJA2 and Hero11, could affect the conformation of LCD at the single-molecule level. As discussed above, we were cautious to over-interpret how our FRET observations correlate to specific conformations, leading to this simplistic model. We do not intend for our explanation of “extended versus collapsed” in the model to explain all structural dynamics of the LCD; rather, we wanted to highlight the characteristic low FRET state promoted by DNAJA2 and Hero11. We believe that the experiment plan explained above will address the Reviewer’s concerns in full, and we thank the Reviewer again for helping us to significantly improve our manuscript.

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

      - In a recent study (PLosBiol, 2020) the same authors described an interesting class of proteins they call 'Hero'. Based on their analyses in cultured cells and transgenic Drosophila models the authors concluded that 'Hero' proteins protect against protein instability and aggregation. So far, this class of proteins has not been analyzed by independent groups.

      In the current manuscript, they mainly confirm their own previous finding that Hero 11 prevents There are several concerns about the presented data:

      We thank the Reviewer for their critical comments on our current manuscript.

      - Based on the filter trap assays shown in figure 1 and 3 the authors conclude that DNAJB8 and Hero11 specifically interfere with the aggregation of TDP-43. However, they do not show that the expression levels of TDP-43 are not altered by the co-expression of the different proteins and are comparable in the different samples. In order to make a relevant statement about the anti-aggregation activity of the analyzed proteins, the ratio between soluble and aggregated TDP-43 has to be analyzed.

      We would like to clarify that the Reviewer means DNAJA2, not DNAJB8. Following the Reviewer’s advice, we will perform Western Blotting combined with sedimentation assays, where the soluble and aggregate fraction from cells is separated by centrifugation and analyzed to examine the expression levels. We will also perform colocalization experiments and Pulse Shape Analysis (PulSA), a flow cytometry-based method that can be used to study protein localization patterns in cell, which will provide insight into the anti-aggregation activities.

      - The FRET assays shown in figures 2 and 4 indicate a slightly higher FRET efficiency in the presence of Hero11 and DNAJA2 and Hero11. The authors postulate that is phenomenon is causally linked to the activity of Hero11 to prevent aggregation of TDP-43. First, it remains unclear whether the slight increase is really significant. Second, I could not find any experimental evidence to support the assumption that a more collapse conformation of the TDP-43 LCD measured in single molecule FRET assays, correlates with an increased aggregation tendency of TDP-43.

      We apologize that we are not sure what the Reviewer refers to by “a slightly higher FRET efficiency in the presence of Hero11 and DNAJA2 (and Hero11).” We would like to clarify that, in the presence of Hero11 and DNAJA2, what we observed was a very low (not slightly higher) FRET efficiency of ~0 (Figure 2g and h), suggesting an extended conformation. In contrast, the aggregation-prone A315T variant of TDP-43 shows a very high FRET efficiency of ~0.9 (Figure 4a), which indicates a collapsed conformation.

      A minor comment, if the authors would like to compare the specific activity of different proteins, they should use equal molarities of the different proteins and not equal amounts.

      As the Reviewer suggests, we will include experiments at equal molarities in the revision.

      - For a one-way ANOVA, the response variable residuals have to be normally distributed. With an n = 3 this cannot be tested. Thus, the quantifications of the results shown in figure 1 and 3 are not reliable.

      We thank the Reviewer for their critical comment on the statistical analysis. We would like to clarify that statistically significant differences in aggregation between conditions compared to a control are based on Dunnett’s test. While ANOVA is typically first performed to test for any significant difference in means before performing a post-hoc test, Dunnett’s test is independent and can be performed without ANOVA.

      Following the Reviewer’s advice, we carefully re-examined our assumption of normality for this data. It is reasonable to perform Dunnett’s test on a sample size of n = 3, and it is generally safe to assume that data from three independent experiments will be reasonably normally distributed. In support of this, performing Kolmogorov-Smirnov test on our data in Figure 1 showed none of the groups differ significantly from normal distributions with the respective mean and standard deviation (p-values greater than 0.05). Thus, we believe it is reasonable to assume the data are normally distributed, the residuals normally distributed, and our statistical analyses reliable. This analysis will be included in the revision to support the normality assumption.

      However, even if we did not assume a normal distribution of our data in Figure 1, we still would have obtained statistically significant differences; If we had relied on a Kruskal-Wallis test as a non-parametric equivalent of ANOVA, thus making no assumption of normality, we would have seen p = 0.005176, a value much lower than our significance level of α = 0.05, indicating sufficient evidence that there is a difference in aggregation among these groups.

      - The title is imprecise and overstate the presented data:

      'canonical chaperone' suggest that their results are valid for chaperones in general. However, they only tested DNAJA2 in the single -molecule FRET assay. Moreover, HAPA8, another canonical chaperone, obviously had an opposite effect on TDP-43 aggregation (Fig.1). Similarly, they only tested Hero11. Thus, 'canonical chaperone' has to be replaced by 'DNAJA2' and 'a heat-resistant obscure (Hero) protein' by 'Hero11'. Similarly, the term 'conformational modulation' is not as concise one would one expect for the title of a research paper.

      We would like to clarify that the Reviewer means HSPA8 (not HAPA8). According to the Reviewer’s suggestion, we will change the title to “DNAJA2 and Hero11 mediate similar conformational extension and aggregation suppression of TDP-43”.

      Reviewer #3 (Significance (Required)):

      In a recent study (PLosBiol, 2020) the same authors described an interesting class of proteins they call 'Hero'. Based on their analyses in cultured cells and transgenic Drosophila models the authors concluded that 'Hero' proteins protect against protein instability and aggregation. So far, this class of proteins has not been analyzed by independent groups.

      In the current manuscript, they mainly confirm their own previous finding that Hero 11 prevents aggregation of TDP-43 and present very few new data that would provide new insights. Specifically, only the FRET assays shown in figure 2 and 4 are really new, which, by the way, could easily be shown in one figure.

      We thank the Reviewer for their critical evaluation of our current study. As the Reviewer suggests, we believe our smFRET results provide new insights into how Hero11 and DNAJA2 function. We would like to emphasize that, rather than confirming our previous findings, our current manuscript mainly addresses a critical point that remained unknown in our previous study by investigating the mechanism of how Hero proteins prevent aggregation. Moreover, to our knowledge, this is the first observation of the TDP-43 LCD, with the effect of a pathogenic mutation, at the single-molecule level.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      • In a recent study (PLosBiol, 2020) the same authors described an interesting class of proteins they call 'Hero'. Based on their analyses in cultured cells and transgenic Drosophila models the authors concluded that 'Hero' proteins protect against protein instability and aggregation. So far, this class of proteins has not been analyzed by independent groups.

      In the current manuscript, they mainly confirm their own previous finding that Hero 11 prevents There are several concerns about the presented data: - Based on the filter trap assays shown in figure 1 and 3 the authors conclude that DNAJB8 and Hero11 specifically interfere with the aggregation of TDP-43. However, they do not show that the expression levels of TDP-43 are not altered by the co-expression of the different proteins and are comparable in the different samples. In order to make a relevant statement about the anti-aggregation activity of the analyzed proteins, the ratio between soluble and aggregated TDP-43 has to be analyzed. - The FRET assays shown in figures 2 and 4 indicate a slightly higher FRET efficiency in the presence of Hero11 and DNAJA2 and Hero11. The authors postulate that is phenomenon is causally linked to the activity of Hero11 to prevent aggregation of TDP-43. First, it remains unclear whether the slight increase is really significant. Second, I could not find any experimental evidence to support the assumption that a more collapse conformation of the TDP-43 LCD measured in single molecule FRET assays, correlates with an increased aggregation tendency of TDP-43. A minor comment, if the authors would like to compare the specific activity of different proteins, they should use equal molarities of the different proteins and not equal amounts. - For a one-way ANOVA, the response variable residuals have to be normally distributed. With an n = 3 this cannot be tested. Thus, the quantifications of the results shown in figure 1 and 3 are not reliable. - The title is imprecise and overstate the presented data: 'canonical chaperone' suggest that their results are valid for chaperones in general. However, they only tested DNAJA2 in the single -molecule FRET assay. Moreover, HAPA8, another canonical chaperone, obviously had an opposite effect on TDP-43 aggregation (Fig.1). Similarly, they only tested Hero11. Thus, 'canonical chaperone' has to be replaced by 'DNAJA2' and 'a heat-resistant obscure (Hero) protein' by 'Hero11'. Similarly, the term 'conformational modulation' is not as concise one would one expect for the title of a research paper.

      Significance

      In a recent study (PLosBiol, 2020) the same authors described an interesting class of proteins they call 'Hero'. Based on their analyses in cultured cells and transgenic Drosophila models the authors concluded that 'Hero' proteins protect against protein instability and aggregation. So far, this class of proteins has not been analyzed by independent groups.

      In the current manuscript, they mainly confirm their own previous finding that Hero 11 prevents aggregation of TDP-43 and present very few new data that would provide new insights. Specifically, only the FRET assays shown in figure 2 and 4 are really new, which, by the way, could easily be shown in one figure.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this manuscript, the authors build on their findings (Tsuboyama 2020) that electrostatically charged IDPs (Heros) can protect proteins from denaturation and aggregation. In their previous work, they demonstrate that these Hero proteins could decrease the fraction of insoluble GFP-TDP43∆NLS in mammalian cell lines and that this mode of action was related to the electrostatic charge of the proteins and not sequence dependent. Although this protective mode of action appears to be similar to that of canonical chaperones, it is unclear how the Hero proteins compare. In this study, the authors compare Hero11 to a panel of canonical chaperones in their cell-based assays and show that it prevents aggregation in a comparable way to DNJA2. It appears that Hero11 decreases the GFP-TDP43∆NLS aggregates better than some other chaperones. They then utilise their expertise in smFRET analysis (Tsuboyama, 2018) to compare what effect DNJA2 and Hero11 (along with Hero11KR-->G (non-charged control)) have on the dynamic structures of the GFP-TDP43∆NLS (labelled with complementary fluorophores in the LCD domain). Based on analysis of the WT GFP-TDP43∆NLS and the A315T GFP-TDP43∆NLS, the authors suggest that the presence of Hero11 and DNJA2 maintain the LCD-domain of TDP43 in an extended conformation and that by doing so, aggregation can be prevented (as assessed in the cell-based assay).

      Despite finding the results very interesting, I feel that the study is preliminary and the conclusions drawn are not fully substantiated by the presented experimental work. Many questions need addressing to validate these findings and conclusions (please see more in the "Significance" section). I have tried to list the main concerns below.

      Questions/concerns:

      Authors used double transient transfections but have not shown quantification of protein levels of the chaperones versus TDP43 - western blots to confirm proper expression (and levels) of the chaperones/Hero protein is crucial without it, we cannot assume that the differences in TDP-43 aggregation are a result of effective chaperoning or due to a lack of expression of any of the chaperone proteins, or high expression of others.

      Authors used quite a high BSA concentration in the smFRET work; it would be useful to see what the TDP43 smFRET trace looks like without BSA incubation (to ensure it is not causing some effect). Also, is there a concentration dependence? The Authors mention they are unable to identify a Hero/TDP43 complex; but if the amount of Hero protein is high (given that it is single molecules tethered), the change in compaction may not relate to the levels/ratios found in the cells (where changes to aggregation are occurring). have the authors considered whether positively charged polymers (poly-Lys) have any impact on the TDP-43 smFRET distribution? Given that the smFRET trace is so heterogeneous, to understand what is happening here would require the comparison of more than 2 variants.

      Although the A315T variant has a very distinct smFRET profile, it is clear that the effects of Hero11KR-->G (that is proposed to have no effect on aggregation or on the smFRET of WT TDP43) has a clear impact on A315T. Why is this?

      The LCD region is prone to PMTs - as the tethered protein is taken from expression in mammalian cells, how can the authors be sure that it has no PMTs? Although a clear difference is observed between WT and A315T in terms of "compactness" of the LCD domain, we cannot assume that the effect of DNAJ2 and Hero11 are the same - in fact, the Hero11 KR-->G control for the A315T is clearly different from the negative control (BSA) and the effect that was seen in WT. As the LCD domain is well-known to be the site of post-translational modifications (ie. Phosphorylation - this would have an effect on an electrostatic Hero11), could the effects be related to changes in PMTs as well?

      The authors mention other studies on DNJA proteins on TDP-43; is the mechanism by which they suppress aggregation known? If the authors want to compare the unknown effects of Hero11, it would be useful to know what DNJ2A is doing, otherwise, the results are still not conclusive, only that "change is similar" in two experiments. What is known about DNJ2A interactions with TDP-43? Did the authors do any pulldown assays to detect a complex in cellulo?

      It is unclear how the findings of the smFRET relate to structural understanding of the LCD-domain of TDP43 (i.e. NMR studies?); is it known whether PTMs are more prominent with the A315T variant as this may explain it's more compact nature? As well, putative helical structure in the LCD domain may lend to the changes in compaction.

      It is unclear how there can be such a prominent FRET ~0 peak and in fact negative values.

      Conclusion is that Hero11 and DNJA2 maintain the TDP43 LCD-domain (soluble protein) in an extended form and that this is linked with the decrease in aggregates found in the cell; however, with the cell-based assay, no analysis to quantify the expression levels of the TDP43 and the chaperones/Hero are presented, and more importantly, no analysis on the complementary soluble fraction (to the filter assay) has been done to show that indeed, these biomolecules maintain the proteins in a soluble form. It is possible that the TDP-43 is being degraded?

      Significance

      Contextually, this study has novelty and potential value for basic research. Firstly, understanding the underlying mechanisms by which Hero protein prevent aggregation would be valuable towards understanding the players in protein homeostasis which can be imbalanced with respect to disease. Secondly, the use of smFRET as a tool in understanding the dynamics of TDP-43 and mutational variants can be powerful in defining structural attributes with pathological consequences in ALS. Although this work shows comparisons between the effect of a canonical chaperone (DNJA2) and Hero11 on the dynamics of monomeric protein and the effect on cellular aggregation, proposing a general mechanism on the data from two TDP-43 variants and a cell-based aggregation assay is premature and more experimental evidence is needed to define the critical link that prevents aggregation of TDP-43 within the cell. Mechanistically, the study does not give a lot of additional insight into the mode of action of Hero11 in the process of preventing aggregation (nor does it explain what DNJA2 is doing and therefore how Hero11 compares and contrasts). The proposed "extended versus collapsed" switch is simplistic and doesn't account for the complexity of TDP-43 structural dynamics. To support their proposed mechanism of action, the authors needs to examine TDP-43 mutational variants (specifically disease-related ones) using their smFRET to understand exactly what the "collapsed" and "extended" data is defining before making the leap that this effect is what is preventing aggregation. There are some structural studies about residual structure in this region (via NMR) that should be considered (https://doi.org/10.1016/j.str.2016.07.007). Although the A315T variant has a very distinct smFRET profile, it is clear that the effects of Hero11KR-->G (that is proposed to have no effect on aggregation or on the smFRET of WT TDP43) has a clear impact on A315T. Why is this? Have the authors considered that the LCD domain of TDP43 is prone to post-translational modifications? Is this variant more phosphorylated - a PMT like phosphorylation is surely to have an impact on interactions with Hero proteins as they are positively charged. Given that the protein is expressed in mammalian cells, it is likely that PMTs have occurred (but the authors should analyse for this).

      With regards to the cell-based aggregation assays, the authors again present a simplified relationship - however, a number of control experiments and additional questions arise. It appears that there is less aggregation with co-expression of some chaperones and the Hero11, but what about the soluble fraction? What is the impact of these biomolecules? Is this that it is maintaining soluble protein, enhancing degradation, propagating soluble oligomers? Equally, how do we know that the levels of the chaperones/Heros and the TDP-43 is the same in each cell - these are transient transfections, and no western blots are shown to confirm the levels of the proteins. In fact, the authors state that "co-transfection of HSP70 (HSPA8), HSP90 (HSP90AB1) or HOP all failed to suppress TDP-43 aggregation compared to GST" and mention that this is in contrast to other studies, but could this be a failure to express these in the cell models? Some western blot/lysate analysis is needed. Chaperones often form complexes with their client proteins, is there any evidence of complexes in these cell models (i.e. using immunoprecipitation)?

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      The submitted manuscript is comparing the effect of individual chaperones and heat-resistant obscure (Hero) proteins on the overall folding of the TDP-43 LCD-domain and its relation to aggregation propensity. Therefore, the authors apply smFRET in order to deduce eventual morphological changes of the LCD-domain from FRET efficiencies. The authors observe that the LCD domain extends its structure upon binding of chaperone/Hero proteins whereas it is collapsed in the absence of those. Furthermore, immunoblotting of filter trap assays indicate that overexpression of chaperones and Hero proteins reduce aggregation of TDP-43 in vivo. Both, the morphological effects on the LCD-domain and the aggregation propensity are significantly enhanced for the TDP-43 A315T mutant. Moreover, the authors tested a charge depleted Hero protein version with reduced "chaperone-like" behaviour. Therefore, the authors conclude that the binding or chaperone activity of the Hero protein is based on its residue specific charges. Finally, the authors conclude that Hero proteins can act similar to chaperones in order to keep protein homeostasis under stress conditions.

      Major comments:

      The similar effect of chaperones and Hero proteins on the morphology of TDP-43 found by the authors is intriguing and the applied experimental procedures seem well described and conducted.

      However, the assumption of the authors that a change in morphology of the LCD-domain by the chaperones and Hero proteins is directly connected to the reduction of TDP-43 aggregation is not entirely clear. Whether an overexpression of individual chaperones and Hero proteins has a direct effect on TDP-43 aggregation cannot be tested in vivo, only. It cannot be excluded that inside the cell the here tested chaperones and Hero proteins control intermediate processes or work as co-factors for other proteins involved in protein homeostasis rather directly influencing the aggregation of TDP-43. Therefore, I recommend in vitro aggregation experiments, using ThT signal as a readout. By adding chaperones, Hero proteins and a negative (BSA or others) control individually, a direct effect on TDP-43 aggregation could be concluded. Those experiments have been extensively used in the field and are quick and straightforward to handle.

      In addition, focusing on the LCD-domain as a main driver for TDP-43 aggregation is limiting this study. In particular, recent studies [1] indicate that the RRM1 and RRM2 sites of TDP-43 have a major impact on TDP-43 gelation and maturation to solid aggregates. Unfortunately, those sites have not been studied in this manuscript.

      As an optional alternative for using Hero11KR->G could be the alteration of buffer conditions and using higher number of salts to promote charge screening. It would be of interest whether the results with the Hero11KR->G could be reproduced with wild type Hero11.

      [1] Lu et al. Nat Cell Biol;24(9):1378-1393 (2022)

      Minor comments:

      Overall, the text is clearly written, and the figures are appropriate.<br /> Whether the activity of individual chaperones or Hero proteins on TDP-43 aggregation "may result in the overall fitness of the cell" or "reinforcing the conformational health of the proteome" is disputable without knowing how the overexpression of certain chaperones or Hero proteins alter the formation of toxic TDP-43 oligomers.

      Significance

      Studying the mechanistic effects of chaperones on aggregating proteins is of major interest for the field in order to understand aging related disbalance of protein homeostasis and the progression of neurological decline, such as seen for amyotrophic lateral sclerosis (ALS). Furthermore, finding homolog proteins, also being able to inhibit protein aggregation, can help to understand overall mechanisms of protein aggregation and processes preventing such fatal behaviour. However, the technique used in this manuscript are not very novel and have been used numerously times before. smFRET is a common technique to look at protein folding/unfolding and is used frequently as a molecular ruler. The manuscript is of interest for the field of protein aggregation and folding, smFRET and neurodegeneration.

      My expertise lies in the field of protein aggregation and inhibition due to chaperones, measuring molecular interactions and neurodegenerative diseases.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      We are obviously very pleased with the general support expressed by the referees, and appreciate their critical comments. We detail below how we propose to respond to their suggestions and queries.

      In view of the fact that my lab is no longer in existence, I will have to rely on the kind generosity of my colleagues at EMBL to host former team members (the two first authors) for a limited period to come back to Heidelberg to carry out any further experimental work that may be needed. This means we will have to limit the work we can do to those experiments with the highest priority. However, we are optimistic that we will be able to obtain indicative results.

      We will also follow most of the referees’ other suggestions and requests for additional data and quantifications, as outlined (or already included) below.

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

      Summary: ASC is the Pyrin/CARD-containing adapter protein that functions as a core component of inflammasome signaling complexes. ASC functions downstream of various NLR- and ALR-inflammasome initiator proteins and upstream of the inflammatory caspases that function as inflammasome effector enzymes. This study uses a novel chimeric construct (Opto-ASC) comprising the Arabidopsis photo-oligomerizable cryptochrome 2 (Cry2-olig) protein with zebrafish ASC to generate transgenic zebrafish larvae wherein ASC oligomerization can be rapidly, dynamically and spatially induced by blue light illumination of either the entire larva or single cells within discrete tissues of an intact larva. Induction of these "opto-inflammasome" complexes is coupled with state-of-the-art, live-cell optical imaging of multiple single cell and integrative tissue parameters to assay various modes of regulated cell death within the peridermal and basal cellular layers of the larval skin. This experimental model was further combined with genetic manipulation of the expression of various zebrafish inflammatory or apoptotic caspases, as well as the two zebrafish members of the gasdermin family of pore-forming proteins which can mediate disruption of plasma membrane permeability without (pre-lytic) or with (pyroptosis) progression to lytic cell death.

      The main results of the study are: 1) introduction of a novel experimental system for dynamic and spatially resolved ASC oligomerization and speck formation within the cells of intact epithelial tissues of a living organism; 2) the ability of these optically induced ASC oligomers/specks to drive multiple modes of regulated cell death which exhibit some (but not all) features of lytic pyroptosis or non-lytic apoptosis depending on cell type and tissue location; 3) the ability of the dying epithelial cells containing optically-induced ASC specks to coordinate rapid adaptive responses in adjacent non-dying cells to maintain integrity/ continuity of skin epithelial barrier; and 4) unexpectedly, no obvious role for either of the two zebrafish gasdermins in the regulated cell death responses.

      Major Comments:

      1. Are the claims and the conclusions supported by the data or do they require additional experiments or analyses to support them? The major goal of this MS is to present a new experimental model (optogenetic activation of ASC oligomerization in transgenic zebrafish) that has the potential to provide new insights regarding the multiple mechanisms by which ASC can regulate inflammasome/ cell death signaling responses in the context of an intact organism. As noted above, some of the observed results are unexpected (e.g., lytic cell death independent of the zebrafish gasdermins in particular epithelial cells) and may reflect mechanisms unique to zebrafish as a non-mammalian vertebrate model versus the mammalian experimental systems (murine and human) that have informed most of our current understanding of how ASC coordinates inflammasome and cell death responses. However, the authors have used rigorous genetic approaches to rule out trivial explanations for the unexpected observations. Thus, no major additional experiments are required to support the claims and conclusions presented in the MS.

      2. Are the suggested experiments realistic in terms of time and resources? Yes. It would help if you could add an estimated time investment for substantial experiments: A few weeks.

      3. Are the data and the methods presented in such a way that they can be reproduced? Are the experiments adequately replicated and statistical analysis adequate? Yes.

      4. Are the experiments adequately replicated and statistical analysis adequate? Yes.

      Minor comments

      1. Specific experimental issues that are easily addressable:

      There's a significant concern with the use of LDC7559 (line 387) as a putative small molecule inhibitor of gasdermin D function to test roles (or lack thereof) of the zebrafish gasdermins in the ASC-triggered lytic cell death responses. A recent study (Amara et al. 2021. Cell. PMID34320407) has reported that LDC7559 does not inhibit gasdermin D (and possibly other gasdermins) but rather acts as an allosteric activator of PFKL (phosphofructosekinase-1 liver type) in neutrophils and thereby suppress generation of the NADPH required for the phagocytic oxidative burst and consequent NETosis. Thus, use of LDC7559 as a presumed gasdermin inhibitor in the current MS is problematic and should be deleted. As an alternative pharmacological approach to suppress gasdermin function, the authors might consider the use of either disulfiram (Hu et al. 2020. Nature Immunology. PMID32367036) and/or dimethylfumarate (Humphries et al. Science. 2020. PMID32820063). These would be straightforward additional experiments.

      We have ordered the reagents to do these experiments. We are optimistic that we will obtain data that will strengthen this part of the ms and be a pointer for future studies by others.

      We propose to keep the information on LDC7559 included, but to discuss the reservations the referee lists above - otherwise, others might ask why we did not even try this inhibitor. .

      Are prior studies referenced appropriately? there are some problems; see below. 2a. One paper is cited twice in lines 724-726 and 727-729. 2b. Another paper is cited twice in lines 790-792 and 793-795. 2c. No journal is included for the referenced study by Shkarina et al in lines 827-828. 2d. No journal is included for the referenced study by Stein et al in lines 831-832. 2e. No journal is included for the referenced study by Masumoto et al in lines 793-795. 2f. No journal is included for the referenced study by Kuri et al in lines 774-775.

      We are embarrassed about these omissions and mistakes and have corrected them..

      Are the text and figures clear and accurate? Generally, yes but with a few exceptions noted below: 3a. line 28: "morphological distinct" should read "morphologically distinct" 3b. line 161: this sentence contains in parentheses "for how long?" I think this was a comment by one author that wasn't removed from the final submitted MS 3c. line 945: spelling "balck" > "black" 3d. line 268: "whereas showed a delayed speck formation": the authors need to specify what model/ condition showed a delayed speck formation 3e. line 262: spelling "egnerated" > "generated"

      Thank you, all corrected.

      CROSS-CONSULTATION COMMENTS I also agree with the comments of the other 2 reviewers. Between the 3 sets of comments and suggestions, the aggregate review will provide the authors with a suitable range of feasible recommendations that will improve an already strong MS.

      Reviewer #1 (Significance (Required)):

      1. General assessment: As noted above, this the major goal of this MS is to present a new experimental model (optogenetic activation of ASC oligomerization in transgenic zebrafish) that has the potential to provide new insights regarding the multiple mechanisms by which ASC can regulate inflammasome/ cell death signaling responses in the context of an intact organism. The authors have used rigorous genetic approaches to rule out trivial explanations for the unexpected observations. In general, the MS describes an elegant model system that will provide a platform for identifying new mechanisms of ASC-dependent inflammasome signaling and regulated cell death.

      2. Advance: This MS describes a highly novel experimental model. Zebrafish are increasingly being used as a genetically tractable model for inflammasome signaling within integrated tissues of intact organism. As noted above, the advances are technical but also conceptual. Future application of this novel model is likely to yield identification of new mechanisms for ASC function in innate immunity and regulated cell death within the context of tissue homeostasis and host defense.

      3. Audience: Basic research and discovery.

      4. Please define your field of expertise with a few keywords to help the authors contextualize your point of view: My group investigates multiple aspects of inflammasome signaling biology at the cellular level with an emphasis on cell-type specific roles for gasdermins in coordinating downstream innate immune responses to inflammasome activation in various myeloid leukocytes (macrophages, dendritic cells, neutrophils, eosinophils, mast cells).

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

      Programmed cell death is critical for host defense and tissue homeostasis. How dead cells initiate cellular responses in the microenvironment with neighbouring cells in vivo is still largely unknown. The authors have chosen a Zebrafish model to tackle this question, given that this model shows advantages for imaging and addresses these pathways in a complex in vivo setting. Their recent development of light-induced activation of caspases (published in JEM) enabled them to investigate cellular responses to a specific type of cell death in vivo at a single cell resolution. In this study, the author further developed a light-induced activation of ASC to specifically look at inflammasome activation-mediated cell death in vivo. The authors have successfully established this system in zebrafish and also observed that Opto-Asc-induced cell death showed different phenotypes as compared to Opto-caspase-a/b-induced cell death. However, it is not really clear why. I have a few specific comments to be addressed or discussed.

      1. In Fig.3 and Fig.4, the majority of Opto-Asc localizes to the plasma membrane but not endogenous Asc. It seems that tagging affects its localization, which could potentially explain its slow kinetics in oligomerization.

      That is an interesting suggestion. The membrane enrichment is indeed reproducible and we have no full explanation for it. However, ASC itself seems to have some affinity for the cell cortex as seen by its association with the apical actin ridges in keratinocytes in the resting state (see e.g. figure 3A). Affinity of ASC for actin is also documented in the literature:(F-actin dampens NLRP3 inflammasome activity via flightless-1 and LRRFIP2 OPEN; https://doi.org/10.1038/srep29834).

      Perhaps the fusion to the optogenetic module somehow enhances the affinity through the initial dimerization. But we can only speculate and have no further evidence that would allow reliable conclusions.

      In Fig.7, the authors showed that deletion of Caspb, but not Caspa, affected the apical extrusion, without affecting cell death. This may indicate that other caspases, like Caspase-8 or/and caspase-3 were involved. This could be addressed through deletion of Caspase-8 or/and caspase-3.

      These are experiments we had in fact done. Unfortunately, they did not allow us to address the question, because the deletions resulted in embryonic lethality. We have added this information to the text.

      It is very surprising that Opto-Asc-mediated cell death is not dependent on Gasdermins, at least in Caspb-dependent apically extruded dead cells.

      Indeed – but see comment by and our response to reviewer 1. We hope to be able to provide additional data.

      CROSS-CONSULTATION COMMENTS I agree with the other two reviewers and don't have further comments.

      Reviewer #2 (Significance (Required)):

      The Opto-Asc zebrafish model developed in this study will enable us to specifically look at inflammasome-mediated cell death in vivo. This model is more physiologically relevant compared to Opto-caspase1 model.

      Audience interested in physiological function of inflammasome activation, but it is questionable whether such a tool will address mechanisms in mammalian cells. Eventually, more evidence for the latter could be provided.

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

      In this article, de Carvahlo and colleagues describe a novel optogenetic tool allowing single cell and temporally controlled induction of ASC clusters in vivo (in zebrafish), a central adaptator protein of the inflammasome complex which is involved in the induction of pyroptosis. This alternative mode of programmed cell death is involved in pathogen response and promote cell swelling and the release of pro-inflammatory factors. Previous works have shown that the inflammasome activation is associated with the formation of a large cluster of ASC protein (called speck) which promotes then the recruitment and the activation of caspase 1. Specks were previously characterised by the same group in vivo (in zebrafish larvae) and could be induced by the overexpression of ASC protein. This however was not compatible with fine spatio-temporal control of speck formation, thus preventing very refined characterisation of the dynamics and the distinction of the cell autonomous and non-cell autonomous effects.

      By fusing ASC to the blue-light sensitive oligomerising protein Cry2-olig under the control of a heat shock promoter, they could induce time controlled induction of speck at the single cell level, which is then followed by cell extrusion and cell death both in the periderm and the basal cell of the skin of zebrafish larvae. Doing so, they could characterise the dynamics of speck formation as well as key paramters affecting its dynamics and the subsequent extrusion. While ASC activation led to apical or basal extrusion in the periderm layer followed by non-apoptotic cell death, it triggers basal extrusion and apoptosis in the basal layer. Importantly, periderm cell elimination does not seem to strictly follow all the features of pyroptosis as it does not require GSDM, and relies on Caspb (not Caspa). It is also associated with strong Calcium release both in the dying and neighbouring cells.

      The authors performed a very careful characterisation of the tools and the optimisation of the condition to form speck and eliminate cells. The experiments are very well performed with all the necessary controls. The results, while to some extend still hard to fully interpret for some aspects, illustrate the plasticity of cell death and cell extrusion, which include several very interesting observations on the direction of extrusion, putative compensatory modes of cell death upon Caspase1 perturbation and the difference of response to ASC clustering depending on the tissue layer. While it is not the main point of this study, the observation that the direction of extrusion can vary very significantly in different genetic backgrounds is also extremely interesting.

      The atypical cell elimination revealed in the system may require further characterisation in the future and suggest that the tools may not be the best to study bona fide pyroptosis. However, I don't believe there is always such strict separation between the modes of cell death and I am sure that it could lead to very interesting insights on inflammasome formation, extrusion and charcaterisation of downstream signalling in vivo, so overall this will be a very interesting resource for the community working on inflammasome, cell death and extrusion.

      I have some suggestions that could help to better characterise the mode of elimination as well as the mechanism of speck formation. I have also some suggestions for comparison with other published results as well as some text editing.

      Main points :

      1. So far, it remains a bit unclear how the authors define precisely speck versus any aggregate and the light induced clusters of Cry2 olig. Is it related to the timescale of formation and/or the lifetime of the aggregates? Is it related to their size?

      There Is no ‘formal’ definition of an inflammatory speck apart from it being the unusually large aggregates that ASC forms once it is activated. Light-induced clusters of Cry2Olig alone, or of Cry2olig fusions with proteins that do not normally oligomerize are much smaller (extensive documentation in the literature).

      A speck is thus a stable aggregate of ASC which is usually around 1 µm in size and is able to activate downstream caspases. But neither of these aspects alone are unique to ASC: prion-like structures can also be large aggregates (indeed ASC-specks have been compared to prions), and much smaller molecular assemblies can activate caspases. Thus ‘speck’ is more an operational definition, and ‘natural’ specks do have both of these properties, but as our experiments show, the properties can actually be separated. I would rather not try to narrow or change the definition, but leave any further discussion to the experts in the field.

      Figure 4E shows a number of variants of ‘speck’-like and other multimers: ASC-mKate and Opto-ASC form large single specks in the presence of endogenous ASC. Opto-ASC specks are only slightly smaller than those formed by endogenously tagged ASC-GFP (see also Supplementary Figure 2E.. Opto-PYD recruits endogenous ASC and becomes incorporated into a speck of approximately the same size, while Opto-CARD does so less efficiently. All of these kill cells. In the absence of endogenous ASC, Opto-ASC forms much smaller specks, and very many in each cell, but these are still functional as seen by the fact that they still kill cells (not the large spot at t = 60 min in the right half of Fig. 4E is not a speck, but the contracted dying cell). Both Opto-PYD and Opto-CARD also form only the small aggregates (quantification will be included), with Opto-PYD still killing the cell by virtue of its ability to recruit caspases via their PYD, whereas Opto-CARD does not.

      Since the authors use most of the time constant blue light illumination, could they also assess how long the speck remains after stopping blue light exposure and quantify their lifetime (relative to the CRY2olig cluster lifetime)?

      Briefly, any speck that contains a functional ASC moiety remains stable and does not disassemble once the blue light is turned off. In skin cells it is not possible to make quantitative measurements because they are killed by the speck. Opto-ASC specks remain stable until they are taken up by macrophages, as originally reported for ASC-GFP specks in Kuri et al. 2017.

      Stability can best be assessed in muscle cells, which do not die upon speck formation. The figure below shows that specks begin to form within minutes of a short pulse of illumination and remain stable (and indeed grow further) for at least 60 min.

      Here is an example:

      Revisions Figure A:

      __Stability of __Opto-ASC specks in muscle cells after exposure to a single pulse of blue light

      Specks in muscle cells expressing Opto-AscTg(mCherry-Cry2olig-asc) are induced by a single illumination with blue light (488nm) at t = 0 for 32 seconds. Multiple oligomers begin to form within 6 minutes, continue to gradually increase in number and, and remain until the end of the movie (60 mins).

      Cell outlines in the overlying epithelium labeled by AKT-PH-GFP are faintly visible in the first frame. Scale bar is 20 mm.

      Similarly could they provide some comparison of the size and localisation of CRY2 olig clusters compared to the speck.

      For size, see above. In addition, the size of the Cry2 oligomers as well as of Opto-ASC specks can vary with expression levels.

      For location, Cry2olig clusters are usually distributed throughout the cell, as seen in most of the right panels in Fig 4E, and in earlier work in cultured cells (e.g. Taslimi et al 2014). ASC specks can form anywhere in the cell, while Cry2olig-ASC has a preference for the cell cortex, but this is not absolute. In keratinocytes, but not in basal cells, the speck usually forms close to the lateral membrane. In the absence of endogenous ASC no real speck is formed but Opto-ASC in this case shows no clear localisation of Opto-ASC to the membrane.

      In view of the variation we see, a strict quantification is difficult: what would be the ‘correct’ definition of classes to look at? To make statistically significant statements, we would need an enormous number of examples in which we could control for all of the variation of expression levels, cell size, day to day variation etc, and we currently don’t have these. We hope the qualitative evidence in the micrographs we show represents the differences well, and we will be happy to provide a larger number of images, if the referees feel this would be helpful.

      With the non functional CRY2olig Asc fusion (Cter fusion), do they still see transient olig2 clustering which then reverse when blue light illumination is gone? I think it might be useful to clarify these points in the main text since most of the quantifications are based on speck localisation/numbering, so their characteristics have to be very well defined.

      That would be interesting to work out, but after our initial experiments with this construct, we did not pursue this further, since it was not a pressing issue at the time. If we can fit this into our planned experimental time table, we will re-assess it. However, while of interest, we feel these data would not add substantially to what we know at this point.

      1. In all the snapshots of speck formation, there seems to be a relative enrichment of the ASC signal at the cytoplasmic membrane (relative to the cytoplasm) prior to strong speck formation. This seems specific of optoASC as it does not seem to happen for the endogeneous ASC or upon overexpression of ASC-mKate (both in this study and in the previous study published by the same group). Is this apparent membrane enrichment something reproducible? (I see that on pretty much every example of this manuscript). If so what could be the explanation? Is there an actual recruitment at the membrane or is it because the membrane/cortical pool takes longer to be recruited in the speck (hence looking relatively more enriched at intermediate time points).

      See our speculations in response to point 1 of the first referee.

      We too would really like to understand this, but see no easy and efficient way of testing it at this point.

      1. There is also a very distinctive ring accumulation that seems to match with apical constriction and/or a putative actomyosin ring (since this is perfectly round, it could match with a structure with high line tension) (see Figure 1E, Figure 3B, Figure 4D...). Is it something already known? Could the authors comment a bit more on this? This could suggest that ASC accumulates in actomyosin cortex, which would be a very interesting property.

      We see that we had failed to be clear about this.

      There are two types of actin-labelled rings that appear around dying cells. One is formed by the epithelial cells that surround the dying cell. This structure becomes visible as soon as the cell begins to shrink. That it is formed by the surrounding cells is clear from mosaics where the dying cell does not express the actin marker (e.g. suppl. Figure 4A) and the parts of the ring are seen only in the subset of surrounding cells that do express the marker. This ring is also not circular, but follows the polygonal shape of the shrinking cell. We believe that this is the contractile structure that closes the wound, as observed in many other cases of wound healing.

      The other is the one the referee describes here. It is formed within the dying cell, as shown by the fact that it is visible in labelled cells when all the surrounding cells are negative for the marker. The other difference is that it appears only once the dying cell has already contracted considerably and begins to round up and be extruded (most clearly seen in Fig. 1E). The third referee had raised a similar point in relation to the same structure seen in Fig. 6C, and we provide below the requested analysis. It relies on resolution in the y-axis, which is unsatisfactory, but nevertheless, it is clear that this ring is in a plane above the apical surface of the epithelium (marked by the red membrane marker, i.e is present in the detaching cell. It may well simply be actin appearing in the entire cortex of the cell as it rounds up and looking like a ring when seen from above. A completely different method for imaging would have to be set up to document this reliably, but we hope that these explanations help to clarify the confusion we may have created.

      We do not see this accumulation in cells that leave the epithelium towards the interior (see figure in the response to ‘minor points’ below).

      In the end, since cell death can also occur without visible speck formation, I am wondering if they are eventually the most relevant structure to be quantified. Is it because speck can be dissolved upon caspase activation and could it relates to the speed at which caspase are activated (which may not leave enough time for strong aggregation and visible speck formation)? I believe it would help to get more explanation/discussion on this point.

      As already mentioned above, it is indeed not obvious what the significance of the large speck is (and it is extremely puzzling why it is that normally one a single one forms in each cell). We agree that it is not necessarily functionally relevant for the signalling outcome to quantify this property – but nor was this the purpose of this work. Regardless of what kind of aggregate is formed, the optogenetic tool allows the induction of ASC-dependent cell death, and therefore the study of the ensuing cellular events.

      The compensatory mechanisms that lead to cell death/extrusion despite depletion of caspb is very interesting. Could the authors use some pan caspase inhibitor (like zvad FMK) to confirm that this block opto-ASC cell death also in this context? Alternatively could they check the status of effector caspase activation using live probe (nucview) or immunostaining in the context of caspb depletion?

      Those would be interesting avenues to pursue. However, for the reason stated above (Leptin lab closing down, members of fish group no longer at EMBL), we are forced to restrict ourselves to the most important experiments, and think we should prioritize the ones mentioned above.

      1. If I understand well, Figure 7C on the right side suggest that the double KO cells don't extrude (if indeed "no change" mean no extrusion, by the way this nomenclature may deserve some clarification in the legend). I don't think these results are mentioned at any point in the main text, but it would be important to include them (since this is an important control).

      This interpretation is in fact correct, and we have changed the labelling in the figure to ‘no immediate death’

      1. Waves of calcium following cell death and cell extrusion have been previously characterised (Takeushi et al. Curr Biol 2020, Y Fujita group). Interestingly, in this previous article they observed waves of calcium near Caspase8 induced death (in MDCK) as well as near laser induced death in zebrafish, while apparently the authors don't see such Calcium waves upon Caspase8 activation in the zebrafish here. I think it would be important to include a comparison of the authors results with this previous paper in the discussion

      We have included this in our discussion.

      There is also a previous study which characterised the impact of caspase1 on cell extrusion (Bonfim Melo et al. Cell Report 2022, A. Yap lab) which promotes apical extrusion in Caco2 cells. I think it would also be important to include this work in the discussion and to compare with the results obtain here in vivo.

      We have included this in our discussion.

      Other minor points:

      1. Line 439: are the numbers given in percentage? if these are absolute numbers, it is out of how many cells ? Same remark line 445: what are the number of cases representing? (percentage?)

      We have rephrased this to make it unambiguous.

      Figure 5: could the authors show periderm and basal cell extrusion with the same type of markers? (membrane or actin or ZO1)? This would help to really compare accurately the morphology and the remodellings associated.

      We used Utr-mNeonGreen to lable actin both in periderm and basal cells. Actin labeling of extruded periderm cells is shown in figure 6C, actin labeling of a dying basal cells and the overlying periderm cells is shown in supplementary figure 5A.

      Is there any obvious differences in cell size or characteristic cell shape between the classic lab strains (golden, AB, AB2B2) and the WIK and experiment strain used here? I do acknowledge that this is clearly not the focus of this study, but given this striking difference (which is related to an important question in the field of extrusion), it would interesting to mention this if there is anything obvious.

      We will make these measurements and include the data.

      1. Figure 6C: what is exactly the localisation in Z of this strong actin accumulation observed during apical extrusion? Is it apical or is it rather on the basal side of the cell? A lateral view of actin could be useful in this figure for all the different conditions described.

      See response to ‘main point 3’ above.

      The images that show this are below. However, even from these images it is hard to appreciate the locations. They are in fact much easier to see by following the movies over time, and through the z-sections at any given time point. We will of course submit the movies with the manuscript.

      Revisions figure B:

      Localization of actin in the yz and xz planes in Opto-Asc-induced cell death and Opto-caspase-8-induced apoptosis

      Orthogonal projections of images of apically (A) and basally (B, C) extruded cells at four time points from time lapse recordings. Each time point shows the x-z plane and the orthogonal yz and and xz planes, in which the apical sides of the epithelium faces the x-z image.

      Actin is labeled with mNeonGreen-UtrCH (cyan), plasma membranes and internal membranes by lyn-tagRFP (magenta). Actin is initially concentrated in the apical cortical ridges of periderm cells.

      1. Apically extruded cell after death is induced by Opto-Asc. As the cell dies actin is lost from the apical ridges and accumulates in the cell cortex in a plane above the original apical surface of the epithelium
      2. Basally extruded cell after death is induced by Opto-Asc. Actin is retained in the apical ridges as the cell shrinks and moves below the epithelium within the dying cell.
      3. Basally extruded cell after death is induced by Opto-Caspase 8. The apical surfaces forms a transient dome in which the actin ridges remain intact before the dying cell is internalized. .

      Figure S3B: could the authors show the utrophin-neonGreen channel separatly? Is there a ring of actin in the dying cell? Also are the membrane protrusion formed more basally? (I suspect this is a z projection, but this would need to be specified in the legend).

      1. Figure 4A legend: I guess the authors meant red arrowheads rather than frame ? This has been corrected

      2. I list below a number of typos I could find in the main text

      Thanks for noticing these, we have corrected all of these, as well as further typos we found.

      Line 29: in Line 30: but Line 151 : from the ...[...] (tissue ?) Line 161: there is most likely a text commenting that was not removed (for how long?) Line 262: generated (egnrtd) Line 268: whereas showed a delay (the subject is missing) Line 269: a point is missing Line 362: which the lack Line 368: a point is missing Line 400: a space is lacking "cellsdepending" Line 438: shrinkwe (space) Line 459 : or I infections Line 525: there is a point missing.

      CROSS-CONSULTATION COMMENTS I generally agree with all the comments raised by the other reviewers which partially overlap with comments I had (see for instance referee two for the role of other caspases and the membrane localisation of the probe).

      Reviewer #3 (Significance (Required)):

      In this article, de Carvahlo and colleagues describe a novel optogenetic tool allowing single cell and temporally controlled induction of ASC clusters in vivo (in zebrafish), a central adaptator protein of the inflammasome complex which is involved in the induction of pyroptosis. This alternative mode of programmed cell death is involved in pathogen response and promote cell swelling and the release of pro-inflammatory factors. Previous works have shown that the inflammasome activation is associated with the formation of a large cluster of ASC protein (called speck) which promotes then the recruitment and the activation of caspase 1. Specks were previously characterised by the same group in vivo (in zebrafish larvae) and could be induced by the overexpression of ASC protein. This however was not compatible with fine spatio-temporal control of speck formation, thus preventing very refined characterisation of the dynamics and the distinction of the cell autonomous and non-cell autonomous effects.

      By fusing ASC to the blue-light sensitive oligomerising protein Cry2-olig under the control of a heat shock promoter, they could induce time controlled induction of speck at the single cell level, which is then followed by cell extrusion and cell death both in the periderm and the basal cell of the skin of zebrafish larvae. Doing so, they could characterise the dynamics of speck formation as well as key paramters affecting its dynamics and the subsequent extrusion. While ASC activation led to apical or basal extrusion in the periderm layer followed by non-apoptotic cell death, it triggers basal extrusion and apoptosis in the basal layer. Importantly, periderm cell elimination does not seem to strictly follow all the features of pyroptosis as it does not require GSDM, and relies on Caspb (not Caspa). It is also associated with strong Calcium release both in the dying and neighbouring cells.

      The authors performed a very careful characterisation of the tools and the optimisation of the condition to form speck and eliminate cells. The experiments are very well performed with all the necessary controls. The results, while to some extend still hard to fully interpret for some aspect, illustrate the plasticity of cell death and cell extrusion, which include several very interesting observations on the direction of extrusion, putative compensatory modes of cell death upon Caspase1 perturbation and the difference of response to ASC clustering depending on the tissue layer. While it is not the main point of this study, the observation that the direction of extrusion can vary very significantly in different genetic backgrounds is also extremely interesting.

      The atypical cell elimination revealed in the system may require further characterisation in the future and suggest that the tools may not be the best to study bona fide pyroptosis. However, I don't believe there is always such strict separation between the modes of cell death and I am sure that it could lead to very interesting insights on inflammasome formation, extrusion and charcaterisation of downstream signalling in vivo, so overall this will be a very interesting resource for the community working on inflammasome, cell death and extrusion.

      My expertise are in cell extrusion, optogenetics, apoptosis and epithelial mechanics. I am not a specialist of the inflammasome and pyroptosis.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this article, de Carvahlo and colleagues describe a novel optogenetic tool allowing single cell and temporally controlled induction of ASC clusters in vivo (in zebrafish), a central adaptator protein of the inflammasome complex which is involved in the induction of pyroptosis. This alternative mode of programmed cell death is involved in pathogen response and promote cell swelling and the release of pro-inflammatory factors. Previous works have shown that the inflammasome activation is associated with the formation of a large cluster of ASC protein (called speck) which promotes then the recruitment and the activation of caspase 1. Specks were previously characterised by the same group in vivo (in zebrafish larvae) and could be induced by the overexpression of ASC protein. This however was not compatible with fine spatio-temporal control of speck formation, thus preventing very refined characterisation of the dynamics and the distinction of the cell autonomous and non-cell autonomous effects.

      • By fusing ASC to the blue-light sensitive oligomerising protein Cry2-olig under the control of a heat shock promoter, they could induce time controlled induction of speck at the single cell level, which is then followed by cell extrusion and cell death both in the periderm and the basal cell of the skin of zebrafish larvae. Doing so, they could characterise the dynamics of speck formation as well as key paramters affecting its dynamics and the subsequent extrusion. While ASC activation led to apical or basal extrusion in the periderm layer followed by non-apoptotic cell death, it triggers basal extrusion and apoptosis in the basal layer. Importantly, periderm cell elimination does not seem to strictly follow all the features of pyroptosis as it does not require GSDM, and relies on Caspb (not Caspa). It is also associated with strong Calcium release both in the dying and neighbouring cells.

      • The authors performed a very careful characterisation of the tools and the optimisation of the condition to form speck and eliminate cells. The experiments are very well performed with all the necessary controls. The results, while to some extend still hard to fully interpret for some aspects, illustrate the plasticity of cell death and cell extrusion, which include several very interesting observations on the direction of extrusion, putative compensatory modes of cell death upon Caspase1 perturbation and the difference of response to ASC clustering depending on the tissue layer. While it is not the main point of this study, the observation that the direction of extrusion can vary very significantly in different genetic backgrounds is also extremely interesting.

      • The atypical cell elimination revealed in the system may require further characterisation in the future and suggest that the tools may not be the best to study bona fide pyroptosis. However, I don't believe there is always such strict separation between the modes of cell death and I am sure that it could lead to very interesting insights on inflammasome formation, extrusion and charcaterisation of downstream signalling in vivo, so overall this will be a very interesting resource for the community working on inflammasome, cell death and extrusion.

      I have some suggestions that could help to better characterise the mode of elimination as well as the mechanism of speck formation. I have also some suggestions for comparison with other published results as well as some text editing.

      Main points:

      1. So far, it remains a bit unclear how the authors define precisely speck versus any aggregate and the light induced clusters of Cry2 olig. Is it related to the timescale of formation and/or the lifetime of the aggregates ? Is it related to their size ? Since the authors use most of the time constant blue light illumination, could they also assess how long the speck remains after stoping blue light exposure and quantify their lifetime (relative to the CRY2olig cluster lifetime) ? Similarly could they provide some comparison of the size and localisation of CRY2 olig clusters compared to the speck. With the non functional CRY2olig Asc fusion (Cter fusion), do they still see transient olig2 clustering which then reverse when blue light illumination is gone ? I think it might be useful to clarify these points in the main text since most of the quantifications are based on speck localisation/numbering, so their characteristics have to be very well defined.

      2. In all the snapshots of speck formation, there seems to be a relative enrichment of the ASC signal at the cytoplasmic membrane (relative to the cytoplasm) prior to strong speck formation. This seems specific of optoASC as it does not seem to happen for the endogeneous ASC or upon overexpression of ASC-mKate (both in this study and in the previous study published by the same group). Is this apparent membrane enrichment something reproducible ? (I see that on pretty much every example of this manuscript). If so what could be the explanation ? Is there an actual recruitment at the membrane or is it because the membrane/cortical pool takes longer to be recruited in the speck (hence looking relatively more enriched at intermediate time points).

      3. There is also a very distinctive ring accumulation that seems to match with apical constriction and/or a putative actomyosin ring (since this is perfectly round, it could match with a structure with high line tension) (see Figure 1E, Figure 3B, Figure 4D...). Is it something already known ? Could the authors comment a bit more on this ? This could suggest that ASC accumulates in actomyosin cortex, which would be a very interesting property.

      4. In the end, since cell death can also occur without visible speck formation, I am wondering if they are eventually the most relevant structure to be quantified. Is it because speck can be dissolved upon caspase activation and could it relates to the speed at which caspase are activated (which may not leave enough time for strong aggregation and visible speck formation) ? I believe it would help to get more explanation/discussion on this point.

      5. The compensatory mechanisms that lead to cell death/extrusion despite depletion of caspb is very interesting. Could the authors use some pan caspase inhibitor ( like zvad FMK) to confirm that this block opto-ASC cell death also in this context ? Alternatively could they check the status of effector caspase activation using live probe (nucview) or immunostaining in the context of caspb depletion ?

      6. If I understand well, Figure 7C on the right side suggest that the double KO cells don't extrude (if indeed "no change" mean no extrusion, by the way this nomenclature may deserve some clarification in the legend). I don't think these results are mentioned at any point in the main text, but it would be important to include them (since this is an important control).

      7. Waves of calcium following cell death and cell extrusion have been previously characterised (Takeushui et al. Curr Biol 2020, Y Fujita group). Interestingly, in this previous article they observed waves of calcium near Caspase8 induced death (in MDCK) as well as near laser induced death in zebrafish, while apparently the authors don't see such Calcium waves upon Caspase8 activation in the zebrafish here. I think it would be important to include a comparison of the authors results with this previous paper in the discussion

      8. There is also a previous study which characterised the impact of caspase1 on cell extrusion (Bonfim Melo et al. Cell Report 2022, A. Yap lab) which promotes apical extrusion in Caco2 cells. I think it would also be important to include this work in the discussion and to compare with the results obtain here in vivo.

      Other minor points:

      1. Line 439: are the numbers given in percentage ? if these are absolute numbers, it is out of how many cells ? Same remark line 445 : what are the number of cases representing ? (percentage ?)

      2. Figure 5: could the authors show periderm and basal cell extrusion with the same type of markers ? (membrane or actin or ZO1) ? This would help to really compare accurately the morphology and the remodellings associated.

      3. Is there any obvious differences in cell size or characteristic cell shape between the classic lab strains (golden, AB, AB2B2) and the WIK and experiment strain used here ? I do acknowledge that this is clearly not the focus of this study, but given this striking difference (which is related to an important question in the field of extrusion), it would interesting to mention this if there is anything obvious.

      4. Figure 6C: what is exactly the localisation in Z of this strong actin accumulation observed during apical extrusion ? Is it apical or is it rather on the basal side of the cell ? A lateral view of actin could be useful in this figure for all the different conditions described.

      5. Figure S3B: could the authors show the utrophin-neonGreen channel separatly ? Is there a ring of actin in the dying cell ? Also are the membrane protrusion formed more basally ? (I suspect this is a z projection, but this would need to be specified in the legend).

      6. Figure 4A legend : I guess the authors meant red arrowheads rather than frame ?

      7. I list below a number of typos I could find in the main text

      Line 29: in

      Line 30: but

      Line 151 : from the ...[...] (tissue ?)

      Line 161: there is most likely a text commenting that was not removed (for how long?)

      Line 262: generated (egnrtd)

      Line 268: whereas showed a delay (the subject is missing)

      Line 269: a point is missing

      Line 362: which the lack

      Line 368: a point is missing

      Line 400: a space is lacking "cellsdepending"

      Line 438: shrinkwe (space)

      Line 459 : or I infections

      Line 525: there is a point missing.

      CROSS-CONSULTATION COMMENTS

      I generally agree with all the comments raised by the other reviewers which partially overlap with comments I had (see for instance referee two for the role of other caspases and the membrane localisation of the probe).

      Significance

      In this article, de Carvahlo and colleagues describe a novel optogenetic tool allowing single cell and temporally controlled induction of ASC clusters in vivo (in zebrafish), a central adaptator protein of the inflammasome complex which is involved in the induction of pyroptosis. This alternative mode of programmed cell death is involved in pathogen response and promote cell swelling and the release of pro-inflammatory factors. Previous works have shown that the inflammasome activation is associated with the formation of a large cluster of ASC protein (called speck) which promotes then the recruitment and the activation of caspase 1. Specks were previously characterised by the same group in vivo (in zebrafish larvae) and could be induced by the overexpression of ASC protein. This however was not compatible with fine spatio-temporal control of speck formation, thus preventing very refined characterisation of the dynamics and the distinction of the cell autonomous and non-cell autonomous effects.

      • By fusing ASC to the blue-light sensitive oligomerising protein Cry2-olig under the control of a heat shock promoter, they could induce time controlled induction of speck at the single cell level, which is then followed by cell extrusion and cell death both in the periderm and the basal cell of the skin of zebrafish larvae. Doing so, they could characterise the dynamics of speck formation as well as key paramters affecting its dynamics and the subsequent extrusion. While ASC activation led to apical or basal extrusion in the periderm layer followed by non-apoptotic cell death, it triggers basal extrusion and apoptosis in the basal layer. Importantly, periderm cell elimination does not seem to strictly follow all the features of pyroptosis as it does not require GSDM, and relies on Caspb (not Caspa). It is also associated with strong Calcium release both in the dying and neighbouring cells.

      • The authors performed a very careful characterisation of the tools and the optimisation of the condition to form speck and eliminate cells. The experiments are very well performed with all the necessary controls. The results, while to some extend still hard to fully interpret for some aspect, illustrate the plasticity of cell death and cell extrusion, which include several very interesting observations on the direction of extrusion, putative compensatory modes of cell death upon Caspase1 perturbation and the difference of response to ASC clustering depending on the tissue layer. While it is not the main point of this study, the observation that the direction of extrusion can vary very significantly in different genetic backgrounds is also extremely interesting.

      • The atypical cell elimination revealed in the system may require further characterisation in the future and suggest that the tools may not be the best to study bona fide pyroptosis. However, I don't believe there is always such strict separation between the modes of cell death and I am sure that it could lead to very interesting insights on inflammasome formation, extrusion and charcaterisation of downstream signalling in vivo, so overall this will be a very interesting resource for the community working on inflammasome, cell death and extrusion.

      • My expertise are in cell extrusion, optogenetics, apoptosis and epithelial mechanics. I am not a specialist of the inflammasome and pyroptosis.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Programmed cell death is critical for host defense and tissue homeostasis. How dead cells initiate cellular responses in the microenvironment with neighbouring cells in vivo is still largely unknown. The authors have chosen a Zebrafish model to tackle this question, given that this model shows advantages for imaging and addresses these pathways in a complex in vivo setting. Their recent development of light-induced activation of caspases (published in JEM) enabled them to investigate cellular responses to a specific type of cell death in vivo at a single cell resolution. In this study, the author further developed a light-induced activation of ASC to specifically look at inflammasome activation-mediated cell death in vivo. The authors have successfully established this system in zebrafish and also observed that Opto-Asc-induced cell death showed different phenotypes as compared to Opto-caspase-a/b-induced cell death. However, it is not really clear why. I have a few specific comments to be addressed or discussed.

      1. In Fig.3 and Fig.4, the majority of Opto-Asc localizes to the plasma membrane but not endogenous Asc. It seems that tagging affects its localization, which could potentially explain its slow kinetics in oligomerization.

      2. In Fig.7, the authors showed that deletion of Caspb, but not Caspa, affected the apical extrusion, without affecting cell death. This may indicate that other caspases, like Caspase-8 or/and caspase-3 were involved. This could be addressed through deletion of Caspase-8 or/and caspase-3.

      3. It is very surprising that Opto-Asc-mediated cell death is not dependent on Gasdermins, at least in Caspb-dependent apically extruded dead cells.

      CROSS-CONSULTATION COMMENTS

      I agree with the other two reviewers and don't have further comments.

      Significance

      The Opto-Asc zebrafish model developed in this study will enable us to specifically look at inflammasome-mediated cell death in vivo. This model is more physiologically relevant compared to Opto-caspase1 model.

      Audience interested in physiological function of inflammasome activation, but it is questionable whether such a tool will address mechanisms in mammalian cells. Eventually, more evidence for the latter could be provided.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      ASC is the Pyrin/CARD-containing adapter protein that functions as a core component of inflammasome signaling complexes. ASC functions downstream of various NLR- and ALR-inflammasome initiator proteins and upstream of the inflammatory caspases that function as inflammasome effector enzymes. This study uses a novel chimeric construct (Opto-ASC) comprising the Arabidopsis photo-oligomerizable cryptochrome 2 (Cry2-olig) protein with zebrafish ASC to generate transgenic zebrafish larvae wherein ASC oligomerization can be rapidly, dynamically and spatially induced by blue light illumination of either the entire larva or single cells within discrete tissues of an intact larva. Induction of these "opto-inflammasome" complexes is coupled with state-of-the-art, live-cell optical imaging of multiple single cell and integrative tissue parameters to assay various modes of regulated cell death within the peridermal and basal cellular layers of the larval skin. This experimental model was further combined with genetic manipulation of the expression of various zebrafish inflammatory or apoptotic caspases, as well as the two zebrafish members of the gasdermin family of pore-forming proteins which can mediate disruption of plasma membrane permeability without (pre-lytic) or with (pyroptosis) progression to lytic cell death.

      The main results of the study are:

      1) introduction of a novel experimental system for dynamic and spatially resolved ASC oligomerization and speck formation within the cells of intact epithelial tissues of a living organism;

      2) the ability of these optically induced ASC oligomers/specks to drive multiple modes of regulated cell death which exhibit some (but not all) features of lytic pyroptosis or non-lytic apoptosis depending on cell type and tissue location;

      3) the ability of the dying epithelial cells containing optically-induced ASC specks to coordinate rapid adaptive responses in adjacent non-dying cells to maintain integrity/ continuity of skin epithelial barrier; and

      4) unexpectedly, no obvious role for either of the two zebrafish gasdermins in the regulated cell death responses.

      Major Comments:

      1. Are the claims and the conclusions supported by the data or do they require additional experiments or analyses to support them? The major goal of this MS is to present a new experimental model (optogenetic activation of ASC oligomerization in transgenic zebrafish) that has the potential to provide new insights regarding the multiple mechanisms by which ASC can regulate inflammasome/ cell death signaling responses in the context of an intact organism. As noted above, some of the observed results are unexpected (e.g., lytic cell death independent of the zebrafish gasdermins in particular epithelial cells) and may reflect mechanisms unique to zebrafish as a non-mammalian vertebrate model versus the mammalian experimental systems (murine and human) that have informed most of our current understanding of how ASC coordinates inflammasome and cell death responses. However, the authors have used rigorous genetic approaches to rule out trivial explanations for the unexpected observations. Thus, no major additional experiments are required to support the claims and conclusions presented in the MS.

      2. Are the suggested experiments realistic in terms of time and resources? Yes. It would help if you could add an estimated time investment for substantial experiments: A few weeks.

      3. Are the data and the methods presented in such a way that they can be reproduced? Are the experiments adequately replicated and statistical analysis adequate? Yes.

      4. Are the experiments adequately replicated and statistical analysis adequate? Yes.

      Minor comments:

      1. Specific experimental issues that are easily addressable:

      There's a significant concern with the use of LDC7559 (line 387) as a putative small molecule inhibitor of gasdermin D function to test roles (or lack thereof) of the zebrafish gasdermins in the ASC-triggered lytic cell death responses. A recent study (Amara et al. 2021. Cell. PMID34320407) has reported that LDC7559 does not inhibit gasdermin D (and possibly other gasdermins) but rather acts as an allosteric activator of PFKL (phosphofructosekinase-1 liver type) in neutrophils and thereby suppress generation of the NADPH required for the phagocytic oxidative burst and consequent NETosis. Thus, use of LDC7559 as a presumed gasdermin inhibitor in the current MS is problematic and should be deleted. As an alternative pharmacological approach to suppress gasdermin function, the authors might consider the use of either disulfiram (Hu et al. 2020. Nature Immunology. PMID32367036) and/or dimethylfumarate (Humphries et al. Science. 2020. PMID32820063). These would be straightforward additional experiments.

      1. Are prior studies referenced appropriately? there are some problems; see below.

      2a. One paper is cited twice in lines 724-726 and 727-729.

      2b. Another paper is cited twice in lines 790-792 and 793-795.

      2c. No journal is included for the referenced study by Shkarina et al in lines 827-828.

      2d. No journal is included for the referenced study by Stein et al in lines 831-832.

      2e. No journal is included for the referenced study by Masumoto et al in lines 793-795.

      2f. No journal is included for the referenced study by Kuri et al in lines 774-775.

      1. Are the text and figures clear and accurate? Generally, yes but with a few exceptions noted below:

      3a. line 28: "morphological distinct" should read "morphologically distinct"

      3b. line 161: this sentence contains in parentheses "for how long?" I think this was a comment by one author that wasn't removed from the final submitted MS

      3c. line 945: spelling "balck" > "black"

      3d. line 268: "whereas showed a delayed speck formation": the authors need to specify what model/ condition showed a delayed speck formation

      3e. line 262: spelling "egnerated" > "generated"

      CROSS-CONSULTATION COMMENTS

      I also agree with the comments of the other 2 reviewers. Between the 3 sets of comments and suggestions, the aggregate review will provide the authors with a suitable range of feasible recommendations that will improve an already strong MS.

      Significance

      1. General assessment:

      As noted above, this the major goal of this MS is to present a new experimental model (optogenetic activation of ASC oligomerization in transgenic zebrafish) that has the potential to provide new insights regarding the multiple mechanisms by which ASC can regulate inflammasome/ cell death signaling responses in the context of an intact organism. The authors have used rigorous genetic approaches to rule out trivial explanations for the unexpected observations. In general, the MS describes an elegant model system that will provide a platform for identifying new mechanisms of ASC-dependent inflammasome signaling and regulated cell death.

      1. Advance:

      This MS describes a highly novel experimental model. Zebrafish are increasingly being used as a genetically tractable model for inflammasome signaling within integrated tissues of intact organism. As noted above, the advances are technical but also conceptual. Future application of this novel model is likely to yield identification of new mechanisms for ASC function in innate immunity and regulated cell death within the context of tissue homeostasis and host defense.

      1. Audience:

      Basic research and discovery.

      1. Please define your field of expertise with a few keywords to help the authors contextualize your point of view:

      My group investigates multiple aspects of inflammasome signaling biology at the cellular level with an emphasis on cell-type specific roles for gasdermins in coordinating downstream innate immune responses to inflammasome activation in various myeloid leukocytes (macrophages, dendritic cells, neutrophils, eosinophils, mast cells).

  2. Jan 2023
    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Response to reviewers

      Reviewer #1

      Reviewer #1 (evidence, reproducibility and clarity (required)):

      Winter et al. present a study of Ebola virus fusion in the acidic environment of the late endosome. Based on cryo-ET of Ebola virions undergoing entry into cells, they note that the VP40 matrix is disassembled and dissociated from the viral membrane in virions seen in the endosome. Subsequent in vitro and computational analyses suggest that protons diffuse across the viral membrane and neutralize anionic lipids on the inner leaflet. They argue that this loss of negative charge reduces the affinity of VP40 for the viral membrane. They further suggest that VP40 dissociation from the viral membrane precedes GP-mediated membrane fusion and contributes to reduction in the energy barrier for membrane stalk formation. Whereas most studies have focused on the importance of acidic pH in triggering GP conformational changes during fusion, the present work contributes new appreciation for VP40-membrane interactions.

      We would like to thank the reviewer for all the insightful comments and appreciation of the novelty.

      In the cryo-ET experiments aimed at visualizing Ebola entry, do the authors see evidence of viral membrane fusion? There is no mention of this in the text. Knowing that the virions that show disassembly of the VP40 matrix are in fact the virions that productively enter cells would support the conclusions of the study. As is stands, one is forced to wonder whether the virions that show VP40 disassembly prior to fusion ultimately fuse.

      *We first note that the EBOV virions shown in Figure 1 entering host cells were captured by cryo-ET at 48 hours post infection and resulted from 2-3 rounds of infection, thus the virions can productively enter the cells by micropinocytosis. Virions that are not able to undergo membrane fusion would be processed in the lysosomes and would not be detectable by cryo-ET at 48 hours post infection. In addition, the virions captured in late endosomes contain nucleocapsids, hence these virions are likely infectious. Together, this is good evidence that we really see events after successful membrane fusion. *

      *We fully agree with the reviewer that capturing a fusion event would provide further proof that fusion depends on prior disassembly of the VP40 matrix layer. To address this, we acquired additional data on cells infected at different time-points post-infection (15 cells imaged); regrettably, we have not been successful in capturing a membrane fusion event, presumably due its fast kinetics. In this study we are technically limited with the amount of the virus we can use for infection in BSL4. The current dataset was generated at an MOI of 0.1 and this makes capturing entry events difficult as we would need an MOI of at least 100-1000 to increase the chances of capturing such a rare event. *

      *Considering the technical difficulties to perform the experiment under BSL4 conditions, we have in addition performed a similar experiment using EBOV VLPs at high concentration (estimated MOI > 100) composed of VP40 and GP (Fig. S5). Despite the high VLP concentration, we could only find 2 tomograms out of 18 tomograms showing VLP entry events. These clearly show that the VP40 matrix is disassembled in VLPs residing in endosomes. The same lamellae displayed sites of viral fusion as evident from enlarged endosomal membrane surfaces studded with GPs facing endosomal lumina. Hence, this new data supports our results that VLPs that undergo VP40 disassembly are able to fuse. We have included the new supplementary figure S5 and added the following sentence to the main text: *

      Lines 96-102: “We were not able to capture virions residing in endosomes in the process of fusing with the endosomal membrane, presumably because virus membrane fusion is a rapid event. However, in a similar experiment using EBOV VLPs composed of VP40 and GP, we could confirm the absence of ordered VP40 matrix layers in VLPs inside endosomal compartments. Moreover, we were able to capture one fusion event and several intracellular membranes studded with luminal GPs, indicating that fusion had taken place (Fig. S5).”

      In the cryo-ET experiments that evaluate VP40 disassembly in vitro, why do the authors leave out NP from their VLP preparations? There is some evidence in the literature (Li et al., JVI 2016) that NP is necessary to form particles with native morphology. If the authors feel that NP is not necessary for their experiment, perhaps this could be noted.

      *Thank you very much for this important comment. Throughout this study, we mainly focused on the fate of the VP40 matrix during entry and thus reduced the complexity of the VLPs used to the minimum – VP40 and GP, so indeed NP was left out before. To address the role of the nucleocapsid in Ebola VLPs uncoating, we have now also included data on VLPs prepared by expression of nucleocapsid components (NP, VP24 and VP35) in addition to GP and VP40. Cryo-ET analysis of these VLPs showed that VLPs mainly contain loosely coiled nucleocapsid. This is consistent with a study by Bharat et al 2012, which shows that compared to virions, VLPs displayed heterogeneous nucleocapsid assembly states and reduced incorporation of nucleocapsids. It is important to note that VLPs containing nucleocapsid also displayed disassembled VP40 matrices at low pH (Fig. S7). Hence, nucleocapsid proteins do not influence the VP40 disassembly driven by low pH and GP-VP40 VLPs can be used as model to study VP40 uncoating. *

      *We included a statement shown on lines 150-153: “We further repeated the experiment using VLPs composed of VP40, GP and the nucleocapsid proteins NP, VP24 and VP35, and observed the same low pH-phenotype described above. These results show that nucleocapsid proteins do not influence the VP40 disassembly driven by low pH.” *

      The authors argue that acidic pH neutralizes the charge of PS phospholipids, thereby removing the electrostatic interactions of basic residues in VP40 and PS. They also note in the Methods section that 7 amino acids in VP40 are predicted by PROPKA to be protonated at pH 4.5. If the authors feel that protonation of these 7 amino acids is not involved in the loss of affinity for PS, this could be stated explicitly and justified. Could the protonation of these 7 amino acids contribute to disassembly of the VP40 lattice, rather than dissociation from the membrane?

      Thank you for this interesting comment. We note that the amino acids predicted to be protonated (*E76, E325, H61, H124, H210, H269, H315, see below) are far away from the interaction interface with the membrane and also away from the intra-dimerization domain. Hence, they do not likely contribute to the loss of affinity for PS but may contribute to conformational changes that facilitate the disassembly of the VP40 matrix. For clarification, we have added the following statement to the methods section: *

      Lines 541-544: “Importantly, these residues are located away from the interaction interface of VP40 with the membrane and their protonation accordingly does not influence membrane-binding. However, protonation of these residues may contribute to conformational changes that facilitate the VP40 matrix disassembly.

      Minor: Figure S5C is difficult to interpret. The red frame on the bars that indicates data acquired at low pH is nearly invisible. Better might be to indicate explicitly (ie, with words) the pH at which data were acquired.

      Thank you very much for this comment. We have changed the design of the graph accordingly. Please note that the figure numbering has changed and that Figure S5C is now Figure S6C.* * Reviewer #1 (significance (required)): The significance of the study stems from the idea that the VP40 lattice and its interaction with the viral membrane plays a direct role in facilitating viral fusion. To my knowledge, this has not been previously addressed. The significance would be considerably increased if the authors were able to demonstrate by cryo-ET that the virions with disassembled VP40 were in fact the virions that productively fused. Nonetheless, this work should be of broad interest to researchers studying viral fusion as it may represent a phenomenon relevant to numerous viruses that enter cells via the endocytic route.

      Reviewer #2 Reviewer #2 (evidence, reproducibility and clarity (required)):

      The manuscript by Winter et al., entitled "The Ebola virus VP40 matrix undergoes endosomal disassembly essential for membrane fusion" describes the structural aspects of the events that precede Ebola virus (EBOV) membrane fusion in late endosome and virion uncoating in the cytosol. By combining state-of-the-art cryo-electron tomography (cryo-ET) with biophysical and computational techniques, they have elucidated the pivotal role of the ebolaviral matrix virion protein 40 (VP40) in modulating the fusion process, in particular discovering that disassembly of the VP40 ordered lattice is low pH-driven, occurs despite the absence of a viral ion channel within the filovirus envelope and takes place through the weakening of VP40 interactions with lipids at the interface between the ebolaviral envelope and matrix. Overall, the manuscript is well written and the research work is very well conceived, with solid orthogonal experimental approaches that mutually validate their respective results. It is opinion of this reviewer that the paper contributes to the elucidation of a key step in the EBOV infection cycle and that it will be of great interest for the readership of Review Commons and for the community of structural biologists. Therefore, I recommend the publication of this paper, however after some minor revision to the text, the figures and the figure legends, which show inconsistencies in the terminology used, the acronyms and could be easily improved by some little graphical editing.

      Thank you very much for your positive feedback and your comments.

      Comments:

      • By starting their abstract and introduction sessions with the term "Ebola viruses" the authors are (on purpose?) preparing the reader to the implicit statement that their findings could be a paradigm model for the other members of the Ebolavirus genus. This is an exciting picture, especially in perspective of VP40-targeting drugs development. Therefore, although conclusions in this sense would probably require further studies, I encourage the authors to implement their figure 3 (or related supplementary figure) with a multiple-sequence alignment, and the relative text in the manuscript, by showing if and how much the basic patch at the C-terminus of VP40 is conserved within the Ebolavirus genus, especially the residues Lys224, Lys225, Lys274 and Lys275.

      Thank you very much for this comment. We have added a corresponding sequence alignment highlighting the high conservation of the basic patch of amino acids across all Ebola virus species (Suppl. Fig. S6). In the text, we refer to the sequence conservation as follows:

      Lines 213-215: “These interactions are driven by basic patches of amino acids which are highly conserved across all EBOV species (Fig. S8 H), further emphasizing their importance in adaptable membrane binding.”

      • It is a bit inconvenient for the reader to follow how a story unfolds while jumping back and forth between figures, and this is why I would recommend to move the period of the sentence at lines 88-91 to the session where figure 5 is discussed.

      *We refer in fact to Figure 1 and fixed the reference accordingly (line 95). *

      • Please, avoid the use of the slang "Ebola" without the apposition "virus", and make the text consistent throughout the manuscript by only using the acronym of each term after it was introduced for the first time.

      Thank you for this comment. We have thoroughly revised the use of technical terms.

      Minor revisions: Line 1: "matrix protein undergoes" We refer here to the entire VP40 matrix layer composed of many VP40 proteins and not to single VP40 proteins (as the individual proteins do not disassemble, but their macromolecular assembly does). For clarification, we changed the title to “matrix layer undergoes”.

      Line 19: "the matrix viral protein 40 (VP40)" We have corrected the statement.

      Line 18: considering that a virus "exists" in the form of a virion while temporarily located outside the cell, and as a "molecular entity" consisting of viral proteins and nucleic acids organised in macromolecular complexes during its life cycle inside the infected cell, this reviewer encourages the authors to rephrase as follows: " Ebola viruses (EBOVs) virions are filamentous particles, ..." Thank you for your suggestion. We have rephrased it to: „Ebola viruses (EBOVs) assemble into filamentous virions“ (line 18).

      Lines 35-36 and line 40: "that is determined by the matrix made up by the viral protein 40 (VP40), which drives ..." And then, directly use the acronym VP24 at line 40

      We have corrected the statement.

      Line 40: as VP24 and VP35 interact with NP but do not interact with the ssRNA genome, please rephrase as follows "the nucleoprotein (NP) which encapsidates the ssRNA genome, and the viral proteins VP24 and VP35 which, together with NP, form the nucleocapsid"

      We have corrected the statement.

      Lines 47-48: "...fusion glycoprotein (GP)...[...] the ebolaviral envelope"

      We have corrected the statement.

      Line 51: "...remarkably long virion of EBOVs undergoes..."

      We have rephrased the statement: line 55: “…remarkably long EBOV virions undergo…”

      Line 63: "... in vitro, and in endo-lysosomal compartments in situ, by cryo-electron..."

      We have corrected the statement.

      Lines 70-71: " to shed light on EBOVs ... [...] with EBOV (Zaire ebolavirus species, Mayinga strain) in biosafety level 4 (BSL4) containment"

      We have corrected the statement.

      Line 72: chemically fixed by? (PFA and GA acronyms have been annotated in figure 1, but should be first mentioned in their explicit form in the text)

      We have now mentioned annotations for GA and PFA both in the main text and in the figure legend in their explicit forms.

      Line 73 (cryo-FIB)

      We have corrected the acronym.

      Line 80: EBOV virions

      We have corrected the statement.

      Figure 1A and line 97: for consistency with the terminology used in the main text, should be perhaps in the second step preferred the term "vitrification" instead of cryofixation? Readers not familiar with the field could be confused by the use of the two synonyms

      We have replaced the term as suggested.

      Lines 92-93: "...these data indicate [...] and suggest..."

      We have corrected the statement.

      Figure 1C and line 100: in the color legend EBOV is annotated as dark teal, however in the segmentation of the reconstructed tomogram there are three objects, one of which in dark teal is evidently a portion of EBOV virion inside the endosome, and other two are in different shades of green. What are those? Please, could author specify their identity in the figure legend with their corresponding color code? The same applies to supplementary figure S2 (see comment below).

      Thank you very much for this comment. All three green objects are EBOV virions. For clarification, we have added numbers 1-3 to the figure and legend and adjusted the text in the legend accordingly (lines 109-110).

      Line 95: "...tomography of EBOV virions..."

      We have corrected the statement.

      Line 98: "...showing EBOV virions..." (This reviewer refers to the use of the term 'EBOVs' as for different species within the genus rather than for different EBOV particles within a dataset)

      We have corrected the statement.

      Line 105: "... a purified EBOV before..." *We realized a mistake in our phrasing: the virion shown in Fig. 1 H is not purified, but a virion found adjacent to the plasma membrane of an infected cell. We have changed the phrasing accordingly (lines 117-118). *

      Line 110 and 113: "...EBOV matrix..." And "EBOV virus-like particles (VLP)"

      We have corrected the statement.

      Line 140, 141, 145 and 147: "EBOV VLPs" and "EBOV VLP"; idem at lines 188-189, 209 and anywhere else in the manuscript (including figure 4A) We have corrected the use of “EBOV VLP(s)” as suggested.

      Line 235: "influenza virus ion channel..."

      We have corrected the statement.

      Line 249: please, use directly the above-introduced acronym for the detergent

      We have revised the use of acronyms.

      Figure 5F: in plot's X axis label: thermolysin (T)?

      Yes, this is correct and stated in the figure legend.* * Line 342: "EBOV have remarkably long..."

      We have corrected the statement.

      Line 420 "...matrix-specific"

      We have corrected the spelling error.

      Line 464: "grids"

      We have corrected the spelling error.

      Line 465: "for cryo-FIB milling"

      We have corrected the statement.

      Line 611: "influenza virus M2 ..." (Please, from which influenza virus strain does the gene come from? Alternatively, which is the NCBI Protein and/or UniProt database code?)

      We have added the information to the Methods (line 648): “….A/Udorn/307/1972 (subtype H3N2))…”

      Line 623: please, use the above-designated acronym for the detergent

      *We have used the acronym as suggested. *

      Line 716: "...based on cryo-ET..." We have corrected the statement.

      Line 718: "influenza virus" We have corrected the term.

      Line 734: "cryo-ET data" We have corrected the term.

      Fig. S8: for consistency with the main text, "thermolysin" We have corrected the spelling of thermolysin throughout the manuscript.

      Fig. S2, C and F: are these EBOV virions (as mentioned in the figure title) or EBOV VLPs (as the legends in the two panels of this figure seem to suggest)? Please, the authors should clarify

      Thank you very much for spotting this mistake! These are indeed EBOV virions and we have changed the legends within the figure accordingly.

      Line 1046: "malleable lipid envelope of the EBOV"; this adjective sounds confusing; the reviewer encourages the authors to rephrase for more clarity.

      We have removed the adjective „malleable”.

      Reviewer #2 (significance (required)): see above.

      __Reviewer #3__Reviewer #3 (evidence, reproducibility and clarity (required)):

      Winter and colleagues describe the molecular architecture of Ebola virus during entry into host cells. The main claims of the paper are that VP40 is disassembled prior to fusion. Disassembly is driven by the low pH environment in the endosomes. PH-induced uncoating works via "passive equilibration" because the Ebola virus envelope does not contain an ion channel. The authors conclude that structural remodeling of VP40 acts as a molecular switch coupling uncoating to fusion. The main novel results of the manuscript are: In situ cryo-ET of endosomal compartments shows EBOV particles with intact condensed nucleocapsids and disordered protein densities that may relate to detached VP40. Five EBOV particles were imaged in the endosome and all had detached VP40 layers. Controls, budding virions and extracellular virions showed intact VP40 layers. Incubation of VP40-Gp VLPs with a pH 4.5 buffer leads to the disorder of the VP40 matrix in vitro, which is independent of Gp presence in the VLPs. MD simulation showed VP40 dimer binding to model membranes containing 30 % PS at pH7 and reduced binding at pH 4.5. Lipidomics revealed the lipid composition of VP40-Gp VLPs demonstrating 9% PS.

      VP40-PHluorin fusions were used to determine acidification of VLPs in vitro and to calculate a permeability coefficient of 1.2 Å sec-1, which is quite low compared to the permeability of the plasma membrane (345 Å sec-1). Next they modeled membrane fusion showing that fusion is more favorable after VP40 disassembly, especially favoring stalk formation. The authors propose further that fusion pore opening is more favorable in the presence of VP40. The authors claim that strong interactions of lipids with VP40 stabilizes the hemifusion intermediate. VP40 Gp VLPs can enter host cells independent of pH once Gp has been activated by thermolysin.

      We thank the reviewer for these interesting comments and valuable suggestions.

      Some of the results are over interpreted and require appropriate modifications.

      Main points that need to be addressed: Imperfections of the membrane could be induced by proteins. Does acidification of the virion depend on GP and its transmembrane region? This can be tested with chimeric GP replacing its TM by unrelated trimeric TMs.

      We agree that this is important to consider. We have addressed this question in Fig. 2 K using VLPs composed of VP40 alone. These VLPs lack GP and still display luminal acidification as evident from the disassembled VP40 matrix when incubated at low pH. Therefore, acidification does not depend on GP. For clarification, we have adjusted the following sentence in the discussion:

      Lines 410-413: “Using VLPs of minimal protein composition (VP40 and GP, and VP40 alone), we show that VP40‑disassembly, i.e. the detachment of the matrix from the viral envelope is triggered by low endosomal pH (Fig. 2). This indicates that VP40 disassembly does not depend on structural changes of other viral proteins, including GP, and is driven solely by the acidic environment.*” *

      Virus entry assays, line 292. The low pH is not only used for Gp cleavage, but induces the conformational changes leading to the post fusion conformation of Gp2. The authors need to check what happens to Gp once it is cleaved by thermolysin. Is this sufficient to induce the conformational changes in Gp? And if so how does entry of such VLPs work, because once the conformational change is triggered, GP2 will adopt the post fusion conformation which is inactive in fusion. This requires further clarification.

      To our knowledge, there is only one study showing that EBOV GP2 changes conformation at low pH in the form of a re-arrangement of the fusion peptide from an extended loop to a kinked conformation (Gregory et al 2011). Importantly, low pH alone is not sufficient to trigger GP mediated membrane fusion and NPC1 is needed as a trigger for membrane fusion process (Das et al, 2020). Hence proteolytically processed GP requires NPC1 binding to change its conformation to post-fusion state. We addressed this question by using pre-cleaved (= GP2) and low pH- treated VLPs in our entry assay (Fig. 5 F). Since low pH-treated VLPs enter host cells as efficiently as VLPs incubated at neutral pH, and low pH-treated and additionally pre-cleaved VLPs enter even more efficiently, it is highly unlikely that low pH triggers the post-fusion conformation as this should inhibit virus entry (as the reviewer pointed out). In conclusion, low pH does not induce the post-conformation in GP2 and we have included a respective sentence for clarification:

      Lines 339-343: * Since thermolysin-treated EBOV VLPs efficiently enter untreated host cells at neutral and low pH, we further conclude that low pH alone does not induce the GP2 post-fusion conformation, which would inhibit virus entry. Together, this suggests a role of low endosomal pH beyond proteolytic processing of EBOV GP, likely for the disassembly of the VP40 matrix.”*

      In the fusion model, the authors claim that VP40 disassembly is more favorable for stalk formation, which is likely true. However, they also claim that strong VP40 interaction, which I would interpret as VP40 filaments interacting with the membrane, favor fusion pore opening. The tomograms and the in vitro experiments with VLPs indicate that the complete VP40 matrix is detached from the membrane under low pH conditions.

      We would like to stress that the modelling results for hemifusion formation and pore opening are independently calculated but have to be interpreted together because they occur sequentially. Hemifusion precedes formation of the pore and hence even though the model shows that the fusion pore opening is favored in the presence of VP40 interaction, membrane fusion cannot proceed to this stage because hemifusion is blocked until the VP40 matrix layer disassembles from the membrane. We apologize for lack of clarity, and we have added the sentences:

      Lines 315-318: “However, it is important to note that hemifusion precedes pore formation in the membrane fusion pathway. Since the disassembly of the VP40 matrix is required for hemifusion and hence for the initiation of membrane fusion, it determines the outcome of the membrane fusion pathway.*” *

      VLPs are purified. Can the authors exclude the possibility that the purification protocol does not damage the VLP membrane leading to in vitro acidification in a low pH environment? Can some of the assays be repeated with non-purified VLPs?

      *Thank you very much for this important comment. To address this question, we had performed the cryo-ET experiments using purified and unpurified VLPs and found that they are virtually indistinguishable. Importantly, unpurified VLPs also undergo VP40 disassembly. We now show images from unpurified VLPs in a supplementary figure (Fig. S7). Thereby, the manuscript contains data of purified VLPs while we also provide proof that the purification protocol does not influence the disassembly of the VP40 matrix. We added the following explanatory sentence to the main text: *

      Lines 151-156: “*We further repeated the experiment using VLPs composed of VP40, GP and the nucleocapsid proteins NP, VP24 and VP35, and observed the same low pH-phenotype described above (Fig. S5 C). Performing the experiments on unpurified VLPs harvested from the supernatant of transfected cells confirmed that the purification protocol applied did not influence the disassembly of the VP40 matrix (Fig. S7). “ *

      Does acidification only work at pH 4.5?

      *We also attempted to verify the acidification of VLPs at higher pH (~5.5. and ~6.0) by cryo-ET, however, subtle structural differences were difficult to quantify. Considering the lower permeability of the VLP membrane compared to the plasma membrane, we think that acidification occurs indeed also at higher pH (as shown for cells), albeit at slower kinetics. *

      Minor points Line 37: Ruigrok et al. 2000 J Mol Biol showed first that Ebola VP40 requires negatively charged lipids for interaction.

      *Thank you for pointing out this reference. We have included it in the text. *

      Fig. 1f: Is VP40 detaching as a filament?

      We have not observed that VP40 detaches as a filament or a linear segment of multiple VP40 dimers. *Since the VP40 dimer is inherently flexible (Fig. 3, Fig. S8) and can rotate along the N- and C-terminal intra- and inter-dimer interfaces, we believe disassembly occurs in a non-ordered fashion (not as filaments, see also Figure 2 G-K). *

      References 8 and 28 are the same. We have corrected the reference duplication.

      Lipidomics: The authors find only 9% PS in the VLPs. How do these results compare to the composition of other enevloped viruses that have been reported to assemble on negatively charged lipids.

      *We compared the lipid composition of the EBOV VLPs to the lipid composition of influenza viruses and HIV, which both bud from the plasma membrane and require negatively charged lipids. When grown in eggs, the envelope of influenza viruses contains 22-25 % PS (Ivanova et al 2015, Li et al 2011), and approximately 12% when produced from MDCK cells (Gerl et al 2012). The envelope of HIV virions produced from HeLa or MT4 cells contains 10-15% PS. These numbers suggest that the producing cell line strongly influences the lipid composition of the virus particles. Besides differences in the producing cell line, the lower amount of PS found in EBOV VLPs could have multiple implications: first, apart from PS, PIP2 has also been shown to interact specifically with VP40 at budding sites in the plasma membrane (Jeevan et al 2017, Johnson et al 2018) and thus also contributes to virion assembly (potentially allowing for a lower PS concentration); second, as recently shown for paramyxoviruses (Norris et al 2022), binding of PS to viral proteins is not based on charge alone but may include specific binding – in which case a high affinity of viral proteins to PS may allow for a lower PS concentration in the target membrane. Overall, the rather low PS content in Ebola VLPs might be important for VP40 interaction and low pH-driven disassembly. *

      EBO virus was suggested to assemble at lipid rafts. Is this reflected by the lipid composition?

      *Yes, that is correct. A hallmark of lipid rafts is the enrichment of cholesterol and sphingomyelin (~32 mol% cholesterol, ~ 14 mol% sphingomyelin) in the microdomains (Pike et al 2002). The lipid composition of the EBOV VLPs determined in our study (~ 39% cholesterol and ~10 mol% sphingomyelin) is consistent with the assembly at lipid rafts. Minor differences stem from the different cell lines and lipidomic approaches used to determine the lipid species. *

      Reviewer #3 (significance (required)): In summary, the manuscript is of high technical quality and the observation that VP40 detaches from the viral membrane prior to membrane fusion is novel and interesting to the field of virus fusion. How acidification occurs in the absence of an ion channel remains to be determined. The authors provide little insight how this might work. The strong part of the manuscript is the EM part, which shows convincing detachement of the VP40 matrix. I cannot comment too much on the modelling part, which, however, sounds solid.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Winter and colleagues describe the molecular architecture of Ebola virus during entry into host cells. The main claims of the paper are that VP40 is disassembled prior to fusion. Disassembly is driven by the low pH environment in the endosomes. PH-induced uncoating works via "passive equilibration" because the Ebola virus envelope does not contain an ion channel. The authors conclude that structural remodeling of VP40 acts as a molecular switch coupling uncoating to fusion.

      The main novel results of the manuscript are:

      • In situ cryo-ET of endosomal compartments shows EBOV particles with intact condensed nucleocapsids and disordered protein densities that may relate to detached VP40.

      • Five EBOV particles were imaged in the endosome and all had detached VP40 layers. Controls, budding virions and extracellular virions showed intact VP40 layers.

      • Incubation of VP40-Gp VLPs with a pH 4.5 buffer leads to the disorder of the VP40 matrix in vitro, which is independent of Gp presence in the VLPs.

      • MD simulation showed VP40 dimer binding to model membranes containing 30 % PS at pH7 and reduced binding at pH 4.5.

      • Lipidomics revealed the lipid composition of VP40-Gp VLPs demonstrating 9% PS.

      • VP40-PHluorin fusions were used to determine acidification of VLPs in vitro and to calculate a permeability coefficient of 1.2 Å sec-1, which is quite low compared to the permeability of the plasma membrane (345 Å sec-1).

      • Next they modeled membrane fusion showing that fusion is more favorable after VP40 disassembly, especially favoring stalk formation.

      • The authors propose further that fusion pore opening is more favorable in the presence of VP40.

      • The authors claim that strong interactions of lipids with VP40 stabilizes the hemifusion intermediate.

      • VP40 Gp VLPs can enter host cells independent of pH once Gp has been activated by thermolysin.

      • Some of the results are over interpreted and require appropriate modifications.

      Main points that need to be addressed:

      • Imperfections of the membrane could be induced by proteins. Does acidification of the virion depend on GP and its transmembrane region? This can be tested with chimeric GP replacing its TM by unrelated trimeric TMs.

      • Virus entry assays, line 292. The low pH is not only used for Gp cleavage, but induces the conformational changes leading to the post fusion conformation of Gp2. The authors need to check what happens to Gp once it is cleaved by thermolysin. Is this sufficient to induce the conformational changes in Gp? And if so how does entry of such VLPs work, because once the conformational change is triggered, GP2 will adopt the post fusion conformation which is inactive in fusion. This requires further clarification.

      • In the fusion model, the authors claim that VP40 disassembly is more favorable for stalk formation, which is likely true. However, they also claim that strong VP40 interaction, which I would interpret as VP40 filaments interacting with the membrane, favor fusion pore opening. The tomograms and the in vitro experiments with VLPs indicate that the complete VP40 matrix is detached from the membrane under low pH conditions. VLPs are purified. Can the authors exclude the possibility that the purification protocol does not damage the VLP membrane leading to in vitro acidification in a low pH environment?

      • Can some of the assays be repeated with non-purified VLPs?

      • Does acidification only work at pH 4.5?

      Minor points

      • Line 37: Ruigrok et al. 2000 J Mol Biol showed first that Ebola VP40 requires negatively charged lipids for interaction.

      • Fig. 1f: Is VP40 detaching as a filament?

      • References 8 and 28 are the same.

      • Lipidomics: The authors find only 9% PS in the VLPs. How do these results compare to the composition of other enevloped viruses that have been reported to assemble on negatively charged lipids.

      • EBO virus was suggested to assemble at lipid rafts. Is this reflected by the lipid composition?

      Significance

      In summary, the manuscript is of high technical quality and the observation that VP40 detaches from the viral membrane prior to membrane fusion is novel and interesting to the field of virus fusion. How acidification occurs in the absence of an ion channel remains to be determined. The authors provide little insight how this might work.

      The strong part of the manuscript is the EM part, which shows convincing detachement of the VP40 matrix. I cannot comment too much on the modelling part, which, however, sounds solid.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      The manuscript by Winter et al., entitled "The Ebola virus VP40 matrix undergoes endosomal disassembly essential for membrane fusion" describes the structural aspects of the events that precede Ebola virus (EBOV) membrane fusion in late endosome and virion uncoating in the cytosol. By combining state-of-the-art cryo-electron tomography (cryo-ET) with biophysical and computational techniques, they have elucidated the pivotal role of the ebolaviral matrix virion protein 40 (VP40) in modulating the fusion process, in particular discovering that disassembly of the VP40 ordered lattice is low pH-driven, occurs despite the absence of a viral ion channel within the filovirus envelope and takes place through the weakening of VP40 interactions with lipids at the interface between the ebolaviral envelope and matrix. Overall, the manuscript is well written and the research work is very well conceived, with solid orthogonal experimental approaches that mutually validate their respective results. It is opinion of this reviewer that the paper contributes to the elucidation of a key step in the EBOV infection cycle and that it will be of great interest for the readership of Review Commons and for the community of structural biologists. Therefore, I recommend the publication of this paper, however after some minor revision to the text, the figures and the figure legends, which show inconsistencies in the terminology used, the acronyms and could be easily improved by some little graphical editing.

      Comments:

      • By starting their abstract and introduction sessions with the term "Ebola viruses" the authors are (on purpose?) preparing the reader to the implicit statement that their findings could be a paradigm model for the other members of the Ebolavirus genus. This is an exciting picture, especially in perspective of VP40-targeting drugs development. Therefore, although conclusions in this sense would probably require further studies, I encourage the authors to implement their figure 3 (or related supplementary figure) with a multiple-sequence alignment, and the relative text in the manuscript, by showing if and how much the basic patch at the C-terminus of VP40 is conserved within the Ebolavirus genus, especially the residues Lys224, Lys225, Lys274 and Lys275.

      • It is a bit inconvenient for the reader to follow how a story unfolds while jumping back and forth between figures, and this is why I would recommend to move the period of the sentence at lines 88-91 to the session where figure 5 is discussed.

      • Please, avoid the use of the slang "Ebola" without the apposition "virus", and make the text consistent throughout the manuscript by only using the acronym of each term after it was introduced for the first time.

      Minor revisions:

      Line 1: "matrix protein undergoes"

      Line 19: "the matrix viral protein 40 (VP40)"

      Line 18: considering that a virus "exists" in the form of a virion while temporarily located outside the cell, and as a "molecular entity" consisting of viral proteins and nucleic acids organised in macromolecular complexes during its life cycle inside the infected cell, this reviewer encourages the authors to rephrase as follows: " Ebola viruses (EBOVs) virions are filamentous particles, ..."

      Lines 35-36 and line 40: "that is determined by the matrix made up by the viral protein 40 (VP40), which drives ..." And then, directly use the acronym VP24 at line 40

      Line 40: as VP24 and VP35 interact with NP but do not interact with the ssRNA genome, please rephrase as follows "the nucleoprotein (NP) which encapsidates the ssRNA genome, and the viral proteins VP24 and VP35 which, together with NP, form the nucleocapsid"

      Lines 47-48: "...fusion glycoprotein (GP)...[...] the ebolaviral envelope"

      Line 51: "...remarkably long virion of EBOVs undergoes..."

      Line 63: "... in vitro, and in endo-lysosomal compartments in situ, by cryo-electron..."

      Lines 70-71: " to shed light on EBOVs ... [...] with EBOV (Zaire ebolavirus species, Mayinga strain) in biosafety level 4 (BSL4) containment"

      Line 72: chemically fixed by? (PFA and GA acronyms have been annotated in figure 1, but should be first mentioned in their explicit form in the text)

      Line 73 (cryo-FIB)

      Line 80: EBOV virions

      Figure 1A and line 97: for consistency with the terminology used in the main text, should be perhaps in the second step preferred the term "vitrification" instead of cryofixation? Readers not familiar with the field could be confused by the use of the two synonyms

      Lines 92-93: "...these data indicate [...] and suggest..."

      Figure 1C and line 100: in the color legend EBOV is annotated as dark teal, however in the segmentation of the reconstructed tomogram there are three objects, one of which in dark teal is evidently a portion of EBOV virion inside the endosome, and other two are in different shades of green. What are those? Please, could author specify their identity in the figure legend with their corresponding color code? The same applies to supplementary figure S2 (see comment below).

      Line 95: "...tomography of EBOV virions..."

      Line 98: "...showing EBOV virions..." (This reviewer refers to the use of the term 'EBOVs' as for different species within the genus rather than for different EBOV particles within a dataset)

      Line 105: "... a purified EBOV before..."

      Line 110 and 113: "...EBOV matrix..." And "EBOV virus-like particles (VLP)"

      Line 140, 141, 145 and 147: "EBOV VLPs" and "EBOV VLP"; idem at lines 188-189, 209 and anywhere else in the manuscript (including figure 4A)

      Line 235: "influenza virus ion channel..."

      Line 249: please, use directly the above-introduced acronym for the detergent

      Figure 5F: in plot's X axis label: thermolysin (T)?

      Line 342: "EBOV have remarkably long..."

      Line 420 "...matrix-specific"

      Line 464: "grids"

      Line 465: "for cryo-FIB milling"

      Line 611: "influenza virus M2 ..." (Please, from which influenza virus strain does the gene come from? Alternatively, which is the NCBI Protein and/or UniProt database code?)

      Line 623: please, use the above-designated acronym for the detergent

      Line 716: "...based on cryo-ET..."

      Line 718: "influenza virus"

      Line 734: "cryo-ET data"

      Fig. S8: for consistency with the main text, "thermolysin"

      Fig. S2, C and F: are these EBOV virions (as mentioned in the figure title) or EBOV VLPs (as the legends in the two panels of this figure seem to suggest)? Please, the authors should clarify

      Line 1046: "malleable lipid envelope of the EBOV"; this adjective sounds confusing; the reviewer encourages the authors to rephrase for more clarity.

      Significance

      see above.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Winter et al. present a study of Ebola virus fusion in the acidic environment of the late endosome. Based on cryo-ET of Ebola virions undergoing entry into cells, they note that the VP40 matrix is disassembled and dissociated from the viral membrane in virions seen in the endosome. Subsequent in vitro and computational analyses suggest that protons diffuse across the viral membrane and neutralize anionic lipids on the inner leaflet. They argue that this loss of negative charge reduces the affinity of VP40 for the viral membrane. They further suggest that VP40 dissociation from the viral membrane precedes GP-mediated membrane fusion and contributes to reduction in the energy barrier for membrane stalk formation. Whereas most studies have focused on the importance of acidic pH in triggering GP conformational changes during fusion, the present work contributes new appreciation for VP40-membrane interactions.

      • In the cryo-ET experiments aimed at visualizing Ebola entry, do the authors see evidence of viral membrane fusion? There is no mention of this in the text. Knowing that the virions that show disassembly of the VP40 matrix are in fact the virions that productively enter cells would support the conclusions of the study. As is stands, one is forced to wonder whether the virions that show VP40 disassembly prior to fusion ultimately fuse.

      • In the cryo-ET experiments that evaluate VP40 disassembly in vitro, why do the authors leave out NP from their VLP preparations? There is some evidence in the literature (Li et al., JVI 2016) that NP is necessary to form particles with native morphology. If the authors feel that NP is not necessary for their experiment, perhaps this could be noted.

      • The authors argue that acidic pH neutralizes the charge of PS phospholipids, thereby removing the electrostatic interactions of basic residues in VP40 and PS. They also note in the Methods section that 7 amino acids in VP40 are predicted by PROPKA to be protonated at pH 4.5. If the authors feel that protonation of these 7 amino acids is not involved in the loss of affinity for PS, this could be stated explicitly and justified. Could the protonation of these 7 amino acids contribute to disassembly of the VP40 lattice, rather than dissociation from the membrane?

      • Minor: Figure S5C is difficult to interpret. The red frame on the bars that indicates data acquired at low pH is nearly invisible. Better might be to indicate explicitly (ie, with words) the pH at which data were acquired.

      Significance

      The significance of the study stems from the idea that the VP40 lattice and its interaction with the viral membrane plays a direct role in facilitating viral fusion. To my knowledge, this has not been previously addressed. The significance would be considerably increased if the authors were able to demonstrate by cryo-ET that the virions with disassembled VP40 were in fact the virions that productively fused. Nonetheless, this work should be of broad interest to researchers studying viral fusion as it may represent a phenomenon relevant to numerous viruses that enter cells via the endocytic route.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Manuscript number: RC-2022-01756

      Corresponding author(s): Wenya, Hou

      1. General Responses

      Dear Editors and Reviewers,

      We deeply appreciate all critical comments and constructive suggestion from all Reviewers, which have inspired us to conceive at least 8 new important experiments and mathematic analysis/modeling (shown in dark red). In addition, we will include more repeats with quantification for spot assays (with more HU doses) and biochemical experiments as well as language revision (shown in orange).

      Below we only list the general response to the Major Concerns raised by at least two Reviewers:

      • To perform mathematic analysis of the single-cell quantitative data (Fig 4, Fig 5 and Fig S4) (Analysis #1).

      50% Sic1 degradation time from Sic1peak

      WT SC

      7.62 min

      whi7 whi5 SC

      7.91 min

      WT HU

      36 min

      whi7 whi5 del HU

      7.49 min

      50% nuclear exit time of Whi5

      WT SC

      4.69 min

      rad53Δsml1Δ SC

      7.60 min

      WT HU

      22.33 min

      rad53Δsml1ΔHU

      13.41 min

      Table R1. 50% Sic1 degradation time calculated from Sic1peak and 50% nuclear exit time of Whi5 based on the experimental data shown in Fig 5 and Fig 4, respectively.

      (2) To reinterpret the HU-induced extension of G1/S transition with an updated model (Analysis #2).

      (3) predict that like WHI7/5 overexpression, CKS1 deletion (PMID: 7958905) or sic1 mutants with longer destruction timing (T2,5S-VLLPP or T2,5S-RXL reported in Fig. 6C, PMID: 32296067), can suppress the HU sensitivity of rad53 mutants according to our model. Moreover, their suppression effects should be epistatic to WHI7/5 overexpression. Alternatively, the dosage suppression of WHI7/5 might be reversed by CKS1 overexpression or sic1 mutants with shorter destruction timing (unfortunately no such mutant has been reported yet). We will perform this set of genetic experiment to test these predictions and thereby functionally reinforce the Whi7/5-Cks1-Sic1 axis (Experiment #1).

      (4) do DNA replication profiling to examine the number of origin firing or replication capacity (Experiment #2).

      (5) To address the suppression effect of phosphorylation in Fig 2E. We agree that the phenotypes of the A-mutants of Whi7 have a weak difference compared with WT, but become much stronger (5-fold difference between two dilutions) compared with the D-mutants. As shown lately in Fig 3, phosphorylation solely facilitates protein stabilization/total levels, which can be masked by ectopic overexpression from an extra plasmid. Moreover, phosphorylation does NOT enhance Whi7’s interaction with Cks1. We should tune down the contribution of phosphorylation and focus more on the stability/protein level. Furthermore, we will do competition assays using A-/D- mutants with GFP and RFP labels (Experiment #3), and add back whi7 13A or 13D in its endogenous locus in the whi7Δwhi5Δ double mutant to test the effect on Sic1 turnover (Experiment #4).

      (6) To add more repeats with quantification for spot assays (with more HU doses) and biochemical experiments (shown in orange).

      Besides reinforcing the current model, these experiments, analysis and re-interpretation may help to clarify two concepts which remain elusive in current version:

      • S-CDK activation can switch from an abrupt/all-or-none pattern under normal condition to a gradually flattened one under replication stress.
      • Consequently, the Whi7/5-Cks1-S-CDKs axis may determine replication capacity and/or number of origin firing. Thus, we did not include a preliminary revision this time due to significant changes. We plan to request at least 6 months for an extensive full revision (e.g., from a short letter to a regular article) to improve this study to a higher level with more general significance. Therefore, we request a revision opportunity from The EMBO Journal.

      2. Point-to-point responses

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

      SUMMARY

      This work begins with a heterologous screen, introducing human genes in double mec1,sml1 yeast deletants, which are alive, but sensitive to hydroxyurea. The readout was mec1,sml1 proliferation in the presence of hydroxyurea. They found that mec1,sml1 yeast mutants carrying the human RB1 gene (a G1/S transcriptional repressor) proliferated on hydroxyurea. Then, they test if known yeast G1/S transcriptional repressors (Whi5 and Whi7) could have similar effects if provided at higher than normal levels (they did). With this initial result, followed up by a variety of experiments, the authors then go on to propose that replication stress, which activates Mec1 and Rad53, triggers the phosphorylation of Whi7 (by Mec1) and Whi5 (by both Rad53 and Mec1) blocking their eviction from the nucleus, allowing them instead to bind and inhibit Cks1, a Cdk processivity factor, needed for the complete phosphorylation and degradation of a Cdk inhibitor, Sic1. This is different from published work a decade earlier in mammalian cells (ref. 37; Bertoli et al.), which showed that upon replication stress, Chk1 phosphorylates G1/S transcriptional repressors to maintain G1/S transcription, which could help genome stability. Here, the authors propose that replication stress could block the G1/S transition. While the model and some of the experiments are interesting, the rationale for some experiments was shaky, and the data do not fully support the conclusions.

      MAJOR POINTS

        • Any cell that undergoes DNA replication must have already destroyed Sic1. It has been known for 25+ years that targeting Sic1 is the only necessary function of G1/Cdk to enable DNA replication (PMID: 8755551). Sic1 does not reappear until the M/G1 transition. Hence, in the authors' model, where cells are already in the S phase, how can multisite phosphorylation and degradation of Sic1 be the critical and final output of the pathway they propose when there shouldn't be any Sic1 around, to begin with? Why would a cell that has already completed Start and the G1/S transition, is in the S phase and experiencing replication stress, care about going through the G1/S? A: Yes, S-CDK activity is regarded as an abrupt or so-called “all-or-none transition” due to a relative short half-life of Sic1 controlled by a robust double-negative feedback loop (PMID: 24130459; 23230424). Sic1 degradation requires multi-phosphorylation events including prime phosphorylation by G1-CDKs, two opposing multi-phosphorylation by S-CDK complex (Clb5–Cdk1–Cks1), one to trigger phosphodegrons and the other to terminate the degron route (PMID: 32296067). The timing and speed (or “sharpness”) of Sic1 degradation is determined by G1-CDKs and S-CDKs, respectively (PMID: 24130459 and PMID: 32296067). Sic1 degradation is not an instantaneous “all-or-none” event even under the optimal growth conditions. The Sic1 destruction timing calculated from Start (defined as 50% nuclear exit of Whi5) is about 14.2 min, whereas the time between Start and Sic1peak is about 5 min from independent studies (Fig 4G, PMID: 24130459; Fig. 6C, PMID: 32296067; Fig. 7B, 32976810). Similarly, the 50% Sic1 degradation time calculated from Sic1peak (50% of Sic1peak) is about 8 min for WT and whi7, in agreement with the results in Figure 2E, PMID: 24130459. However, in the presence of HU, the 50% of Sic1peak time remains constant (7.49 min) in whi7Δwhi5Δ cells but becomes greater than 36 min in WT. Meanwhile, the 50% nuclear exit time of Whi5 (Start) is about 22 min in WT compared to 13 min in rad53Δsml1*Δ upon HU treatment.

      50% Sic1 degradation time from Sic1peak

      WT SC

      7.62 min

      whi7 whi5 SC

      7.91 min

      WT HU

      36 min

      whi7 whi5 del HU

      7.49 min

      50% nuclear exit time of Whi5

      WT SC

      4.69 min

      rad53Δsml1Δ SC

      7.60 min

      WT HU

      22.33 min

      rad53Δsml1ΔHU

      13.41 min

      Table R1. 50% Sic1 degradation time calculated from Sic1peak and 50% nuclear exit time of Whi5 based on the experimental data shown in Fig 5 and Fig 4, respectively.

      Therefore, G1/S transition is a “transition zone” (from Start to 50% of Sic1peak) rather than a single borderline. The key finding of this study is that in the presence of HU, Sic1 degradation speed/sharpness is significantly reduced (Figure 5), mechanistically due to the inhibition of S-CDK-Cks1 by Whi7/5. This eventually reflects a flattened S-CDK activity curve, no longer an “all-or-none activation” any more upon replication stress. S-CDKs phosphorylate the two essential targets (Sld2 and Sld3) to enable DNA replication. Therefore, the Sic1 levels determine the S-CDK activities, which in turn determine the DNA replication capacity (the maximal amount of DNA a cell can synthesize per unit time). In sum, under optimal conditions, the S-CDK activity appears an abrupt/sharp transition and cells replicate DNA in its maximum capacity (i.e., minimal S phase length). When cells encounter replication stress (HU), S-CDK is activated very slowly (very low Sic1 destruction speed) and replicate DNA with a low capacity (slow fork speed and/or few origin firing) to meet the limited resource. Recently, the de Bruin group demonstrates that replication capacity can be tuned by E2F-dependent transcription (includes S-Cyclin genes) in mammalian cells (PMID: 32665547).

      Inspired by these questions, we plan to

      (1) perform mathematic analysis of the single-cell quantitative data (Fig. 5 and S4) (Analysis #1).

      (2) reinterpret the HU-induced extension of G1/S transition with an updated model (Analysis #2).

      (3) predict that like WHI7/5 overexpression, CKS1 deletion (PMID: 7958905) or sic1 mutants with longer destruction timing (T2,5S-VLLPP or T2,5S-RXL reported in Fig. 6C, PMID: 32296067), can suppress the HU sensitivity of rad53 mutants according to our model. Moreover, their suppression effects should be epistatic to WHI7/5 overexpression. Alternatively, the dosage suppression of WHI7/5 might be reversed by CKS1 overexpression or sic1 mutants with shorter destruction timing (unfortunately no such mutant has been reported yet). We will perform this set of genetic experiment to test these predictions and thereby functionally reinforce the Whi7/5-Cks1-Sic1 axis (Experiment #1).

      (4) do DNA replication profiling to examine the number of origin firing or replication capacity (Experiment #2).

      • The results in Figure 2C are confusing and difficult to interpret. For example, comparing lane 8 (WT without hydroxyurea) to lane 7 (WT with hydroxyurea), it appears that there is more phosphorylated Whi7 in lane 7 (hydroxyurea treatment) than in lane 8 (no treatment). But, the ratio of phosphorylated/unphosphorylated Whi7 is not that different (there is very little unphosphorylated Whi7 in lane 8). Same problem when comparing lanes 3 and 4. I understand that they later show that Whi7 is stabilized by hydroxyurea, but from the data in this figure, what exactly can they conclude here?*

      A: Yes, phosphorylation is a bit confusing according to the current statement. Without HU, Whi7 is phosphorylated by G1-CDKs with a much less total protein level as well. With HU, whi7 is phosphorylated by Mec1 and Rad53, because Whi7-P largely disappeared in rad53 mutant (lane 1) and 13A (with all putative Mec1-Rad53 sites mutated, lane 5). Lanes 3 and 4 are biological repeats of Lanes 7-8 with less loading. We will clarify our statement.

      • Their data in Figure 2E show that phosphorylation of Whi7 is not required for suppressing the lethality of rad53,sml1 cells treated with hydroxyurea. Cells carrying Whi7-41A (lacking all possible phosphorylations) suppressed nearly as well as wild-type Whi7 did. The purported differences in the suppression are minuscule at best and not evident at the dilutions tested. It is not clear at all how they can conclude that phosphorylation of Whi7 has anything to do with the ability of Whi7 overexpression to suppress the lethality of rad53,sml1 cells.*

      A: Yes, we agree that the phenotypes of the A-mutants of Whi7 have a weak difference compared with WT, but become much stronger (5-fold difference between two dilutions) compared with the D-mutants. As shown lately in Fig 3, phosphorylation solely facilitates protein stabilization/total levels, which can be masked by ectopic overexpression from an extra plasmid. Moreover, phosphorylation does NOT enhance Whi7’s interaction with Cks1.

      Anyway, we should tune down the contribution of phosphorylation and focus more on the stability/protein level. Furthermore, we will do competition assays using A-/D- mutants with GFP and RFP labels __(Experiment #3) __and add back whi7 13A or 13D in its endogenous locus in the whi7

      • For all the arguments they make about this new role of Whi5 and Whi7 at Start, they do not examine size homeostasis or the kinetics of cell cycle progression in any of their experiments and their mutants, with or without hydroxyurea treatment.*

      A: Good suggestion. We will examine size homeostasis, budding index or the cell cycle progression in the related experiments (Experiment #5). In Fig. S5, we only showed the cell cycle progression profiles in wild-type cells carrying an extra copy of Whi7 WIQ or Whi7 WIQ ΔC. WIQ mutant (without Swi6 binding activity) significantly slowed the cell cycle progression under normal conditions.

      • The Sic1 stability experiments they show in Figure 5 are nice. They would need to be extended to their various mutants, including their Whi7 phosphomutants, to make a case for phosphorylation by Rad53 and Mec1 in this process.*

      A: Thanks, very good suggestion, we will add back whi7 13A or 13D in its endogenous locus in the whi7Δwhi5Δ double mutant (Experiment #4), to avoid the effects of overexpression.

      MINOR POINTS

        • The language is awkward. Editing for style will be necessary.* A: We will request language editing.
      1. They use different hydroxyurea doses in the experiments they show, making it difficult to conclude much when comparing different figures. Why aren't they consistent from experiment to experiment?*

      A: Sorry for the confusing. We used at least three HU concentration gradients in each experiment, but only showed one of them to save the space for a short article. Notably, S. cerevisiae has a much broader range of HU doses (up to 300 mM) than other species (less than 10 mM). We’ll add other Figures during revision.

      **Referees cross-commenting**

      Overall, all reviews are well-aligned. The points raised by the other reviewers are valid, and the reviews are thorough and detailed. I don't know whether the authors will be able to respond since the list is quite long. Even if they do, the manuscript will look very different. I do not have anything else to add.

      Reviewer #1 (Significance (Required)):

      The manuscript presents some interesting data, most notably the role of Whi7 and Whi5 in the stability of Sic1 in vivo and the various in vitro experiments the authors present. The advance is conceptual and mechanistic, offering a different and unanticipated model for the role of these proteins at Start, under replication stress. Unfortunately, the significance of the manuscript is limited. A convincing case for their model and its importance has not been made. For example, their data in Figure 2E, measuring the ability of phosphomutants to suppress the lethality of rad53,sml1 cells upon replication stress, is underwhelming and undermines the importance of the study, particularly to a wider audience.

      A: Thanks for the suggestion, we will improve the model as discussed above.

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

      Summary:

      Jin et al demonstrate a novel type of regulation of the G1/S transition in response to hydroxyurea stress. They approach this by first screening a library of human proteins (cDNA on yeast plasmids) for repressors of the mec1 or rad53 HU sensitivity. HU inhibits ribonucleotide reductase and thus lowers dNTP pools needed for S-phase. This slows replication and leads to stalled replication forks, triggering a "replication stress" response, which is executed by the kinases Mec1 and Rad53. Deletions of mec1 or rad53 are viable in unstressed conditions (with additional sml1 deletion), but are lethal on even low doses of HU. One main hit that rescued this lethality was the human G1/S inhibitor RB. They then went on to confirm that also the yeast analogs Whi5 and Whi7 can rescue mec1 or rad53 lethality when overexpressed. To track down the mechanism, the authors do a variety of genetic and biochemical assays. The resulting model is that Mec1 and Rad53 phosphorylate and stabilize Whi7, which binds to and inhibits the S-phase-CDK complex via the processivity factor Cks1. So on top of acting as a transcriptional repressor, Whi7 (and probably also Whi5) is also a direct interactor and inhibitor of CDK. The binding of Whi7 to Cks1-Clb5/6-CDK prevents the hyperphosphorylation and degradation of the inhibitor Sic1, and thus slows the G1/S transition in response to HU.

      Major comments:

      - Are the key conclusions convincing?

      ->Overall I think the sum of the evidence supports the suggested model, individual claims though are on somewhat shaky grounds based often on single replicates, see below.

      My main conceptual issue may be somewhat just a "semantic" problem. In my understanding "replication stress" refers to stalled replications forks and/or large stretches of single-strand DNA which then triggers a checkpoint response. So how would slowing the G1/S transition help to deal with "replication stress", if replication is not yet happening in these cells? I am assuming Mec1 senses dNTP depletion also in the absence of replication and that is how Mec1 and Rad53 become active in G1. But then maybe the model and the arguments can be phrased differently? What exactly is slowing down Sic1 degradation doing for the cell? Replenishing dNTP pools before the first origins fire? Or is maybe Sic1 not the most important target of this regulation? Maybe also during S-phase, partially inhibiting CDK is beneficial, maybe to stretch out origin firing... or?

      A: Thank you, very good suggestion. This also helps to address the Major Point 1 raised by Reviewer #1. This also reminds us about the work from Pasero’s group demonstrating that Mec1 is activated at the onset of normal S phase by low dNTPs (PMID: 32169162). We will revise the text, and do DNA replication profiling __(Experiment #2) __to examine the number of origin firing or replication speed. Also see response to Point 1 of Reviewer #1.

      - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

      ->Most of the work is done on Whi7 and then some Whi5 in the end, I would tone down on the Whi5 claims a bit.

      A: Very good suggestion. We have to include Whi5 in the story because it plays a redundant role with Whi7.

      - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

      -> Since the authors are clearly able to do quantitative live cell imaging, I do not understand why they do not quantify Whi7 concentrations and localization in response to HU instead of using Western blots of synchronized cells. This would make the whole thing much more credible, especially given the current lack of replicates (see below). This would also allow correlating the timing and amount of the Whi7 response with the stabilizing of Sic1 in single cells.

      A: Yes, we tried but did not see Whi7-GFP clearly because of its very low protein abundance, which is also not shown in literature as far as we know. Only overexpressed Whi7 fluorescence detection(PMID: 33443080).

      ->The causality of phosphorylation being required for stabilization seems plausible from the genetics, but is far from clear in the western blots. Here, concentration increase seems to precede phosphorylation. Could this due to induced Whi7 transcription?

      A: Good suggestion. We will detect Whi7 mRNA levels through qPCR (Experiment #6).

      ->Many if not most claims are based on single replicates. See below.

      - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

      -I am not suggesting any different types of experiments or new methods, so it should be doable within a few weeks.

      - Are the data and the methods presented in such a way that they can be reproduced?

      -I would suggest the authors spell out all of their experimental procedures instead of referring to "as described previously". I think everyone knows the pains of going on a wild goose chase of following references to the original method description.

      A: Good suggestion. I will described all experimental procedures to replace "as described previously".

      - Are the experiments adequately replicated and statistical analysis adequate?

      -The key weakness of this entire paper is imho that many claims are based on single experiments, that are neither replicated nor quantified. For example, all the co-IPs (such as 1E or 3F) should be replicated and the ratio of bait to target quantified and averaged.

      A: Good suggestion. We will show the biological repeats and quantification.

      -If a claim is made regarding increased phosphorylation in vivo, then again this should be replicated and the ratio of phosphorylated to unphosphorylated bands quantified. In many Whi7 gels it looks like it is mainly the total amount of the protein that is changing rather than the phosphorylation state. But again, by eye and from a single replicate, this is hard to tell.

      A: Good suggestion. We will add more repeats.

      -A similar thing holds true for the spot assays. Spot assays are great to show lethality and rescue as in the first figure. But making semi-quantitative claims of different degrees of "partial rescue" from a single spot assay is a bit speculative. This seems especially true since the authors are using different and seemingly random HU concentrations for every spot assay, which suggests that the effect is not very robust and can only be seen in very specific concentration ranges. If e.g. the degree of rescue between WT, A and D mutants or truncations matters for the model/the storyline, then more quantitative growth or competition assays should be added.

      A: Good suggestion. sorry for the confusing. We used at least three HU concentration gradients in each experiment, but only showed one of them to save the space for a short article. Notably, S. cerevisiae has a much broader range of HU doses (up to 300 mM) than other species (less than 10 mM). We’ll add other Figures during revision, and do competition assays using A-/D- mutants with GFP and RFP labels

      Minor comments:

      - Specific experimental issues that are easily addressable.

      ->At least some of the alpha-factor release experiments should contain infos on budding index and/or DNA content to understand see the delay in timing by HU addition.

      A: Good suggestion. We will examine size homeostasis, budding index or the cell cycle progression in the related experiments (Experiment #5).

      - Are prior studies referenced appropriately?

      ->Seems fine from the G1/S side, but I don't know the Mec1/Rad53 literature well enough to judge.

      - Are the text and figures clear and accurate?

      ->The authors could do another round of proofing figures and legends. For example, Fig 5C contains scale bars that are not defined, blot 3E has an asterix labeling that is not defined, the model in 5E has misspelled "degradation"...

      A: We will proofread and revise the full text again.

      - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

      -> The authors use a lot of different mutants (especially for Whi7). Even for someone who knows the proteins fairly well, it is hard to remember throughout the text which abbreviation is relating to which mutations and which function that is addressing. Maybe occasionally remind the reader of what the mutant is or use terms like Whi7non-binding rather than WIQ.

      A: Thank you for your suggestion. We will add (TF non-binding) after WIQ.

      ->The text could also use another round of proof-reading. The overall flow of the storyline is easily comprehensible, but sometimes there is a sudden switch of topics or new proteins come out of nowhere. Some expressions are used in a way that is not common English.

      A: We will request language editing.

      Reviewer #2 (Significance (Required)):

      - Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

      ->This study is a major conceptual contribution to understanding G1/S regulation in perturbed conditions (assuming the results can be replicated and quantified as detailed above). That Whi7 (and maybe Whi5) directly inhibit Clb5/Clb6-CDK through Cks1 binding is an important addition/modification to the current model and may well be important beyond genotoxic stress.

      A: Thanks and we’ll reinforce it with more repeats and quantification.

      - Place the work in the context of the existing literature (provide references, where appropriate).

      ->The authors do this reasonably well.

      - State what audience might be interested in and influenced by the reported findings.

      -> Anyone in the yeast cell cycle/replication field should find this interesting. It should also have important implications for the mammalian cell cycle/replication/DNA damage field.

      - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

      ->I am well familiar with G1/S control and all the methods used in the study. I am not an expert on replication stress/DNA damage/ checkpoint signaling.

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

      Summary

      In their manuscript "Transcription-independent hold of the G1/S transition is exploited to cope with DNA replication stress", Jin et al. intend to show that Retinoblastoma-like G1/S transcriptional repressors can also work as S-CDK-Cks1 inhibitors in response to DNA replication stress, hence prolongating the G1/S transition to enable cells to deal with replication stress. In particular, they aim to identify the mechanism by which Whi7/Srl3 (Suppressor of Rad53 Lethality) rescues the lethality of rad53 yeast mutants. Even though their very first experiment is performed using human RB1, the remainder or the work is performed in the yeast model organism. Experimental methods used include mostly immunoprecipitation experiments (Western blots), spot assays, and some single cell microscopy (not specified if widefield or confocal).

      Major comments

      1) the authors refer to a cross-species screen where they aim to detect human proteins that rescue, upon overexpression, the yeast mec1Dsml1D and rad53Dsml1D lethality (of note, not mec1D/rad53D: why?). They identify hRB1 this way. But the entire screen data is missing, either is the analysis pipeline and "hit selection thresholds" (if applicable). Then no more experiments are performed on human cells or using human proteins. In my opinion this cross-specie approach is not necessary, or not developed enough.

      A: Yes, we have only performed a pilot screen based on the growing on 4 mM HU. We consider removing it. The reason to use mec1Dsml1D for genetic screen is that mec1D/rad53D cells are dead even without HU, whereas dissection assays do not fit for large-scale screening.

      2) Moreover, the interpretation of the data provided as a whole is strongly complicated by the variability in the HU doses used to trigger the Mec1/Rad53 response. While most immunoprecipitation experiments are performed with 200mM, spot assays are performed at various HU concentrations ranging from 3 to 21mM (and exploring the entire range). Sometimes HU concentrations differ on the same Figure panels. Downstream effects of such diverse HU concentrations might also be very diverse and due to this it is difficult to get an understanding of how the different experiments fit together.

      A: Sorry for the confusing. We used at least three HU concentration gradients in each experiment, but only showed one of them to save the space for a short article. Notably, S. cerevisiae has a much broader range of HU doses (up to 300 mM) than other species (less than 10 mM). Spot assays (HU are persistent) are mostly done in the mec1Dsml1D and rad53Dsml1D background (sensitive to 4 mM HU), whereas the IP experiments (only 2-3 h treatment and then removal) are mainly performed in WT or at least in comparison with WT background (resistant up to 250 mM HU). We’ll add other Figures during revision.

      3) Likewise, some experiments are performed only on rad53D backgrounds, or only on mec1D backgrounds (e.g. Fig1B and Fig1F, respectively), while results are claimed valid for the two gene deletion backgrounds.

      A: Thank you. We will add some “not shown data” and remove the invalid claims without data.

      4) Finally, the experiments performed in this study and/or their quantitative analysis are insufficient to support several of the claims, and results are often "over-interpreted". Below I have listed some of such insufficient experiments/analyses, in regard of the interpretation that the authors make of each piece of data.

      - Fig1B could indeed show that Whi7 could rescue rad53D lethality but it is hard to judge from just one tetrad. Many tetrads should be shown to exclude "random sampling" effects.

      A: Thank you. We will add more repeats and remove over-statements. Fig 1B was carried out for at least 12 tetrads but the original picture has been unintentionally lost. We can repeat it if necessary, but the result was validated by the plasmid shuffling experiment (Fig 1C).

      - Fig1F indeed shows that the rescue effect of Whi7 overexpression on mec1Dsml1D lethality in HU does not require its G1/S transcription factor-binding motif (GTB); however, it does not prove that it is independent on any putative effects that Whi7 could have on transcription (it could affect other transcription factors, or even the same ones via other domains).

      A: Good suggestion. As far as we know, there are no reports proving that Whi7 binds to other transcription factors. To rule out this possibility, we will detect whether overexpression of WHI7 affects the transcription of representative G1/S genes (Experiment #7).

      - FigS2A does not really support the authors' claim that Whi7 is hyperphosphorylated upon HU-treatment: the first lane before HU treatment already show the same hyperphosphorylated bands than the second lane (see "darker exposure"); however, the signal intensity is clearly lower so the overall levels of Whi7 are clearly increased by HU, rather than the relative fractions of phosphorylated species.

      A: Yes, we will modify the statement as suggested.

      - Fig2B shows that HU-dependent increase in Whi7 levels is partially abrogated in rad53Dsml1D and mec1Dsml1D mutant backgrounds, which demonstrates that Whi7 upregulation requires either Rad53 or Sml1, and Mec1 or Sml1, but not Rad53/Mec1 as claimed by the authors.

      A: Thank you, we will revise the statement. The only known function of Sml1 is a small unstructured protein inhibitor of Rnr1.

      - Likewise, Fig2B does not show any significant Whi7 phosphorylation following HU-treatment in the whi7-13AP mutant with all CDK consensus sites mutated to alanine. There is indeed a slightly slower migrating band appearing as acknowledge by the authors, which also appears in the mec1Dsml1D and rad53Dsml1D backgrounds. Again here, higher Whi7 levels in the WT background make the comparison with mec1Dsml1D and rad53Dsml1D backgrounds almost impossible. Quantification of the blots, including normalization of the signals of each phosphorylated band to the total signal, could help. But overall this figure does not demonstrate any Mec1/Rad53-dependent Whi7 phosphorylation following HU treatment. The phostag gel Fig2C might show the same result, as the differences in phosTag signals between different conditions might just simply reflect the differences in total amount of Whi7 between those same conditions. However, I acknowledge that Figs 2D and S2C shows Rad53- and Mec1-triggered Whi7 phosphorylation in vitro, but the conditions of this experiments likely differ a lot from in vivo context (kinase levels, competing substrates, presence of co-factors...).

      A: Thank you, we will quantify the blotting as suggested.

      - Along the same lines, Fig3E seems to show that truncation of Whi7 C terminus slightly reduces its efficiency in pulling down Cks1 (indicating reduced interaction). However, the total amount of WT Whi7 in the pull down seems to exceed the total amount of Whi7-DeltaC protein, which could in part explain the difference in Cks1 signal. Here again, quantification of the WB signals and adequate normalization would maybe make this figure more convincing.

      A: Good suggestion. We will show the biological repeats and quantification.

      - Fig4A-B (Whi5 GFP data): the cell representing the absence of HU shows Whi5 nuclear export and therefore likely passes through G1/S; the HU-treated cell shown as example does not export Whi5 from the nucleus, certainly because it does not pass G1/S. IMHO this demonstrates that the G1/S transition is delayed in HU-treated cells (as shown previously), irrespective of any role of Whi5 or Whi7 in this delay.

      - Likewise, Fig4C shows the absence of HU-induced delay in Whi5 nuclear export in rad53Dsml1D cells; however, while the authors claim this indicates "Rad53-dependent nuclear detention of Whi5", it is equally plausible that it indicates that rad53Dsml1D cells do not delay the G1/S transition under HU treatment.

      A: good comments. We should claim both possibilities at this stage. Previous studies mainly show delays in the Start stage (e.g., down-regulate SBF transcription). CLN1/2 deletion is known to delay DNA replication in a Sic1-dependent manner albeit with unknown mechanism in the S-CDK activation stage.

      - The same ambiguity holds for Fig5A,B (Sic1-GFP quantification in whi5Dwhi7D double deletion strain following release from alpha factor block): indeed Sic1 is degraded fast after release from alpha factor block both in presence of HU, while in WT cells Sic1 is not immediately degraded in presence of HU. While authors claim that "Whi7 and Whi5 significantly slow down the Sic1 degradation", this result could also likely reflect that whi5Dwhi7D cells pass G1/S even in this context, and therefore that whi5 or whi7 or both have a role in maintaining cells in G1, not showing any direct implication of Whi5/Whi7 in Sic1 degradation.

      A: good comments. It only provides some indirect hints. For instance, whi5Dwhi7D cells pass G1/S in a same timing as WT in the absence of HU (Fig. S4), indicating that the role of Whi5/7 in the G1/S delay is related to additional checkpoint function, not normal G1 maintaining function. Moreover, it should be combined with other results, for example, dosage suppression effects in the presence of HU and inhibitory effects in the absence of HU. Direct evidence of Whi5/Whi7 in Sic1 degradation and Cks1 inhibition comes only from the biochemical experiments shown in Fig 3E-3H.

      - FigS5: the authors show here that overexpression of Whi7-WIQ (that does not bind SBF) slows down the G1/S transition following release from alpha factor blockade, but this data does not demonstrate anything related to the role of Whi7 in the DNA replication stress response. Indeed, since Whi7 sequesters Cln3 in the ER (independent of any putative role on transcription regulation), its overexpression could simply reflect an increased sequestering of Cln3 pool. What does this result become in a cln3D background?

      A: Very good suggestion. We will check whether cln3Δ affects the suppression effect of Whi7 (Experiment #8).

      Due to the fundamental concerns raised above in the interpretation of the data, it is difficult to predict the outcome of more controlled experiments that would aim to prove the same statements. This makes the estimation of the time and resources required to complete the study almost impossible.

      Minor comments

      Owing to the major comments above, an important re-structuration of the study is required, and minor comments I may have on this version are likely to be irrelevant to the revised manuscript.

      Reviewer #3 (Significance (Required)):

      The study aims to establish a molecular link between the progression through the G1/S transition and the DNA damage and DNA replication stress responses. Establishing molecular links between different phases of the cell cycle is an important question in basic research, and might be of interest for a broad range of cell biologists, even though the study is conducted in a model organism (budding yeast). The link proposed involves G1/S inhibitors Whi5 and Whi7, that would bind and inhibit the Cks1 subunit of S-CDK complexes, downstream of Rad53 and Mec1 signaling. The authors confirm some known results (e.g., Whi7 overexpression bypasses rad53 lethality in presence of HU) and gather new pieces of data using well-established methods (immunoprecipitation, spot assays, fluorescence microscopy). However, many experiments reported in this study are not sufficient to support the authors' claims, and therefore the novel mechanistic insight that this study ambitions to provide is not established.

      My scientific background being more in bio-imaging than in biochemistry, it is possible that I missed some hands-on experience to correctly interpret artefacts on western blots, however I do not feel like I missed sufficient expertise to evaluate any section of the manuscript.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Summary

      In their manuscript "Transcription-independent hold of the G1/S transition is exploited to cope with DNA replication stress", Jin et al. intend to show that Retinoblastoma-like G1/S transcriptional repressors can also work as S-CDK-Cks1 inhibitors in response to DNA replication stress, hence prolongating the G1/S transition to enable cells to deal with replication stress. In particular, they aim to identify the mechanism by which Whi7/Srl3 (Suppressor of Rad53 Lethality) rescues the lethality of rad53 yeast mutants. Even though their very first experiment is performed using human RB1, the remainder or the work is performed in the yeast model organism. Experimental methods used include mostly immunoprecipitation experiments (Western blots), spot assays, and some single cell microscopy (not specified if widefield or confocal).

      Major comments

      1. the authors refer to a cross-species screen where they aim to detect human proteins that rescue, upon overexpression, the yeast mec1Dsml1D and rad53Dsml1D lethality (of note, not mec1D/rad53D: why?). They identify hRB1 this way. But the entire screen data is missing, either is the analysis pipeline and "hit selection thresholds" (if applicable). Then no more experiments are performed on human cells or using human proteins. In my opinion this cross-specie approach is not necessary, or not developed enough.
      2. Moreover, the interpretation of the data provided as a whole is strongly complicated by the variability in the HU doses used to trigger the Mec1/Rad53 response. While most immunoprecipitation experiments are performed with 200mM, spot assays are performed at various HU concentrations ranging from 3 to 21mM (and exploring the entire range). Sometimes HU concentrations differ on the same Figure panels. Downstream effects of such diverse HU concentrations might also be very diverse and due to this it is difficult to get an understanding of how the different experiments fit together.
      3. Likewise, some experiments are performed only on rad53D backgrounds, or only on mec1D backgrounds (e.g. Fig1B and Fig1F, respectively), while results are claimed valid for the two gene deletion backgrounds.
      4. Finally, the experiments performed in this study and/or their quantitative analysis are insufficient to support several of the claims, and results are often "over-interpreted". Below I have listed some of such insufficient experiments/analyses, in regard of the interpretation that the authors make of each piece of data.

      5. Fig1B could indeed show that Whi7 could rescue rad53D lethality but it is hard to judge from just one tetrad. Many tetrads should be shown to exclude "random sampling" effects.

      6. Fig1F indeed shows that the rescue effect of Whi7 overexpression on mec1Dsml1D lethality in HU does not require its G1/S transcription factor-binding motif (GTB); however, it does not prove that it is independent on any putative effects that Whi7 could have on transcription (it could affect other transcription factors, or even the same ones via other domains).
      7. FigS2A does not really support the authors' claim that Whi7 is hyperphosphorylated upon HU-treatment: the first lane before HU treatment already show the same hyperphosphorylated bands than the second lane (see "darker exposure"); however, the signal intensity is clearly lower so the overall levels of Whi7 are clearly increased by HU, rather than the relative fractions of phosphorylated species.
      8. Fig2B shows that HU-dependent increase in Whi7 levels is partially abrogated in rad53Dsml1D and mec1Dsml1D mutant backgrounds, which demonstrates that Whi7 upregulation requires either Rad53 or Sml1, and Mec1 or Sml1, but not Rad53/Mec1 as claimed by the authors.
      9. Likewise, Fig2B does not show any significant Whi7 phosphorylation following HU-treatment in the whi7-13AP mutant with all CDK consensus sites mutated to alanine. There is indeed a slightly slower migrating band appearing as acknowledge by the authors, which also appears in the mec1Dsml1D and rad53Dsml1D backgrounds. Again here, higher Whi7 levels in the WT background make the comparison with mec1Dsml1D and rad53Dsml1D backgrounds almost impossible. Quantification of the blots, including normalization of the signals of each phosphorylated band to the total signal, could help. But overall this figure does not demonstrate any Mec1/Rad53-dependent Whi7 phosphorylation following HU treatment. The phostag gel Fig2C might show the same result, as the differences in phosTag signals between different conditions might just simply reflect the differences in total amount of Whi7 between those same conditions. However, I acknowledge that Figs 2D and S2C shows Rad53- and Mec1-triggered Whi7 phosphorylation in vitro, but the conditions of this experiments likely differ a lot from in vivo context (kinase levels, competing substrates, presence of co-factors...).
      10. Along the same lines, Fig3E seems to show that truncation of Whi7 C terminus slightly reduces its efficiency in pulling down Cks1 (indicating reduced interaction). However, the total amount of WT Whi7 in the pull down seems to exceed the total amount of Whi7-DeltaC protein, which could in part explain the difference in Cks1 signal. Here again, quantification of the WB signals and adequate normalization would maybe make this figure more convincing.
      11. Fig4A-B (Whi5 GFP data): the cell representing the absence of HU shows Whi5 nuclear export and therefore likely passes through G1/S; the HU-treated cell shown as example does not export Whi5 from the nucleus, certainly because it does not pass G1/S. IMHO this demonstrates that the G1/S transition is delayed in HU-treated cells (as shown previously), irrespective of any role of Whi5 or Whi7 in this delay.
      12. Likewise, Fig4C shows the absence of HU-induced delay in Whi5 nuclear export in rad53Dsml1D cells; however, while the authors claim this indicates "Rad53-dependent nuclear detention of Whi5", it is equally plausible that it indicates that rad53Dsml1D cells do not delay the G1/S transition under HU treatment.
      13. The same ambiguity holds for Fig5A,B (Sic1-GFP quantification in whi5Dwhi7D double deletion strain following release from alpha factor block): indeed Sic1 is degraded fast after release from alpha factor block both in presence of HU, while in WT cells Sic1 is not immediately degraded in presence of HU. While authors claim that "Whi7 and Whi5 significantly slow down the Sic1 degradation", this result could also likely reflect that whi5Dwhi7D cells pass G1/S even in this context, and therefore that whi5 or whi7 or both have a role in maintaining cells in G1, not showing any direct implication of Whi5/Whi7 in Sic1 degradation.
      14. FigS5: the authors show here that overexpression of Whi7-WIQ (that does not bind SBF) slows down the G1/S transition following release from alpha factor blockade, but this data does not demonstrate anything related to the role of Whi7 in the DNA replication stress response. Indeed, since Whi7 sequesters Cln3 in the ER (independent of any putative role on transcription regulation), its overexpression could simply reflect an increased sequestering of Cln3 pool. What does this result become in a cln3D background? Due to the fundamental concerns raised above in the interpretation of the data, it is difficult to predict the outcome of more controlled experiments that would aim to prove the same statements. This makes the estimation of the time and resources required to complete the study almost impossible.

      Minor comments

      Owing to the major comments above, an important re-structuration of the study is required, and minor comments I may have on this version are likely to be irrelevant to the revised manuscript.

      Significance

      The study aims to establish a molecular link between the progression through the G1/S transition and the DNA damage and DNA replication stress responses. Establishing molecular links between different phases of the cell cycle is an important question in basic research, and might be of interest for a broad range of cell biologists, even though the study is conducted in a model organism (budding yeast). The link proposed involves G1/S inhibitors Whi5 and Whi7, that would bind and inhibit the Cks1 subunit of S-CDK complexes, downstream of Rad53 and Mec1 signaling. The authors confirm some known results (e.g., Whi7 overexpression bypasses rad53 lethality in presence of HU) and gather new pieces of data using well-established methods (immunoprecipitation, spot assays, fluorescence microscopy). However, many experiments reported in this study are not sufficient to support the authors' claims, and therefore the novel mechanistic insight that this study ambitions to provide is not established.

      My scientific background being more in bio-imaging than in biochemistry, it is possible that I missed some hands-on experience to correctly interpret artefacts on western blots, however I do not feel like I missed sufficient expertise to evaluate any section of the manuscript.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      Jin et al demonstrate a novel type of regulation of the G1/S transition in response to hydroxyurea stress. They approach this by first screening a library of human proteins (cDNA on yeast plasmids) for repressors of the mec1 or rad53 HU sensitivity. HU inhibits ribonucleotide reductase and thus lowers dNTP pools needed for S-phase. This slows replication and leads to stalled replication forks, triggering a "replication stress" response, which is executed by the kinases Mec1 and Rad53. Deletions of mec1 or rad53 are viable in unstressed conditions (with additional sml1 deletion), but are lethal on even low doses of HU. One main hit that rescued this lethality was the human G1/S inhibitor RB. They then went on to confirm that also the yeast analogs Whi5 and Whi7 can rescue mec1 or rad53 lethality when overexpressed. To track down the mechanism, the authors do a variety of genetic and biochemical assays. The resulting model is that Mec1 and Rad53 phosphorylate and stabilize Whi7, which binds to and inhibits the S-phase-CDK complex via the processivity factor Cks1. So on top of acting as a transcriptional repressor, Whi7 (and probably also Whi5) is also a direct interactor and inhibitor of CDK. The binding of Whi7 to Cks1-Clb5/6-CDK prevents the hyperphosphorylation and degradation of the inhibitor Sic1, and thus slows the G1/S transition in response to HU.

      Major comments:

      • Are the key conclusions convincing?

      Overall I think the sum of the evidence supports the suggested model, individual claims though are on somewhat shaky grounds based often on single replicates, see below.

      My main conceptual issue may be somewhat just a "semantic" problem. In my understanding "replication stress" refers to stalled replications forks and/or large stretches of single-strand DNA which then triggers a checkpoint response. So how would slowing the G1/S transition help to deal with "replication stress", if replication is not yet happening in these cells? I am assuming Mec1 senses dNTP depletion also in the absence of replication and that is how Mec1 and Rad53 become active in G1. But then maybe the model and the arguments can be phrased differently? What exactly is slowing down Sic1 degradation doing for the cell? Replenishing dNTP pools before the first origins fire? Or is maybe Sic1 not the most important target of this regulation? Maybe also during S-phase, partially inhibiting CDK is beneficial, maybe to stretch out origin firing... or? - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

      Most of the work is done on Whi7 and then some Whi5 in the end, I would tone down on the Whi5 claims a bit. - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

      Since the authors are clearly able to do quantitative live cell imaging, I do not understand why they do not quantify Whi7 concentrations and localization in response to HU instead of using Western blots of synchronized cells. This would make the whole thing much more credible, especially given the current lack of replicates (see below). This would also allow correlating the timing and amount of the Whi7 response with the stabilizing of Sic1 in single cells.

      The causality of phosphorylation being required for stabilization seems plausible from the genetics, but is far from clear in the western blots. Here, concentration increase seems to precede phosphorylation. Could this due to induced Whi7 transcription?

      Many if not most claims are based on single replicates. See below. - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

      I am not suggesting any different types of experiments or new methods, so it should be doable within a few weeks. - Are the data and the methods presented in such a way that they can be reproduced?

      I would suggest the authors spell out all of their experimental procedures instead of referring to "as described previously". I think everyone knows the pains of going on a wild goose chase of following references to the original method description. - Are the experiments adequately replicated and statistical analysis adequate?

      The key weakness of this entire paper is imho that many claims are based on single experiments, that are neither replicated nor quantified. For example, all the co-IPs (such as 1E or 3F) should be replicated and the ratio of bait to target quantified and averaged.

      If a claim is made regarding increased phosphorylation in vivo, then again this should be replicated and the ratio of phosphorylated to unphosphorylated bands quantified. In many Whi7 gels it looks like it is mainly the total amount of the protein that is changing rather than the phosphorylation state. But again, by eye and from a single replicate, this is hard to tell.

      A similar thing holds true for the spot assays. Spot assays are great to show lethality and rescue as in the first figure. But making semi-quantitative claims of different degrees of "partial rescue" from a single spot assay is a bit speculative. This seems especially true since the authors are using different and seemingly random HU concentrations for every spot assay, which suggests that the effect is not very robust and can only be seen in very specific concentration ranges. If e.g. the degree of rescue between WT, A and D mutants or truncations matters for the model/the storyline, then more quantitative growth or competition assays should be added.

      Minor comments:

      • Specific experimental issues that are easily addressable.

      At least some of the alpha-factor release experiments should contain infos on budding index and/or DNA content to understand see the delay in timing by HU addition. - Are prior studies referenced appropriately?

      Seems fine from the G1/S side, but I don't know the Mec1/Rad53 literature well enough to judge. - Are the text and figures clear and accurate?

      The authors could do another round of proofing figures and legends. For example, Fig 5C contains scale bars that are not defined, blot 3E has an asterix labeling that is not defined, the model in 5E has misspelled "degradation"... - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

      The authors use a lot of different mutants (especially for Whi7). Even for someone who knows the proteins fairly well, it is hard to remember throughout the text which abbreviation is relating to which mutations and which function that is addressing. Maybe occasionally remind the reader of what the mutant is or use terms like Whi7non-binding rather than WIQ.

      The text could also use another round of proof-reading. The overall flow of the storyline is easily comprehensible, but sometimes there is a sudden switch of topics or new proteins come out of nowhere. Some expressions are used in a way that is not common English.

      Significance

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

      This study is a major conceptual contribution to understanding G1/S regulation in perturbed conditions (assuming the results can be replicated and quantified as detailed above). That Whi7 (and maybe Whi5) directly inhibit Clb5/Clb6-CDK through Cks1 binding is an important addition/modification to the current model and may well be important beyond genotoxic stress. - Place the work in the context of the existing literature (provide references, where appropriate).

      The authors do this reasonably well. - State what audience might be interested in and influenced by the reported findings.

      Anyone in the yeast cell cycle/replication field should find this interesting. It should also have important implications for the mammalian cell cycle/replication/DNA damage field. - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

      I am well familiar with G1/S control and all the methods used in the study. I am not an expert on replication stress/DNA damage/ checkpoint signaling.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary

      This work begins with a heterologous screen, introducing human genes in double mec1,sml1 yeast deletants, which are alive, but sensitive to hydroxyurea. The readout was mec1,sml1 proliferation in the presence of hydroxyurea. They found that mec1,sml1 yeast mutants carrying the human RB1 gene (a G1/S transcriptional repressor) proliferated on hydroxyurea. Then, they test if known yeast G1/S transcriptional repressors (Whi5 and Whi7) could have similar effects if provided at higher than normal levels (they did). With this initial result, followed up by a variety of experiments, the authors then go on to propose that replication stress, which activates Mec1 and Rad53, triggers the phosphorylation of Whi7 (by Mec1) and Whi5 (by both Rad53 and Mec1) blocking their eviction from the nucleus, allowing them instead to bind and inhibit Cks1, a Cdk processivity factor, needed for the complete phosphorylation and degradation of a Cdk inhibitor, Sic1. This is different from published work a decade earlier in mammalian cells (ref. 37; Bertoli et al.), which showed that upon replication stress, Chk1 phosphorylates G1/S transcriptional repressors to maintain G1/S transcription, which could help genome stability. Here, the authors propose that replication stress could block the G1/S transition. While the model and some of the experiments are interesting, the rationale for some experiments was shaky, and the data do not fully support the conclusions.

      Major points

      1. Any cell that undergoes DNA replication must have already destroyed Sic1. It has been known for 25+ years that targeting Sic1 is the only necessary function of G1/Cdk to enable DNA replication (PMID: 8755551). Sic1 does not reappear until the M/G1 transition. Hence, in the authors' model, where cells are already in the S phase, how can multisite phosphorylation and degradation of Sic1 be the critical and final output of the pathway they propose when there shouldn't be any Sic1 around, to begin with? Why would a cell that has already completed Start and the G1/S transition, is in the S phase and experiencing replication stress, care about going through the G1/S?
      2. The results in Figure 2C are confusing and difficult to interpret. For example, comparing lane 8 (WT without hydroxyurea) to lane 7 (WT with hydroxyurea), it appears that there is more phosphorylated Whi7 in lane 7 (hydroxyurea treatment) than in lane 8 (no treatment). But, the ratio of phosphorylated/unphosphorylated Whi7 is not that different (there is very little unphosphorylated Whi7 in lane 8). Same problem when comparing lanes 3 and 3. I understand that they later show that Whi7 is stabilized by hydroxyurea, but from the data in this figure, what exactly can they conclude here?
      3. Their data in Figure 2E show that phosphorylation of Whi7 is not required for suppressing the lethality of rad53,sml1 cells treated with hydroxyurea. Cells carrying Whi7-41A (lacking all possible phosphorylations) suppressed nearly as well as wild-type Whi7 did. The purported differences in the suppression are minuscule at best and not evident at the dilutions tested. It is not clear at all how they can conclude that phosphorylation of Whi7 has anything to do with the ability of Whi7 overexpression to suppress the lethality of rad53,sml1 cells.
      4. For all the arguments they make about this new role of Whi5 and Whi7 at Start, they do not examine size homeostasis or the kinetics of cell cycle progression in any of their experiments and their mutants, with or without hydroxyurea treatment.
      5. The Sic1 stability experiments they show in Figure 5 are nice. They would need to be extended to their various mutants, including their Whi7 phosphomutants, to make a case for phosphorylation by Rad53 and Mec1 in this process.

      Minor points

      1. The language is awkward. Editing for style will be necessary.
      2. They use different hydroxyurea doses in the experiments they show, making it difficult to conclude much when comparing different figures. Why aren't they consistent from experiment to experiment?

      Referees cross-commenting

      Overall, all reviews are well-aligned. The points raised by the other reviewers are valid, and the reviews are thorough and detailed. I don't know whether the authors will be able to respond since the list is quite long. Even if they do, the manuscript will look very different. I do not have anything else to add.

      Significance

      The manuscript presents some interesting data, most notably the role of Whi7 and Whi5 in the stability of Sic1 in vivo and the various in vitro experiments the authors present. The advance is conceptual and mechanistic, offering a different and unanticipated model for the role of these proteins at Start, under replication stress. Unfortunately, the significance of the manuscript is limited. A convincing case for their model and its importance has not been made. For example, their data in Figure 2E, measuring the ability of phosphomutants to suppress the lethality of rad53,sml1 cells upon replication stress, is underwhelming and undermines the importance of the study, particularly to a wider audience.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

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

      In this manuscript, the authors report dorsomedial hypothalamus-specific PR-domain containing protein 13-knockout (DMH-Prdm13-KO) mice recapitulated age-associated sleep alterations such as sleep fragmentation and increased sleep attempts during sleep deprivation (SD). These phenotypes were further exacerbated during aging, with increased adiposity and decreased physical activity, resulting in shortened lifespan. Moreover, overexpression of Prdm13 in the DMH ameliorated sleep fragmentation and excessive sleepiness during SD in old mice. They identified maintaining Prdm13 signaling in the DMH might play an important role to control sleep-wake patterns during aging. These findings are interesting and novel and the evidence they provided looks solid.*

      We deeply appreciate that this reviewer found our findings are interesting and the evidence solid.

      *Major comments 1. The author spent a lot of words on Sirt1 in the introduction. Since Sirt1 regulates Prdm13, is there a link between the two in age-related sleep changes? If so, you can add some results and discussion. *

      Thank you very much for raising this important issue. Our previous study demonstrated that a mouse model with high hypothalamic Sirt1 activity displays reduced number of transitions between wakefulness and NREM sleep (reference # 15), revealing that hypothalamic Sirt1, as well as Prdm13, is involved in the regulation of sleep fragmentation.However, sleep propensity was not altered in Sirt1-overexpressing transgenic mice (reference #13) and DMH-Prdm13-KO mice (Fig. 1). Based on these findings, we added the following sentence in the Results.

      On page 11, line 267-274

      "...... Similarly, a mouse model with high hypothalamic Sirt1 activity displays reduced number of transitions between wakefulness and NREM sleep15, revealing that hypothalamic Sirt1, as well as Prdm13, is involved in the regulation of sleep fragmentation. Sleep propensity was not altered in Sirt1-overexpressing transgenic mice13. Given that the level of hypothalamic Prdm13 and its function decline with age, age-associated sleep fragmentation could be promoted through the reduction of Prdm13/Sirt1 signaling in the DMH, but sleep propensity might be increased by other mechanisms. "

      • In Figure 2e, the author describes n=7-8 in the figure legend, but why do both groups on the column show eight data? Is there something wrong with the statistics? Please check the statistics in the article carefully. *

      We corrected n=7-8 to n=8 in the figure legend of Fig. 2e.

      • DMH is known as one of the major outputs of hypothalamus circadian system and is involved in the circadian regulation of sleep-wakefulness (J.Neurosci. 23, 10691-10702 ; Nat Neurosci 4:732-738). Does Prdm13 correlate with circadian rhythms? The author can add relevant content to the discussion *

      As per this reviewer's suggestion, we added the following sentence in the Discussion on page 20, line 500-508,

      "For instance, it would be of great interest to elucidate whether Prdm13 signaling in the DMH contributes to regulate the circadian system, since the DMH is known to be involved in the regulation of several circadian behaviors32,33. Although DMH-Prdm13-KO mice did not display abnormal period length compared with controls, further studies are needed to address this possibility."

      *Minor comments 1. The immunohistochemical diagram in the paper is not representative enough, as shown in FIG. 2b and c. *

      We apologize that our presentation in Figs. 2a-c was confusing. Although Fig. 2b shows the numbers of cFos cells in the entire region of the DMH (summed up from three DMH regions), the images in Fig. 2c are from one of DMH regions for each condition. To avoid confusion, we revised the legend of Figs. 2a-c and the manuscript in the Results as follows:

      -In the figure legend of Figs. 2a-c

      "a, Total numbers of cFos+ cells ......... b,c, Images of DMH sections at bregma -1.67 mm ......."

      -In the Results on page 7, line 180

      "...... the hypothalamus, the DMH (summed up from bregma -1.67 to -1.91mm) showed a greater number of cFos+ cells during SD compared to SD-Cont (Fig. 2a-c, Supplementary Fig. 2a)..... "

      • In FIG. 5h, the authors showed that the effect of overexpression of Prdm13 was very obvious, but the expression range of the virus after injection was lacking. Is there a fluorescent gene such as GFP on the virus to directly see the expression of the virus in the brain? *

      Unfortunately, we do not hold extra samples to check the distribution of the virus after injection. However, we have established sufficient injection technique to target the DMH using the lentivirus system that we used in this study (Satoh et al Cell Metab 2013).

      • Were mice singly housed or housed in groups? *

      Most of the mice were housed in groups, except for the DR study. We added this information in the section Animal models of the Methods on page 41, line 935

      ".....RIKEN BRC. Most of the mice were housed in groups, except for the DR study. For the DR study ,..... "

      • The part of sleep analysis needs to be further refined. How can REM and NREM in mice be distinguished and according to what criteria? *

      We added the criteria to define NREM and REM in the section Sleep analysis of the Methods on page 42, line 995-998.

      ".......with visual examination. EEG periods dominated by higher amplitude delta wave activity with nuchal muscle atonia were scored as NREM sleep epochs. REM sleep consisted of periods of semi-uniform theta activity EEG with muscle atonia and/or muscle atonia with brief myoclonic twitches. Score was blinded ......"

      • The authors may consider adding more recent literature related to DMH and sleep, such as DOI: 10.1093/cercor/bhac258 * We incorporated this reference to the following sentence in the section Results on page 8, line 194.

      "........ Although DMH neurons are linked to sleep21, aging and longevity .... "

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

      Summary: In this study, Tsuji et al. demonstrate that Prdm13 signaling is involved in the regulation of sleep-wake pattern. They also identified Prdm13 as a transcription factor in the DMH neurons. Major comments: 1. The evidence presented in Fig. 1 of age-related sleep fragmentation is potentially problematic. Although many previous studies have demonstrated fragmented sleep, especially fragmentation of NREM sleep, in aged mice compared to young mice, the data here do not suggest NREM fragmentation, because no change in the NREM bout duration was found. REM, on the other hand, may indeed have fragmentation during the dark phase, but REM only takes a small portion of the total sleep. Therefore, the conclusion that sleep is fragmented in old mice is not fully supported by Fig. 1. I noticed that the authors used 4-6 months old mice as the young group. Mice of this age can hardly be called "young". The females even start to have lowered fertility. This might be one of the reasons for the discrepancy between this and other studies. Repeating these experiments (and others involving the young group) with mice of more appropriate age (usually 2-3 months old) is recommended. Nonetheless, aging-caused sleep change is not new knowledge and has been reported repeatedly. This part of the results should be in the supplementary figures. *

      We deeply appreciate this reviewer's comment. In accordance with this reviewer's suggestion, we carefully reconsidered the age of young mice. Most of published studies used mice at 2 to 4 months of age as the young group [2 to 4-month-old (7 studies), 4.6-month-old (1 study), 6-month-old (1 study), 2 to 6-month-old (1 study)]. Thus, to strictly use mice at 3-4 months of age as the young group, we excluded data of one cohort using mice at 6 months of age (2 mice each age group). Consistent with many previous studies, our revised data demonstrated that sleep fragmentation during NREM sleep is predominantly observed in old mice compared with young mice, particularly during the dark period. Based on these new results, we revised Fig.1, Suppl Fig.1, and all description related to Fig. 1 (manuscript on page 5-7, line 103-171). We would like to keep Fig. 1 as it is. Since most of the previous studies used males but not females, data from females are still lacking in the field (Campos-Beltran and Marshall, Pflugers. Arch., 473:841-851, 2021).

      • The sleep phenotypes in aged mice and in Prdm13-KO mice are clearly distinct from each other. In the old mice (Fig. 1), REM sleep is fragmented but the total amount remains unchanged, and NREM sleep is increased (both bout number and total amount), indicating there may be more REM-to-NREM transitions, which the authors should quantify. However, Fig. 3 shows in Prdm13-KO mice, there is no REM fragmentation. In fact, it even seems to stabilize REM. But NREM duration is shorted, and no change in the total NREM or REM sleep time. These results suggest that the sleep alterations caused by aging and Prdm13-KO might have some overlap but are mostly in parallel and likely through different mechanisms. Therefore, the rationale of connecting Prdm13 signaling to aging-caused sleep changes is questionable. Is there a developmental change of Prdm13 expression in DMH between young and old mice? The authors also showed that Prdm13-KO in old mice caused decrease in NREM duration but has no effect on REM sleep, but in normal old mice, it is REM, but not NREM that has a defect. Prdm13 overexpression also only mildly decreased NREM bout number without affecting the episode duration of either NREM or REM, which can hardly be interpreted as "ameliorating sleep fragmentation". To me, all these results just suggest parallel actions of Prdm13 and aging on sleep, with Prdm13 mostly affecting NREM sleep but aging mostly impairing REM sleep. *

      We deeply appreciate this reviewer's keen eyes. We carefully reassessed REM sleep data in Fig. 3. The revised data showed that whereas the duration of NREM episodes in DMH-Prdm13-KO mice during the dark period were significantly shorter compared to control group, the duration of REM episodes in the KO mice was not significantly altered. Therefore, after revising Fig. 1 and 3, our results showed that both aging and Prdm13-KO similarly affect the duration of NREM sleep episodes. These results suggest that sleep fragmentation, in particular, during NREM sleep, is commonly observed in old mice and DMH-Prdm13-KO mice. In addition to sleep fragmentation during NREM sleep, excessive sleepiness during SD was also commonly observed in old mice and DMH-Prdm13-KO mice. On the other hand, the effect of aging and Prdm13-KO on sleep propensity was distinct from each other. We think that age-associated sleep fragmentation could be promoted through Prdm13 signaling in the DMH, but sleep propensity might be increased by other mechanisms. We described these results and possibilities in the Results, and revised the Abstract as follows:

      On page 11, line 264-274

      "activity in DMH-Prdm13-KO mice (Fig. 3h, Supplementary Fig. 3f-h). Together, sleep fragmentation during NREM sleep and excessive sleepiness during SD are commonly observed in old mice and DMH-Prdm13-KO mice, but the effects of aging and Prdm13-KO on sleep propensity were distinct from each other.............. Given that the level of hypothalamic Prdm13 and its function decline with age16, age-associated sleep fragmentation could be promoted through the reduction of Prdm13/Sirt1 signaling in the DMH, but sleep propensity might be increased by other mechanisms."

      On page 2, line 45-46

      "Dietary restriction (DR), a well-known anti-aging intervention in diverse organisms, ameliorated age-associated sleep fragmentation and increased sleep attempts during SD, whereas these effects of DR were abrogated in DMH-Prdm13-KO mice."

      As this reviewer pointed out, the effect of Prdm13 overexpression on NREM sleep fragmentation seems to be moderate, but we still observed effects on excessive sleepiness during SD. Thus, we revised the manuscript related to Prdm13-overexpression study in the Abstract and Results as follows:

      On page 2, line 47-48

      "Moreover, overexpression of Prdm13 in the DMH ameliorated sleep fragmentation and excessive sleepiness during SD in old mice."

      On page 16, line 387-401

      "Overexpression of Prdm13 in the DMH partially affects age-associated sleep alterations

      ...... (Fig. 5h). The number of wakefulness and NREM sleep episodes in old Prdm13-OE mice were significantly lower, whereas duration of wakefulness in old Prdm13-OE mice tended to be longer than old control mice during the dark period with no change in the duration of NREM episodes (Fig. 5i,j). Intriguingly, .... Thus, the restoration of Prdm13 signaling in the DMH partially rescue age-associated sleep alterations, but its effect on sleep fragmentation is moderate."

      • What is the control manipulation for sleep deprivation? The authors need to clarify this in the Methods. Also, sleep deprivation has confounding effects including but not limited to stress, food deprivation (since food was removed during SD), human experimenter (since a gentle-touch method was used). Without proper controls for these variables, the authors should avoid concluding that the changes they saw at cellular level are due to sleep loss. *

      Thank you very much for this suggestion. We added detailed description for AL-SD (the control manipulation for SD) in the section SD study of the Materials as follows:

      On page 42-43, line 1014-1020

      "Mice for control manipulation (AL-SD) were also individually housed prior to the experiment without SD and food removal. We checked the level of blood glucose in the SD study, and found that the level of blood glucose was indistinguishable between SD and AL-SD groups (126±6 and 131±4 mg/dL, respectively), revealing that nutritional status is equal between these two groups."

      Identification of Prdm13+ cells using neuronal markers should be performed in addition to electrophysiological characterizations.

      We performed immunofluorescence using anti-MAP2 antibody and confirmed that most Prdm13+ cells are neurons. We added this new result in Suppl Fig. 2g.

      • Figs. 6 and 7 seem very disconnected from the main story. Identification of Prdm13 as a transcription factor is potentially interesting, but how does it account for its role in affecting sleep? The criteria of picking Cck, Grp and Pmch out of other candidate genes potentially regulated by Prdm13 and the rationale to investigate these genes seem unclear. More importantly, no evidence was shown regarding how Cck/Grp *

      Base on RNA-sequencing using DMH samples from DMH-Prdm13-KO and control mice, we got several candidate genes as downstream genes of Prdm13. After validating the candidate genes by qRT-PCR, Cck, Grp and Pmch were detected as top-hit genes. We thus further assessed these three genes in this study. Our result showed that Cckexpression in the hypothalamus significantly declines with age. Based on other literature, hypothalamic Cck seems to be involved in sleep control. Therefore, it is conceivable that Prdm13 controls age-associated sleep alterations via modulating Cck expression. However, as this reviewer pointed out, we are still lacking the evidence showing the role of Prdm13/Cck axis in age-associated sleep alterations. We now clearly described the limitation of our study in the Discussion on page 23, line 560-562.

      "However, the detailed molecular mechanisms by which Prdm13 in the DMH regulates age-associated sleep fragmentation and excessive sleepiness during SD still need to be elucidated in future study. "

      *Minor comments: 1. Please note on the images of Fig. 2d what the green fluorescence was. It is very confusing as is, given that it's surrounded by quantifications of c-fos in the figure. *

      The label "Prdm13" was added in Fig. 2d.

      Please note use a different color for Prdm13 in several figure images (e.g., Fig. 2f, g, 7a,d, and Supplementary 2c). Yellow usually means overlap of red and green.

      Since we have four-color images in Fig. 7, we consistently used yellow for Prdm13 throughout the main figures of the paper. At this moment, we would like to keep the current version of images, but we will revise images if the editor of affiliate journal requests this revision.

      • Please note the statistic test results on power spectrum graphs. *

      We added the statistic test results on power spectrum graphs in Figs. 1d, 4c, and 5d.

      • Inconsistency between the graphs in Fig. 3d and the description in the text. Fig. 3d suggests no change in Wake episode duration, significant decrease in Dark phase NREM and significant increase in Dark phase REM, whereas lines 224-227 in the main text state "The duration of wakefulness episodes ... was significantly shorter than control mice during the light period, and the duration of NREM sleep episodes ...was significantly longer ... during the dark period (Fig. 3d)". Which one is correct? Please check. *

      We apologize for this typo and unclear description. We revised the sentence regarding Fig. 3d as follows:

      On page 10, line 242-246

      "The duration of wakefulness episodes in DMH-Prdm13-KO mice was significantly shorter than control mice during the light period between ZT0 to ZT2. The duration of NREM sleep episodes in DMH-Prdm13-KO mice was significantly shorter than control mice during the dark period (Fig. 3d). These results indicate that DMH-Prdm13-KO mice showed mild sleep fragmentation compared with control mice."

      • Fig. 5f, Y-axis title should be EEG SWA. * We corrected it.
    2. Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      In this study, Tsuji et al. demonstrate that Prdm13 signaling is involved in the regulation of sleep-wake pattern. They also identified Prdm13 as a transcription factor in the DMH neurons.

      Major comments:

      1. The evidence presented in Fig. 1 of age-related sleep fragmentation is potentially problematic. Although many previous studies have demonstrated fragmented sleep, especially fragmentation of NREM sleep, in aged mice compared to young mice, the data here do not suggest NREM fragmentation, because no change in the NREM bout duration was found. REM, on the other hand, may indeed have fragmentation during the dark phase, but REM only takes a small portion of the total sleep. Therefore, the conclusion that sleep is fragmented in old mice is not fully supported by Fig. 1. I noticed that the authors used 4-6 months old mice as the young group. Mice of this age can hardly be called "young". The females even start to have lowered fertility. This might be one of the reasons for the discrepancy between this and other studies. Repeating these experiments (and others involving the young group) with mice of more appropriate age (usually 2-3 months old) is recommended. Nonetheless, aging-caused sleep change is not new knowledge and has been reported repeatedly. This part of the results should be in the supplementary figures.
      2. The sleep phenotypes in aged mice and in Prdm13-KO mice are clearly distinct from each other. In the old mice (Fig. 1), REM sleep is fragmented but the total amount remains unchanged, and NREM sleep is increased (both bout number and total amount), indicating there may be more REM-to-NREM transitions, which the authors should quantify. However, Fig. 3 shows in Prdm13-KO mice, there is no REM fragmentation. In fact, it even seems to stabilize REM. But NREM duration is shorted, and no change in the total NREM or REM sleep time. These results suggest that the sleep alterations caused by aging and Prdm13-KO might have some overlap but are mostly in parallel and likely through different mechanisms. Therefore, the rationale of connecting Prdm13 signaling to aging-caused sleep changes is questionable. Is there a developmental change of Prdm13 expression in DMH between young and old mice? The authors also showed that Prdm13-KO in old mice caused decrease in NREM duration but has no effect on REM sleep, but in normal old mice, it is REM, but not NREM that has a defect. Prdm13 overexpression also only mildly decreased NREM bout number without affecting the episode duration of either NREM or REM, which can hardly be interpreted as "ameliorating sleep fragmentation". To me, all these results just suggest parallel actions of Prdm13 and aging on sleep, with Prdm13 mostly affecting NREM sleep but aging mostly impairing REM sleep.
      3. What is the control manipulation for sleep deprivation? The authors need to clarify this in the Methods. Also, sleep deprivation has confounding effects including but not limited to stress, food deprivation (since food was removed during SD), human experimenter (since a gentle-touch method was used). Without proper controls for these variables, the authors should avoid concluding that the changes they saw at cellular level are due to sleep loss.
      4. Identification of Prdm13+ cells using neuronal markers should be performed in addition to electrophysiological characterizations.
      5. Figs. 6 and 7 seem very disconnected from the main story. Identification of Prdm13 as a transcription factor is potentially interesting, but how does it account for its role in affecting sleep? The criteria of picking Cck, Grp and Pmch out of other candidate genes potentially regulated by Prdm13 and the rationale to investigate these genes seem unclear. More importantly, no evidence was shown regarding how Cck/Grp

      Minor comments:

      1. Please note on the images of Fig. 2d what the green fluorescence was. It is very confusing as is, given that it's surrounded by quantifications of c-fos in the figure.
      2. Please note use a different color for Prdm13 in several figure images (e.g., Fig. 2f, g, 7a,d, and Supplementary 2c). Yellow usually means overlap of red and green.
      3. Please note the statistic test results on power spectrum graphs.
      4. Inconsistency between the graphs in Fig. 3d and the description in the text. Fig. 3d suggests no change in Wake episode duration, significant decrease in Dark phase NREM and significant increase in Dark phase REM, whereas lines 224-227 in the main text state "The duration of wakefulness episodes ... was significantly shorter than control mice during the light period, and the duration of NREM sleep episodes ...was significantly longer ... during the dark period (Fig. 3d)". Which one is correct? Please check.
      5. Fig. 5f, Y-axis title should be EEG SWA.

      Significance

      General assessment: There are discrepancies in the evidence presented, and the results were poorly organized. I found the main conclusions of the manuscript not very convincing and the causal links among Prdm13, aging and sleep alterations weak.

      Advance: The identification of DMH Prdm13 in regulating sleep is potentially interesting and of some novelty, but the underlying mechanism and its causal relationship with aging were not clearly elucidated.

      Audience: basic research

      My expertise: sleep, social behavior, hypothalamus, dopamine neuromodulation, neural circuit development, synaptic organization.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      In this manuscript, the authors report dorsomedial hypothalamus-specific PR-domain containing protein 13-knockout (DMH-Prdm13-KO) mice recapitulated age-associated sleep alterations such as sleep fragmentation and increased sleep attempts during sleep deprivation (SD). These phenotypes were further exacerbated during aging, with increased adiposity and decreased physical activity, resulting in shortened lifespan. Moreover, overexpression of Prdm13 in the DMH ameliorated sleep fragmentation and excessive sleepiness during SD in old mice. They identified maintaining Prdm13 signaling in the DMH might play an important role to control sleep-wake patterns during aging. These findings are interesting and novel and the evidence they provided looks solid.

      Major comments

      1. The author spent a lot of words on Sirt1 in the introduction. Since Sirt1 regulates Prdm13, is there a link between the two in age-related sleep changes? If so, you can add some results and discussion.
      2. In Figure 2e, the author describes n=7-8 in the figure legend, but why do both groups on the column show eight data? Is there something wrong with the statistics? Please check the statistics in the article carefully.
      3. DMH is known as one of the major outputs of hypothalamus circadian system and is involved in the circadian regulation of sleep-wakefulness (J.Neurosci. 23, 10691-10702 ; Nat Neurosci 4:732-738). Does Prdm13 correlate with circadian rhythms? The author can add relevant content to the discussion

      Minor comments

      1. The immunohistochemical diagram in the paper is not representative enough, as shown in FIG. 2b and c.
      2. In FIG. 5h, the authors showed that the effect of overexpression of Prdm13 was very obvious, but the expression range of the virus after injection was lacking. Is there a fluorescent gene such as GFP on the virus to directly see the expression of the virus in the brain?
      3. Were mice singly housed or housed in groups?
      4. The part of sleep analysis needs to be further refined. How can REM and NREM in mice be distinguished and according to what criteria?
      5. The authors may consider adding more recent literature related to DMH and sleep, such as DOI: 10.1093/cercor/bhac258

      Significance

      Akiko Satoh's 2015 article "Deficiency of Prdm13, a dorsomedial hypothalamus-enriched gene, mimics age-associated changes in sleep quality and adiposity "influenced the novelty of the study, but the authors went further in terms of details and mechanisms. The audience of the basic research will be influenced by this research.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      1. General Statements [optional]

      We would like to thank the reviewers for taking time in reviewing and commenting on our paper. The comments were very constructive and conscientious, thanks to their expertise in the field. These comments and the revisions would surely make this paper a better and more robust finding in the field.

      The comments were about clearer explanations, increasing the quality of the data and additional experiments for a stronger conclusion, all of which we are eager to accomplish. Now we have sorted out the problems and planned the experiments required in the revision, as detailed below.

      2. Description of the planned revisions

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

      Summary In this manuscript, Komori et al. examined the role of the LRRK2 substrate and regulator Rab29 in the lysosomal stress response. Briefly, in chloroquine (CQ)-treated HEK293 cells the authors observed an apparent LRRK2-independent increased in Rab29 phosphorylation which was accompanied by translocation of Rab29 to lysosomes. Intriguingly, the authors detected a similar increase in Rab29 phosphorylation when Rab29 was tethered to lysosomes in the absence of CQ treatment. Using mass spectrometry, mutagenesis and a phospho-specific anti-body, the authors mapped the CQ-induced phosphorylation site to S185 and demonstrated its independence from LRRK2. Next, the authors found that PKCa was the kinase responsible for S185 phosphorylation and lysosomal translocation of Rab29. Lastly, the authors showed that in addition to PKCa the lysosomal translocation of Rab29 was also regulated by LRRK2. Overall, Komori and colleagues provide interesting new insights into the phosphorylation-dependent regulation of Rab29. However, there are. Number of technical and conception concerns which should be addressed.

      Major points 1) Figure 1F: the localization of Rab29 to lysosomes is not convincing at all. The authors should either provide more representative image examples or image the cells at a higher resolution. The authors should also confirm the CQ-induced lysosomal localization of Rab29 in a different cell type (e.g., HEK293).

      We will replace Fig 1F pictures with slightly more magnified images with higher resolution. We will also include additional cell types (HEK293, and other cells that are predicted to express endogenous Rab29); Reviewer #2 also raised this point (see Reviewer #2 comment on Significance).

      Moreover, the authors should show that prenylation of Rab29 is required for its CQ-induced phosphorylation.

      We will test the effect of lovastatin, a HMG-CoA reductase inhibitor that causes the depletion of the prenylation precursor geranylgeranyl diphosphate from cells (Binnington et al., Glycobiology 2016, Gomez et al, J Cell Biol 2019), or 3-PEHPC, a GGTase II specific inhibitor that also causes the inhibition of Rab prenylation (Coxon FP et al, Bone 2005).

      2) The rapalog-induced increase in Rab29 phosphorylation in Figure 2D is not convincing since there is at least 2-3-fold more Rab29 in FRB-LAMP1 expressing cells compared to their FRB-FIS1 counterparts. An independent loading control is also missing. This is a key experiment and should be properly controlled and quantified. In addition, can CQ treatment drive 2xFKBP GFP-Rab29 from mitochondria to lysosomes (in the presence of rapalog and FRB-Fis1)?

      We will carefully examine another round of rapalog-induced phosphorylation of Rab29, with an independent loading control such as alpha-tubulin. The immunoblot analysis will be made against the intensity of non-p-Rab29. The response to the latter question was described in the section 4 below.

      3) Figure 4A-C: Are these stable Rab29 expressing cells? If not, the quantification of "the size of largest lysosome in EACH cell" becomes very problematic. This analysis should be repeated with stable Rab29 variant cells in a background lacking endogenous Rab29. Furthermore, the LAMP1 signal is too dim to see any convincing colocalization (e.g., with WT) or the lack thereof (e.g., in the case of S185D).

      The cells shown in Figure 4 are HEK293 cells transiently expressing Rab29, and the issue of quantification was described in the section 3 below. We agree that the signal of LAMP1 was dim, and it turned out that the confocal microscope we used had problems with the sensitivity of the red channels. We will be taking another round of these images with a new confocal microscope.

      Lastly, the authors should corroborate their findings with an ultrastructural analysis since the electron microscopy would definitively be more suitable for this type of measurements.

      We are planning to obtain electron microscopic images, according to this reviewer’s request. We plan to invite an expert in electron microscopy analysis as a co-author.

      4) The lysosomal colocalization of Rab29 in Figure 5C is again not convincing. This analysis needs to be repeated with high resolution imaging.

      Again, we will repeat this experiment with a new confocal microscope, with the hope that it would yield better images.

      5) The authors need to show the level of LRRK2 depletion (Figure 6). Given the role of LRRK2 in driving lysosomal Rab29 translocation, the importance of the LRRK2 independent pS185 for this process remains unclear.

      We will add the level of LRRK2 on its knockdown; we have experienced that LRRK2 knockdown usually occurs with more than 50% efficiency every time. The response to the latter comment was described in the section 3 below.

      6) In general, the authors employ an alternative, biochemical assay (e.g., LysoIP) for the lysosomal translocation of Rab29. This would in particular help to clarify the effect of the Rab29 variants and LRRK2 inhibition.

      We have previously shown that the overexpressed Rab29 (and LRRK2) is enriched in the lysosomal fraction from CQ-treated cells, which was performed using dextran-coated magnetite (Eguchi et al, PNAS 2018). Using the same biochemical method, we will show the enrichment of endogenous Rab29 in the lysosomal fraction.

      Minor points

      9) Figure 2C is lacking the control IF staining for mitochondria (to which 2xFKBP-GFP-Rab29 is assumed be recruited upon co-expression with FRB-FIS1).

      We will stain the cells with MitoTracker to ensure that anchoring away of 2xFKBP-GFP-Rab29 by FRB-Fis1 results in mitochondrial localization.

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

      The data in the manuscript convincingly demonstrates that lysosomal overload by Chloroquine treatment induces Rab29 localisation to the lysosomes and that this membrane association is dependent on PKCalpha-dependent phosphorylation at Ser185.

      We have a number of rather minor comments listed below:

      Figure 2

      The increasing levels of non-phosphorylated Rab29 over the indicated time course of AP21967 treatment in Figure 2B are concerning. First, could you provide an explanation for this clear increase in both non-p-Rab29 and p-Rab29 in the phostag but not the normal gel? Second, could all quantifications of p-Rab29 be made relative to the non-p-Rab29?

      We will try another round of rapalog-induced phosphorylation of Rab29, with an independent loading control. The immunoblot analysis will be made against the intensity of non-p-Rab29. Reviewer #1 raised a similar concern on Figure 2D.

      Figure 5

      To further demonstrate that PKCalpha phosphorylates endogenous Rab29 at Ser185, we recommend reperforming the Go3983/PMA treatment in figure B with the anti-p-Ser185 antibody. It may be sufficient to perform the treatment only at 4 or 8 hours, simply to provide stronger evidence regarding the phosphorylation of endogenous Rab29.

      We will give a try, although the anti-phosphorylated protein antibodies that we tried never worked for phos-tag SDS-PAGE. With the conventional western blot, we will be able to try this experiment.

      It is not clear whether the activity of PMA in the assay is due to inhibition of PKCalpha. Are the effects ablated by PKCalpha KD

      We will test the knockdown of PKCalpha, beta, gamma and delta by siRNAs to further narrow down the effects of PKC-dependent phosphorylation of Rab29.

      Reviewer #2 (Significance (Required)):

      These cell biology findings are important in the field as both Rab29 and LRRK2 are implicated in the pathogenesis of Parkinson disease. The phosphorilation of Ser185 of Rab29 by PKCalpha is novel and contributes to our understanding of Rab29 and LKRR2 regulation. One limitation of the study is that is conducted in only two cell types quite unrelated to the disease, so how general and disease relevant are the findings it is not clear. Most of the data are solid. There are two experiments whose results are difficult to interpret and a few controls missing. Also a few issues with quantifications, all of which is described in details above and will need to be fixed prior to publication. My expertise for this paper is in the cell biology of lysosomal function.

      The issue that only two cell types were analyzed was also raised by reviewer #1, so we will examine additional cell types, especially those that are predicted to express endogenous Rab29. Our responses to other issues raised are described elsewhere. Thank you for these insightful comments.

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

      Figure 4A-C: Are these stable Rab29 expressing cells? If not, the quantification of "the size of largest lysosome in EACH cell" becomes very problematic. This analysis should be repeated with stable Rab29 variant cells in a background lacking endogenous Rab29. (Reviewer #1)

      As described in the section 2 above, the cells shown in Figure 4 are HEK293 cells transiently expressing Rab29. We are sorry that the description “the size of largest lysosome in each cell” was misleading. As we analyzed only cells overexpressing GFP-Rab29 that were marked with GFP fluorescence, we believe that transient expression should not be a problem. To avoid any misunderstandings, we have described in Figure 4 legends that only lysosomes in Rab29-positive cells (and all cells expressing Rab29) were included in the analysis of the largest lysosome of each cell.

      Regarding the effect of endogenous Rab29 in Figure 4 experiments, Reviewer #2 similarly raised the issue on whether Rab29 phosphomimetics are acting as dominant active, preventing lysosomal enlargement. On this point, we have previously reported that knockdown of endogenous Rab29 causes the enhancement of lysosomal enlargement upon CQ treatment (Figure 5I,J of Eguchi et al, PNAS 2018), suggesting that the lysosome-deflating effect by phosphomimetics is a dominant active effect rather than dominant negative suppressing endogenous Rab29. This point is considered significant, and thus has been explained in the results section (page 7, lines 168-171).

      Along similar lines: why not all cells in Figure 5E and Figure 5G show Rab29- and LRRK2-positive structures? How do the authors know which of these phenotypes is the prevalent one? (Reviewer #1)

      As for the ratio of cells with Rab29- and LRRK2-positive structures, it seems reasonable given that different cells have different levels of exposure to lysosomal stress and that the response is transient and does not occur simultaneously. The ratio of these positive cells may also vary depending on the cell culture conditions. Since Rab29- and LRRK2-positive structures are rarely seen in control cells, we think this would be a meaningful phenotype even if only 20-30% of cells show such structures. The result that the ratio of localization changes is not 100% is now noted in the results section explaining Figure 1G (page 4-5, lines 108-110) where the immunocytochemical data first appears.

      Given the role of LRRK2 in driving lysosomal Rab29 translocation, the importance of the LRRK2 independent pS185 for this process remains unclear. (Reviewer #1)

      Our data suggested that Rab29 is stabilized on lysosomes only when LRRK2-mediated phosphorylation and S185 phosphorylation both occur on Rab29 molecule (as shown in Figure 7 scheme), so we believe there is no contradiction. We have now described more clearly about this notion at the end of the results section (page 9, lines 235-236).

      It is not clear what the authors mean by "lysosomal overload stress". Since mature lysosomal incoming pathways such as autophagy or endocytosis are disrupted by CQ, it is difficult to picture an overload. Maybe rephrasing would help to clarify this. (Reviewer #1)

      Chloroquine (CQ) is known as a lysosomotropic agent that accumulates within acidic organelles due to its cationic and amphiphilic nature, causing lysosome overload and osmotic pressure elevation, and this is what we call “lysosomal overload stress”. The well-known effects of CQ to disrupt lysosomal incoming pathways are ultimately caused by the above consequences. Also, we have previously reported that lysosomal recruitment of LRRK2 is caused by CQ but not by bafilomycin A1, the latter being an inducer of lysosomal pH elevation, or by vacuolin-1 that enlarges lysosomes without inducing lysosomal overload/pH elevation (Eguchi et al, PNAS 2018), and further found that not only CQ but also other lysosomotropic agents commonly induced LRRK2 recruitment (Kuwahara et al, Neurobiol Dis 2020). We thus have described the effect of CQ as “overload”. However, it is true that we have not provided a clear explanation for readers, so we have added some notes for lysosomal overload stress in the introduction section (page 3, lines 69-71).

      Which cell type is used for the IF analysis in Figure 2C? This information is in general quite sparse. The authors should clearly state the cell type for each experiment/Figure. (Reviewer #1)

      We have added cell type information that was missing in several places in the manuscript. We are very sorry for the inconveniences. For clarification, HEK293 cells were used in Figure 2C.

      Are the images in figure 1F representative? i.e. does Rab29 always colocalise to such enlarged lysosomes upon CQ treatment and does CQ treatment always drastically alter the cellular distribution of Rab29? (Reviewer #2)

      The images in Figure 1F are representative of when Rab29 is recruited, but it is not seen in all cells, and the ratio of recruitment (~80%) is shown in Figure 1G. Reviewer #1 also asked why Rab29 recruitment is not seen in all cells, and we gave the same answer above. It may be reasonable to speculate that different cells have different levels of exposure to lysosomal stress and that the response is transient and does not occur simultaneously. The ratio of these positive cells may also vary depending on the cell culture conditions. For the readers’ clarity, we have added that the ratio of localization change of Rab29 is not 100% and is comparable to that of LRRK2 previously reported (page 4-5, lines 108-110).

      Considering that the "forced localisation technique" induces a non-physiological colocalization of non-endogenous Rab29 to lysosomes, it may be an overestimation to conclude just from these data that phosphorylation of Rab29 occurs on the lysosomal surface. This is also quite in contrast with the later finding that phosphorylation by PKCalpha promotes lysosome localization of Rab29. It seems more reasonable to conclude that Rab29 can be phosphorylated when localised at the lysosomes (as opposed to other organelles such as mitochondria). If the authors feel strongly about this point they might need to find a less non-physiological assay. (Reviewer #2)

      Yes, it could be an overestimation, and as we do not have better means to conduct a less non-physiological assay, we have modified the description from “occurred on the lysosomal surface” to “could occur on the lysosomal surface” (page 5, line 112 (subtitle) and line 128).

      Regarding the comparison with the later finding that phosphorylation by PKCalpha promotes lysosome localization of Rab29, these data (Figure 2 and 5) could be explained with a single speculation: phosphorylation of Rab29 on lysosomal membranes could retain Rab29 on the membranes for a longer time. It is not easy to decipher which comes first, association with membranes or phosphorylation of Rab29, in a physiological assay, but considering reports that show PKCalpha activation happens on membranes (Prevostel et al., J Cell Sci 2000), at least the data favor our conclusion over the idea of PKCalpha phosphorylating Rab29 in the cytoplasm and then promoting lysosomal localization. This point is now clearly described in the discussion (page 10, lines 248-251).

      It is not clear how the Rab29 phosphomimetics are acting as dominant active preventing lysosomal enlargement. Authors should speculate or repeat the experiments in absence of endogenous Rab29 to clarify the matter. (Reviewer #2)

      A similar concern about the effect of endogenous Rab29 was also raised by Reviewer #1 (see above). We have previously reported that knockdown of endogenous Rab29 causes the enhancement of lysosomal enlargement upon CQ treatment (Figure 5I,J of Eguchi et al, PNAS 2018), suggesting that the lysosome-deflating effect by phosphomimetics is a dominant active effect rather than dominant negative suppressing endogenous Rab29. This point is considered important and thus has been explained in the results section (page 7, lines 168-171).

      Overall, there is some missing information regarding repeats for Western blots, such as those in figure 3C, 3D and 3E. Please add indications about repeats in the figure legend or methods. (Reviewer #2)

      We have added the repeat information to each figure legend where it was missing. We are very sorry for the inconveniences.

      The model in figure 7 however seems to suggest that Rab29 associates to lysosomal membranes independently, and is then stabilised at the membranes by LRRK2 and PKCalpha - a point which is not directly supported by the data. (Reviewer #2)

      As noted earlier, we consider that phosphorylation of Rab29 on lysosomal membranes could retain Rab29 on the membranes for a longer time, given the present data and previous reports that phosphorylation of Rab29 is more likely to happen on the lysosomal membrane than in the cytosol. Also, as inhibition of either of the two phosphorylations ends up in disperse Rab29 localization, we have made this figure as a model of what is plausible right now. This explanation is now added in the discussion (page 10, lines 248-251).

      English proofreading should be improved: "CQ was treated to HEK293" (page 4), "As we assumed that this phosphorylation is independent of LRRK2" as an opening line (page 5) (Reviewer #2)

      Thank you for pointing out these incorrect wordings. They were corrected.

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

      In addition, can CQ treatment drive 2xFKBP GFP-Rab29 from mitochondria to lysosomes (in the presence of rapalog and FRB-Fis1)? (Reviewer #1)

      We do not think that a comparison between the affinities of FKBP-rapalog-FRB and Rab29-[unknown factor that directs Rab29 to lysosomes] is necessary, as the former has a Kd in the single digit nM range (Banaszynski et al, JACS 2005), whereas the latter (based on estimations from related PPIs) is estimated to be in the μM range, which shows a much weaker affinity than the former (McGrath et al, Small GTPases 2019). Furthermore, even if Rab29 appears to have migrated from mitochondria to lysosomes as a result of this experiment, one cannot rule out the possibility that a small portion of the mitochondrial membrane was incorporated into the lysosomal membrane that was enlarged by CQ treatment.

      Molecular weight markes should be provided for all immunoblot experiments. (Reviewer #1)

      The immunoblot pictures without molecular weight markers in our paper are all Phos-tag SDS-PAGE blot analyses. Phos-tag SDS-PAGE results in band shifts of phosphorylated proteins, and writing in markers would be misleading. Moreover, previous representative studies heavily using Phos-tag (e.g., Kinoshita et al, Proteomics 2011, Ito et al, Biochemical Journal 2016) also did not show the molecular weight markers. Here we performed phos-tag SDS-PAGE analysis only to find differences in the phosphorylation state of Rab proteins.

      The use of the quantification ratio of cells with Rab29-positive lysosomes in figure 1G might be slightly misleading as it does not allow the reader to understand to what extent Rab29 localisation at lysosomes upon CQ treatment. We recommend using a simpler quantification, such as by measuring the average colocalisation of Rab29 and LAMP1 per cell. (Reviewer #2)

      For figure 5D and 5F, As with figure 1G, we recommend using a more straightforward and impartial method of quantification such as simply measuring the colocalisation of Rab29 with LAMP1. (Reviewer #2)

      Popular colocalization analyses using Pearson’s or Mander’s coefficients would be a good choice if the amounts of Rab29 varied greatly between lysosomes. However, this may not apply in this case; the amount of Rab29 or LRRK2 on each lysosome is considered to saturate quickly and a relatively low amount of them may not be detected on immunofluorescence observations, whereas the probability of finding these structures has been shown to exhibit a moderate sigmoid curve (as seen in Figure 1E or 2H of Eguchi et al., PNAS 2018). Therefore, the amount of Rab29 or LRRK2 could be approximated to a Bernoulli distribution in terms of colocalization with lysosomes, and this is the reason why we chose to quantify “the ratio of cells with Rab29-positive lysosomes”.

      We recommend using a more transparent and simple quantification method, such as average size of lysosomes per cell. (Reviewer #2)

      As one can see in the inset of Figure 4B, unenlarged lysosomes are unfortunately too small for the quantification of their size, much less tell two small lysosomes apart in our experimental settings and laboratory resources, so we decided to analyze the largest lysosome in each cell as a representative of the cells to minimize measurement errors. This measurement only includes GFP-Rab29 positive cells, and by comparing against CQ-untreated cells we intended to increase the validity of this analysis. This quantification method was also used in our previous report (Eguchi et al, PNAS 2018).

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      The data in the manuscript convincingly demonstrates that lysosomal overload by Chloroquine treatment induces Rab29 localisation to the lysosomes and that this membrane association is dependent on PKCalpha-dependent phosphorylation at Ser185.

      We have a number of rather minor comments listed below:

      Figure 1

      • Are the images in figure 1F representative? i.e. does Rab29 always colocalise to such enlarged lysosomes upon CQ treatment and does CQ treatment always drastically alter the cellular distribution of Rab29?
      • The use of the quantification ratio of cells with Rab29-positive lysosomes in figure 1G might be slightly misleading as it does not allow the reader to understand to what extent Rab29 localisation at lysosomes upon CQ treatment. We recommend using a simpler quantification, such as by measuring the average colocalisation of Rab29 and LAMP1 per cell.

      Figure 2

      • Considering that the "forced localisation technique" induces a non-physiological colocalization of non-endogenous Rab29 to lysosomes, it may be an overestimation to conclude just from these data that phosphorylation of Rab29 occurs on the lysosomal surface. This is also quite in contrast with the later finding that phosphorylation by PKCalpha promotes lysosome localization of Rab29. It seems more reasonable to conclude that Rab29 can be phosphorylated when localised at the lysosomes (as opposed to other organelles such as mitochondria). If the authors feel strongly about this point they might need to find a less non-physiological assay.
      • The increasing levels of non-phosphorylated Rab29 over the indicated time course of AP21967 treatment in Figure 2B are concerning. First, could you provide an explanation for this clear increase in both non-p-Rab29 and p-Rab29 in the phostag but not the normal gel? Second, could all quantifications of p-Rab29 be made relative to the non-p-Rab29?

      Figure 3

      • It is not clear how the Rab29 phosphomimetics are acting as dominant active preventing lysosomal enlargement. Authors should speculate or repeat the experiments in absence of endogenous Rab29 to clarify the matter.
      • Overall, there is some missing information regarding repeats for Western blots, such as those in figure 3C, 3D and 3E. Please add indications about repeats in the figure legend or methods.

      Figure 4

      • We recommend using a more transparent and simple quantification method, such as average size of lysosomes per cell.

      Figure 5

      • To further demonstrate that PKCalpha phosphorylates endogenous Rab29 at Ser185, we recommend reperforming the Go3983/PMA treatment in figure B with the anti-p-Ser185 antibody. It may be sufficient to perform the treatment only at 4 or 8 hours, simply to provide stronger evidence regarding the phosphorylation of endogenous Rab29.
      • It is not clear whether the activity of PMA in the assay is due to inhibition of PKCalpha. Are the effects ablated by PKCalpha KD
      • For figure 5D and 5F, As with figure 1G, we recommend using a more straightforward and impartial method of quantification such as simply measuring the colocalisation of Rab29 with LAMP1.

      Figure 6

      • Again, we recommend altering the methods of quantification

      Figure 7

      • The model in figure 7 however seems to suggest that Rab29 associates to lysosomal membranes independently, and is then stabilised at the membranes by LRRK2 and PKCalpha - a point which is not directly supported by the data.

      English proofreading should be improved: "CQ was treated to HEK293" (page 4), "As we assumed that this phosphorylation is independent of LRRK2" as an opening line (page 5),

      Significance

      These cell biology findings are important in the field as both Rab29 and LRRK2 are implicated in the pathogenesis of Parkinson disease. The phosphorilation of Ser185 of Rab29 by PKCalpha is novel and contributes to our understanding of Rab29 and LKRR2 regulation. One limitation of the study is that is conducted in only two cell types quite unrelated to the disease, so how general and disease relevant are the findings it is not clear. Most of the data are solid. There are two experiments whose results are difficult to interpret and a few controls missing. Also a few issues with quantifications, all of which is described in details above and will need to be fixed prior to publication. My expertise for this paper is in the cell biology of lysosomal function.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary

      In this manuscript, Komori et al. examined the role of the LRRK2 substrate and regulator Rab29 in the lysosomal stress response. Briefly, in chloroquine (CQ)-treated HEK293 cells the authors observed an apparent LRRK2-independent increased in Rab29 phosphorylation which was accompanied by translocation of Rab29 to lysosomes. Intriguingly, the authors detected a similar increase in Rab29 phosphorylation when Rab29 was tethered to lysosomes in the absence of CQ treatment. Using mass spectrometry, mutagenesis and a phospho-specific anti-body, the authors mapped the CQ-induced phosphorylation site to S185 and demonstrated its independence from LRRK2. Next, the authors found that PKCa was the kinase responsible for S185 phosphorylation and lysosomal translocation of Rab29. Lastly, the authors showed that in addition to PKCa the lysosomal translocation of Rab29 was also regulated by LRRK2. Overall, Komori and colleagues provide interesting new insights into the phosphorylation-dependent regulation of Rab29. However, there are. Number of technical and conception concerns which should be addressed.

      Major points

      1. Figure 1F: the localization of Rab29 to lysosomes is not convincing at all. The authors should either provide more representative image examples or image the cells at a higher resolution. The authors should also confirm the CQ-induced lysosomal localization of Rab29 in a different cell type (e.g., HEK293). Moreover, the authors should show that prenylation of Rab29 is required for its CQ-induced phosphorylation.
      2. The rapalog-induced increase in Rab29 phosphorylation in Figure 2D is not convincing since there is at least 2-3-fold more Rab29 in FRB-LAMP1 expressing cells compared to their FRB-FIS1 counterparts. An independent loading control is also missing. This is a key experiment and should be properly controlled and quantified. In addition, can CQ treatment drive 2xFKBP GFP-Rab29 from mitochondria to lysosomes (in the presence of rapalog and FRB-Fis1)?
      3. Figure 4A-C: Are these stable Rab29 expressing cells? If not, the quantification of "the size of largest lysosome in EACH cell" becomes very problematic. This analysis should be repeated with stable Rab29 variant cells in a background lacking endogenous Rab29. Furthermore, the LAMP1 signal is too dim to see any convincing colocalization (e.g., with WT) or the lack thereof (e.g., in the case of S185D). Lastly, the authors should corroborate their findings with an ultrastructural analysis since the electron microscopy would definitively be more suitable for this type of measurements.
      4. The lysosomal colocalization of Rab29 in Figure 5C is again not convincing. This analysis needs to be repeated with high resolution imaging. Along similar lines: why not all cells in Figure 5E and Figure 5G show Rab29- and LRRK2-positive structures? How do the authors know which of these phenotypes is the prevalent one?
      5. The authors need to show the level of LRRK2 depletion (Figure 6). Given the role of LRRK2 in driving lysosomal Rab29 translocation, the importance of the LRRK2 independent pS185 for this process remains unclear.
      6. In general, the authors employ an alternative, biochemical assay (e.g., LysoIP) for the lysosomal translocation of Rab29. This would in particular help to clarify the effect of the Rab29 variants and LRRK2 inhibition.

      Minor points

      1. It is not clear what the authors mean by "lysosomal overload stress". Since mature lysosomal incoming pathways such as autophagy or endocytosis are disrupted by CQ, it is difficult to picture an overload. Maybe rephrasing would help to clarify this.
      2. Which cell type is used for the IF analysis in Figure 2C? This information is in general quite sparse. The authors should clearly state the cell type for each experiment/Figure.
      3. Figure 2C is lacking the control IF staining for mitochondria (to which 2xFKBP-GFP-Rab29 is assumed be recruited upon co-expression with FRB-FIS1).
      4. Molecular weight markes should be provided for all immunoblot experiments.

      Significance

      Please see above.

    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      1. General Statements [optional]

      We are grateful to the reviewers for highlighting the novelty of the mechanism we describe for P2Y2 in driving RGD-binding integrin-dependent invasion, and acknowledging its potential in cancer therapy. We thank the reviewers for their valuable and detailed comments, which have allowed us to prepare a significantly stronger and clearer manuscript.

      2. Point-by-point description of the revisions

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


      Summary

      The study identifies P2Y2 as a purinergic receptor strongly associated with hypoxia, cancer expression and survival. A link is found between P2Y2-integrin interaction and cancer invasion, highlighting this as a novel therapeutic target. The mechanism is interesting and general well explored.

      • *

      We thank the reviewer for acknowledging the novelty of the therapeutic target presented in this work.

      • *

      Minor comments

      As P2Y2 is highly expressed by other cell types found with tumours, including vascular endothelium and leukocytes, the authors should reflect on this as a confounding factor in the analysis of adrenocarcinoma gene expression analysis. I appreciate the RNAscope work may resolve this issue to some extent.

      We agree that P2Y2 is known to be expressed in other cell types. RNAscope did not show convincing staining in PDAC normal adjacent tissue (was similar to negative staining), perhaps due to the challenging nature of pancreatic tissue with respect to RNA degradation. We have resolved this issue by including single cell RNA-seq of normal human pancreas for P2Y2 from Protein Atlas (Sup. Fig. 2B), which shows expression in several cell types, mainly endocrine cells, and macrophages. We now mention this in line 142 : “P2Y2 is known to be expressed at low levels in normal tissues but interestingly RNAscope did not detect this. This data suggest 1) the lower limits of the technique compounded by the challenge of RNA degradation in pancreatic tissue and 2) supports that in tumour tissue where it was detected there was indeed overexpression of P2Y2, in line with the bioinformatic data. Interrogating single cell P2Y2 RNA expression in normal PDAC from proteinatlas.org (Karlsson et al., 2021), expression was found at low levels in several cells types, for example in endocrine cells and macrophages (Sup. Fig. 2B).”

      Major comments

      • *

      The authors correctly identify that the level of ATP in the tumour microenvironment can be very high, typically 100uM or so. However, these concentrations are supramaximal for P2Y2 activation, at which ATP has an approximate EC50 of 100nM. Coupled with the fact that many cell types, including cancer cells, constitutively secrete ATP, there is an opportunity to explore the effects of lower ATP concentrations in some assays, or provide some concentration-response relationship to give more confidence of P2Y2-dependent effects.

      • *

      We thank the reviewer for raising this point and we agree that 100 uM can be a high concentration, albeit one that is frequently used throughout the literature. We have now included a concentration-response relationship (Sup. Fig. 2D) showing that ATP causes cytoskeletal changes that are P2Y2 dependent most prominently at 100 uM, the concentration that, as the reviewer has also corroborated, is similar to the concentration of ATP found in tumours.

      Also, the authors describe the use of cancer cells where P2Y2 has been knocked out using CRISPR. Does this KO have an effect on cancer invasion? The effect of ARC should be absent in these cells and give confidence the effects of ARC are P2Y2-dependent, as some off-target effects of this antagonist have been reported. To explore the influence of constitutive P2Y2 activity, the authors should explore the effects of ARC alone in some assays.

      We agree that including more AR-C only experiments would be informative, so we have included a 3D sphere invasion assay with our CRISPR cell line treated with and without AR-C that shows no effect in invasion (p = 0.4413) (Sup. Fig. 3J). We have now also included images of AsPC-1 cells transfected with Lifeact, showing no changes in morphology with AR-C only (Sup. Fig. 2E). We apologise for missing a ‘+’ in one of the supplementary figures which shows AR-C only in AsPC-1 cells has no effect on its own.

      The effects of the CRISPR cell line in invasion are shown in Fig. 3F, showing a significant reduction (p = 0.0005) in invasion.

      The title of the manuscript implies extracellular ATP drives cancer invasion, though in my opinion this statement is not fully explored. Though ATP/UTP are applied at supramaximal concentrations for P2Y2 activation, the influence of ATP in the cell culture microenvironment without exogenous application is not explored. One would predict that scavenging extracellular ATP with apyrase would negatively impact invasiveness and the proximity of integrin and P2Y2 without ATP/UTP application if constitutively secreted ATP is involved. Pharmacological manipulation of ectonucleotidase activity is an alternative. Experimental route to explore this.

      We agree and have changed the title of our article to “Purinergic GPCR-integrin interactions drive pancreatic cancer cell invasion”. Our 3D sphere experiments with the CRISPR cell line show a reduction in invasion without exogenous application of ATP, which we also see to a lesser extent in our siRNA P2Y2 cell line. We have tested our sphere model with apyrase but unfortunately, the buffer used for apyrase to work is not compatible with our gel composition. Pharmacological manipulation would be a very good alternative if the cells used expressed high levels of CD39 or PANX1, which unfortunately they don’t. We hypothesise that most basal extracellular ATP in our 3D spheres comes from hypoxic areas that cause cell death, just as is postulated for tumours.

      Immunoprecipitation experiments of native proteins would be more convincing data that P2Y2 and integrin physically interaction, as opposed to being in close proximity. This would also overcome artifacts of interaction that can be attributed to receptor overexpression.

      We attempted immunoprecipitation experiments but unfortunately ran into several technical difficulties, including the anti-aV antibody working poorly for Western blot. Immunoprecipitation of these proteins has been reported by others (PMID: 25908848), supporting the proposed interaction.

      DNA-PAINT super resolution microscopy allows for quantification of nanoscale distances, and we used this to calculate the distances where physical interaction occurs. The possibility of this close proximity being by chance is accounted for in the computational nearest neighbour distance calculation by calculating points randomly distributed. This random distribution calculation also helps in overcoming artifacts of interaction due to overexpression, as the random distributed points are the same number of points as the proteins detected in each condition for each region of interest. Importantly, we also performed DNA-PAINT in using untransfected AsPC-1 thus endogenous levels (no receptor overexpression or alteration) and saw similar results (Sup. Fig.4A-D), thus we are confident of the interactions reported.

      Finally, we alter the RGD motif, which underpins the physical interaction, and see significant changes that match observations in previous publications using the P2Y2 agonist UTP, mentioned in the discussion: Line 398 “Following ATP stimulation, the number of P2Y2 proteins at the plasma membrane decreased significantly after one hour, implying receptor internalisation, in line with previous work showing P2Y2 at the cell surface was reduced significantly after one hour of UTP stimulation (Tulapurkar et al., 2005).” and Line 408: “P2Y2 affecting cell surface redistribution of αV integrin has been reported, with αV integrin clusters observed after 5 min stimulation with UTP (Chorna et al., 2007)”

      It is currently not clear what the mechanistic relationship between P2Y2 activity, P2Y2-integrin proximity and RGD motif is. Do the authors suggest the RGD domain becomes exposed upon receptor activation? The mechanism is not fully articulated in the discussion.

      We apologise for any lack of clarity in our postulated mechanism, we have now included a more detailed explanation of the mechanism in the discussion : Line 417 “We speculate that by reducing the ability of integrins to bind to the RGD of P2Y2, through receptor internalisation, RGE mutation or through cRGDfV treatment, there is less RGD-triggered integrin endocytosis, hence less integrin recycling and an increase of integrins at the cell surface.”

      Reviewer #1 (Significance (Required)):


      General assessment: A novel mechanism is presented for therapeutic intervention of cancer. The study relies on supramaximal concentrations of agonist and overexpressed receptors. Role of endogenous P2Y2 not fully explored. The study lacks in vivo evidence of the importance of this mechanisms. Cell developed in the study could be used in mouse models to explore effect on tumour growth.

      Advance: Integrin and P2Y2 interactions are already documented but not in context of cancer.

      Audience: basic research

      We thank the reviewer for crediting this work as a novel mechanism for therapeutic intervention of cancer. We trust that the new data provided (as discussed above) have resolved the concerns of the reviewer as we now have provided an explanation for the concentrations used. We do rely on overexpressed receptors for a small portion of our experiments, however, all experiments with overexpressed receptors were then tested in cells with endogenous expression of P2Y2 and used pharmacological means to show the same behaviour. We have now clarified this. We have also included in the discussion a sentence about the mouse experiment performed by Hui et al. with regards to reduced tumour growth when targeting P2Y2: Line 365: “Combination treatment of subcutaneous xenografts of AsPC-1 or BxPC-3 cells with the P2Y2 antagonist AR-C together with gemcitabine significantly decreased tumour weight and resulted in increased survival compared to placebo or gemcitabine monotherapy control (Hu et al., 2019).”

      • *

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

      Summary:

      Considering the fact that most PDAC are characterized by a high level of extracellular purines content, authors decided to study the expression of the 23 genes coding for membrane proteins involved in the binding or transport of purines in available PDAC transcriptomic cohorts. This approach led to the identification of P2Y2, a GPCR, as the best predictor for the worst survival of patients. Using in vitro models, they show that P2Y2 expression is associated with increased invasion capacity of pancreatic cancer cells and that this pro-invasive effect is dependent on the interaction of P2Y2 with αV integrin via the RGD motif.

      Major comments:

      • It is not clear to me why authors decided at one point to perform a GSEA comparing low and high mRNA expression of P2Y2 and why they decided to focus on the potential interaction of P2Y2 with integrin αV. As a GPCR, activation of P2Y2 leads to the activation of several downstream signaling pathways that may directly impact the adhesion, migration, and invasion properties of cells. Moreover, despite the presence of the RGD motif in P2Y2, it is not excluded that it may bind (maybe more efficiently) to other "cell adhesion" molecules.

      We apologise if the link between the GSEA figure and focusing on the potential integrin interaction was not clear. We have now performed GSEA using the panther gene set library, which includes a “Integrin signalling pathway” gene set. This was the top ranked gene set in both cohorts and we have substituted the GSEA figure for this instead (Fig. 2D). We trust that the narrative of the manuscript and our rationale to pursue the importance of integrin interaction is now clear.

      We agree with the reviewer and believe that P2Y2 may bind to other molecules important in cell adhesion. We studied integrin interactions due to the clear relationship of P2Y2 and integrins in patient data, which was not as evident with other binding partners. Furthermore, this relationship is unexplored in cancer and offers novel therapeutic strategies.

      • Similarly, if αV can regulate P2Y2 signaling, what about the regulation of αV signaling pathways by P2Y2? αV integrin has to bind to a β subunit and, depending on the identity of the β subunit, may have distinct regulations and so different impact on cell invasion. How P2Y2 can interfere with these α/β ratios?

      We thank the reviewer for this comment, and have now included western blots showing the impact of P2Y2 treatment on integrin signalling through FAK and ERK (Fig 5). We agree that the β subunit may have distinct regulation and outputs, but this is outwith the scope of our current study.

      • While it has been shown in other studies, in this work, there is no real proof of the interaction between P2Y2 and αV. Only in Figure 4I, where the authors look at the NND We thank the reviewer for raising this point as it has made us realise that our chosen NND of * *

      • Surprisingly, in the absence of ATP, P2Y2 RGE mutant, which should no more interact with αV, show a 2 to 3 fold more vicinity to αV compared to WT P2Y2. How can the authors explain this?

      We agree that this is a suprising, but robust discovery. By altering the RGD motif, there may be less RGD-triggered integrin endocytosis, leading to increased integrins at the surface. We have included this hypothesis in the discussion in Line 417. The RGE mutation has less affinity to integrins, meaning it still retains some ability to bind to integrins. Hence by chance, a higher number of integrins will result in a higher number of interactions with the RGE. We are planning to interrogate the internalisation dynamics in a future study.

      • For DNA-PAINT experiments, the authors only focus on membrane proteins whose amounts are balanced by internalization, recycling and export from internal compartment. As claimed, but not demonstrated by the authors, interaction of P2Y2 and αV may interfere with all these steps, thereby increasing or decreasing the cell surface expression of both proteins. Hence, it would be useful to 1) control proteins levels by western blot, especially for the overexpressed P2Y2, to be sure that they are the same, 2) block internalization and/or export to decipher the important steps.

      • In fact, all these main questions are raised by the authors in the end of the discussion but so far, they only show that the RGD motif has an impact on the biological role of P2Y2 (cell invasion) and on the membrane dynamic of αV and itself.

      We thank the reviewer for the suggestions:

      • In the course of our attempts to perform co-IP for P2Y2 and aV we could confirm that P2Y2 expression levels were equivalent (see Fig below – for reviewers only), but the problems with anti-aV antibodies prevented completion of the experiment. We also show IF staining showing similar levels of P2Y2 for both overexpressed conditions (Sup. Fig. 3K).

      Figure: Immunoprecipitation of P2Y2 showing similar P2Y2 levels in AsPC-1 P2Y2CRISPR cells trasfected with P2Y2RGD or P2Y2RGE and treated with 100 µM of ATP or control for 1 hour. Antibody used: anti-P2Y2 (APR-010, Alomone Labs).

      • As the reviewer highlights, in this work we have focused on the role of P2Y2 in PDAC invasion and have looked at single-molecule resolution membrane dynamics of αV and P2Y2. The different steps of P2Y2 and integrin αV inte